Menu
Image

Storytelling and AI

A Conversation with Morteza Behrooz

We’re clearly at risk of over-saturation on the topic of AI. But I think there’s still more room for understanding human-AI interactions on many more levels. One of those is storytelling.

This post is a summary of a conversation I recently had with my friend and colleague, Morteza Behrooz, about exactly that: the intersection of storytelling and AI. In it, we explore:

  • Evolution of Story Generation
  • The Mechanisms of Story ‘Interestingness‘
  • Applying Evolutionary Storytelling to Digital Environments
  • Narrative as Fundamental to Human Cognition
  • Storytelling as a Memory and Values Retention Tool
  • AI and the Return to Interactive Folk Narrative
  • Storytelling as an Inherently Interactive
  • Co-Creation and Democratization of Narrative
  • The Concept of AI “Slop” and Authenticity
  • Core Components of Effective Storytelling
  • The Future of Storytelling
  • And more…

The full transcript is posted below the video. All opinions are our own.

Mar 20, 2026

Transcript

00:00:00

 
Morteza: Yeah.
Jay Hasbrouck: So I know you and I decided we want to sit down and talk about this a little bit and explore some of it from different dimensions and think about where is their potential, where is there where are there possible, you know, sort of danger zones. Um and really think also about how is it how does it how does AI and storytelling come together at this particular time in history as well? Thinking about like the evolution of different forms of storytelling and where these things come together. So anyway, here we are having this conversation. Um, thank you for agreeing to join and do
Morteza: Thank you. Thank you for the opportunity.
Jay Hasbrouck: this.
Morteza: It’s a it’s an evergreen topic. Uh storytelling, not AI, I guess. Uh maybe AI from this point forward. Who knows? But in human life. Um but yeah, always passionate about that topic and always love talking to
Jay Hasbrouck: Yeah. Yeah.
Morteza: you.
Jay Hasbrouck: So, I wanted to actually just start the conversation by saying by um going into a little bit about how we met because um we were both um at Meta at the time.
 
 

00:01:28

 
Jay Hasbrouck: I I’d been there a little bit and then I started hearing about this person who joined who had explored storytelling and I’m like but from an angle that was very very interesting uh particularly around thinking about storytelling in well I but I’ll let you explain some of your dissertation work but um but I really wanted to I so I reached out I’m like I got to know this guy um and I’m glad I did because you definitely are approaching it and adding make made my understanding of storytelling a richer place particularly in terms of HCI and what that you know what are the dynamics there going on in in human computer interaction. So I’d love to hear your perspective of like you know like
Morteza: Yeah, absolutely. And uh the top perspective is that I joined MEA and I was looking for good collaborators
Jay Hasbrouck: that
Morteza: and an awesome mentor and friend and and I found you too. So, so thanks for always uh you know chatting about the topics and brainstorming. Storytelling has been on my mind since I was in grad school in different angles.
 
 

00:02:30

 
Morteza: Um the very big differentiator compared to now is that there was no LLM back then or at least fine-tune LLMs to reliably generate stories the way LLMs do today. So it was a very different environment in natural language processing NLP to do research on story generation. Uh traditionally in computer science though story generation is not a new topic. Uh people have used very different methods computer science algorithmic methods to generate language that could be presented as a narrative. Um this is often done in different layers. So there is story events that someone wants to represent in any way. Could be a function call in a computer program could be a different syntax altogether. Um and that’s the events that happen in a story that form the basis of the story. Narutologist I think called also fabula if I remember correctly. And then there’s the narration with language on top that has a lot of nuances itself. Um and that’s of course what LLMs have. They go through both layers but the the the narration it’s so easy uh these days at least. And different methods have been used before um you know from reasoning case-based reasoning to analogical reasoning to uh many other classical planning uh million other classical computer science approaches.
 
 

00:03:47

 
Morteza: And in that environment what really interested me was what makes a story interesting uh to be meta about it. And um that of course took me to a long rabbit hole of cognitive science uh just reading uh wasn’t my major and understanding psychology of it down to like 60s and 70s you know uh research on uh what just triggers the mind the right way to be hooked to a story. Um the categories of cognitive interest as they call it or experiential interest as they call it. one more this is a dangerous word but objective and the other one more a little bit more subjective um and and yeah it was really nice and then I tried to combine it with at least methods at the time to how to capture or detect or generate it um I think it still applies today with much more uh you know modern methods if someone wanted to do that kind of line of research in a better more complete way with LLMs of course it has to be done um um very differently. Uh but the basis was trying to understand the mechanisms that generate the perception of interestingness in stories as we listen to it and trying to reliably predictably recreate them with generation methods of story.
 
 

00:05:06

 
Morteza: Whatever the generation method
Jay Hasbrouck: Yeah.
Morteza: is,
Jay Hasbrouck: Yeah. I loved it and it really helped. At the time I was working on like how do we help people tell better stories in a digital environment particularly on social media of course. Um so I was at that point approaching and still do from an evolutionary perspective and thinking about like how do people learn to tell stories and what what are the things that translate well to the digital environment and and interaction with devices like the phone of course. Um, and thinking about like in particular what are the different components of storytelling that are that are formed when we’re children basically and how much of that carries over. So thinking about partial narratives, right? Or think about primitive narratives or um sometimes called focus chains, you know, pieces pieces of storytelling that are that we do actually every day, but we really don’t process them as stories because they really are just like subcomponents, right? and they they’re part of how we put concepts together. Um, and they stick with us.
 
 

00:06:06

 
Jay Hasbrouck: We learned them when we were three and we’re still using them. And then like so I was thinking about how are those um how could those be leveraged in an environment where we’re compiling or putting together some kind of small expression um expression of narrative in particular because it’s not like we’re just like and if you think about it in social media context it’s not like we were uh you don’t you don’t just put up one kernel of content. It’s always in some context and in some subcontext you know and and in some ways it’s referencing many other things. So I that’s one of that’s where I think a lot of our work intersected um while we were
Morteza: Yeah. Yeah. Absolutely. And I think it’s a for me a little bit during my school and definitely more talking to you,
Jay Hasbrouck: there.
Morteza: it was a realization that narrative is kind of everything. It’s it’s for humans. And I mean that in in kind of a literal way just because human cognition almost cannot help but perceive things as a in a narrated way uh to get to the really bare bones of it.
 
 

00:07:10

 
Morteza: And because of the nature of how we’ve evolved through time, our evolved brain has never known anything but to be through time. And events have always been memorized through time and been retold through language through time. And and that creates the very ingredients of nar narration. And like there is evolutionary forces for why um I I dig into them a little bit when I was in school for why that has been a good force evolutionary force for humans. Um for example, having a simulation kind of playground for us is really useful. um thinking and imagining what will happen if I go you know across the hill or across that kind of cave for like early humans um and being able to narrate that event or that imagination is a useful thing to be able to do evolutionary and that acts as a selection factor. So there is even reason to believe that that is something that literally you know was evolved within us and then down to and we see the facets of that and we see the effects of that in all the facets of our life today.
 
 

00:08:14

 
Morteza: It’s in how we communicate how social media works how things like branding and advertisement work, how uh our politics works. Um you know it’s always has to be narrative. Uh in the most dire situation, most happy situation, people care about the narrative that they’re showing. And because we all share that kind of bias, we get numb to it and we don’t see it. But if some if some alien came up, they could very well call this a planet of stories and that wouldn’t be too
Jay Hasbrouck: Totally. Totally.
Morteza: wrong.
Jay Hasbrouck: And there’s like plenty of evidence just to build on that that metaphor and narrative also are um really really effective devices for memory and memory retention. So this it’s not surprising at all that stories are the way we pass down information from generation to generation, right? Like as you mentioned, you know, like the you could imagine it dating back to like, you know, what happen if I go out and fight the lion, you know, like that becomes a story and those that story is filled with tips on how to fight the lion, right?
 
 

00:09:10

 
Jay Hasbrouck: And so like it’s it it’s um we have anchored around that in a lot of ways too.
Morteza: No.
Jay Hasbrouck: There’s there there are a ton of people um Mary Katherine Bateson uh Margaret me’s daughter talks a little bit about the value of metaphor in that sense and how it’s really really um a glue in a lot of ways for attaching messages and attaching you know like values all these other things that we that we prioritize as humans. Um, yeah. So I I wanted to to like make sure that we touched a little bit on like some of
Morteza: Yeah.
Jay Hasbrouck: where we are coming from as we start thinking about AI and storytelling because I think when people think about AI right now and they think about the generative tools that are available oftentimes they think about madeup s*** you know like it’s sort of like it’s sort of fantasy on the to the extreme and it’s really very bubble gum and it’s like it’s not really well I suppose you could imagine it’s it’s very primitive narrative but in many cases it’s really just fantasy or whatever.
 
 

00:10:07

 
Jay Hasbrouck: It’s not uh it’s not we’re not attaching some of the same values or at least what we see sort of in the pop sense,
Morteza: Yep.
Jay Hasbrouck: but there’s much more potential to go beyond that. And one of the things that I think about when I think about AI and storytelling is this idea of folk narrative um and thinking about like this this the long history um of you know this ability to to um pass on information to express values um through through a form of entertainment. So for example, there are there’s lots of evidence that in the in the past that we would enact, you know, as a species would enact a lot of like, you know, this is where plays come from, right? Like the form of storytelling where we’re actually embodying um some of what it is that we want to pass on in terms of narrative. So like parlor tricks in the past, right? Like in the 1800s, politics were really big and these were forms of storytelling that everyone not everyone many people I should say would participate in as like an active way of um of of engaging in in a form of entertainment.
 
 

00:11:13

 
Jay Hasbrouck: Um, and when you think about AI and the scale and digital scale, right, and this this this sort of uh the virality, the the potential for vir virality, I feel like there’s some interesting intersection there between like if you think about the poly tricks is sort of like stories anyone could tell, you know, sort of like if anyone could jump in at any moment and and play with these ideas in the form of
Morteza: Yeah.
Jay Hasbrouck: of uh in the form of a story. We are kind of in that same space here,
Morteza: Yeah.
Jay Hasbrouck: right, with AI like as these tools become increasingly more available, they become also they go from toys in some ways to tools. Um, so I’m I’m curious about your thoughts in that space and particularly because you’re probably dealing with
Morteza: Yeah.
Jay Hasbrouck: this on a regular basis.
Morteza: Yeah. There’s a there’s a picture that comes to my mind.
Jay Hasbrouck: Um,
Morteza: There was like a storyteller. I I’m not sure where I saw it. I think it was a book. I’m going to mention a book.
 
 

00:12:14

 
Morteza: It it might not have been that. It was the storytelling animal. There’s a I forgot the author’s name that I really enjoyed, but I somewhere on credit credit is due that the picture was a brain of the storyteller and a brain of the story listener and they were like connected with all these kind of like arrows that try to more illustrate a point than to create like a framework or you know explain a exact framework of cognitive kind of like or or communication um that is happening. But it was more illustrating the point that storytelling even when one person is primarily telling the story and another person is primarily listening to the story is still a very inherent kind of inherently interactive activity in that the storyteller has this perception of the listener and particularly the perception of the cognitive mechanisms that will you know tickle them the right way, make it interesting. theory of mind, right? Like what they’re tracking, what they know, what they don’t know, what would be interesting, how to tell the story, maybe augment what they know, maybe maybe play onto what exactly they know so that like you incorporate what they know a little bit better or what they care about to make it interesting.
 
 

00:13:24

 
Morteza: Uh, in fact, one of the experiential interests that I mention is personal relevance that um Roger Shank, I believe, like mentioned in 70s or 80s in like a paper about like what makes a story interesting. That was one of the papers that did that. and uh the teller of the story to the extent that they know a general class of audience or a particular person of the audience they would try to incorporate that into making the story be perceived as interesting. In fact, being able to use the skills is the basis um for what makes a good storyteller in our societies. We have George RR Martin who’s a really good storyteller and he knows very well explicitly or implicitly as a skill how to tickle the human brain exactly the right way when you tell fiction in this case. Um and I think that inter so storytelling is a really interactive inherently interactive thing as I learned and I continue to be amazed to learn new new aspects of it uh um every day and and that is carries through to society. I think it’s been an interactive exercise for generations.
 
 

00:14:30

 
Morteza: I think um to tell stories to each other. It becomes a vehicle of communication of values through time as you mentioned elegantly. It is the currency of human communication and human culture kind of like you know um perpagating through time peragating through society. Um and that interactivity is kind of has been the norm. And what it’s interesting you pointed out the previous formats of entertainment the ones that we’ve maybe grown up with are familiar with these days is movies and TV shows which is like because it’s async is not interactive and in the whole in the grand scheme of storytelling
Jay Hasbrouck: Yeah.
Morteza: that’s actually the exception is it is not normal in the human history to have this
Jay Hasbrouck: Yeah.
Morteza: async very one uh you know kind of disconnected in the other direction kind storytelling and one way to think of AI maybe and I’m not sure how exactly that should be framed or will work out but it’s is that that that exception is ending
Jay Hasbrouck: Totally to exactly I wanted that that’s what you made me think of like Hollywood and the professionalization of storytelling is completely an anomaly right it’s for much much further back it’s been democratized right it’s been something and I when you were you were talking about the the relationship between storyteller and listener um one of the things that comes up again and again when I was digging into that especially trying to understand what that interaction looks like.
 
 

00:16:00

 
Jay Hasbrouck: Um, you see coming from organizations like the Moth or some of these storytelling clubs and a lot of what they encourage is experimentation, right? They want folks and part of what the storyteller does is every time they tell the story and this happens almost without even trying. Every time they tell the story, it changes a little bit the pace based on how what the response looks like from the audience, right? So,
Morteza: Yeah.
Jay Hasbrouck: if you think about experimentation being a critical part or this sort of live experimentation, right? And and there’s a lot of evidence on the neurological level that people’s minds are sinking with the storyteller as they’re listening, right? They’re experiencing because they’re mirroring emotions in many ways. Um so if you think about that form of experimentation within that interaction, that’s very much true of like the parlor tricks and the you know this all this kind of entertainment that was we’ll call it folk entertainment. Um not so true with Hollywood, right? but then becoming more true as people start to think about what are the other tools I might use to tell a story with AI, right?
 
 

00:16:58

 
Jay Hasbrouck: Like how can I imagine a new landscape, a new set of a new tale um and make it, you know, or at least uh open the door for more interactive exchange. I’m curious how you guys are thinking about that too because like or how you are thinking about it but I I know also that relates to what you’re doing um because like it’s not still not the same as a persontoperson interaction or it could be I suppose if the if AI is part of that is part of the tooling but if if part of the interaction is also
Morteza: Yeah.
Jay Hasbrouck: AI to human you know like are is it really just a mirror I don’t know I mean like it
Morteza: Yeah. Yeah.
Jay Hasbrouck: It’s not really a mirror,
Morteza: Um Yeah.
Jay Hasbrouck: I guess.
Morteza: It’s not a maybe, you know, the same mirror. It’s a new mirror. It’s easy to or maybe even the first reaction um to call it like a fake mirror. It’s not the same thing. AI could definitely be deployed and used in a way that is not uh you know reminiscent of the parlor parlor tricks and or collective storytelling that we’ve had and you know case in point a lot of products out there are not made for storytelling and you know Chad GBD for example has limitations it can do a lot and and um it’s interesting to think about like what principles of interactive storytelling people have enjoyed and how to how to carry that forward.
 
 

00:18:27

 
Morteza: Um I think one of the things that is interesting and is playing out a little bit in in different platforms is uh the fact that we still have creators and writers telling stories, right? Like we still have that uh kind of like urge to tell a story and it’s still not live as in like one person does not do tell a story to another to a group of audience. Um but that relationship between creator AI and the story consumer is kind of like changing and it’s maybe one or a few steps to towards that like democratization. It’s not quite like you know everyone will you know be able to participate at the same time you know facilitated by AI. I’m sure different forms of that will emerge. um we don’t know but I think uh when I did a study on um 30 something uh writers at the very uh beginning in the very beginning of this was during like the uh writers like kind of striking and being unhappy with how AI was was evolving um very rightly so very right concerns that they had and um you know I just wanted to learn from their concerns and how what could work um if anything and I and what what I observed was during in the course of one session uh of a you know interview session.
 
 

00:19:50

 
Morteza: Sometimes their views change just because of having a chance to having a nuanced conversation about some of the tools, understanding like thinking um outside of outside of like you know what’s presented um uh in the society about like what could offer what this tool can offer to them. Um, and that as that relationship evolves, I think AI could also like, you know, evolve the relationship uh between the creator and the listener and make that interactive kind of like nature a full loop that we could all
Jay Hasbrouck: Yeah,
Morteza: enjoy.
Jay Hasbrouck: I I’m curious what your thoughts are about like just try trying to think through the directionality, right? So like if I’m interacting with an app that’s whatever, let’s say character AI and I’m and I’m you know I let’s say I’m the
Morteza: Yeah.
Jay Hasbrouck: storyteller and I’m getting a response, right? um what is you know like what are the dynamics of storytelling at play there when it does kind of feel like
Morteza: Yeah.
Jay Hasbrouck: it’s a a mirror or whatever or take the opposite direction let’s fast forward 10 years and say imagine a world where you’ve got an AI storyteller and you’ve got an audience right and the AI storyteller is reached a level of sophistication where it can respond to an audience and does adapt its storytelling and change based on the dynamics of that interaction.
 
 

00:21:09

 
Jay Hasbrouck: So, you know, like how how do you describe the the dynamics at play there? You know,
Morteza: Yeah.
Jay Hasbrouck: I’m curious if you’ve thought through any of
Morteza: Yeah.
Jay Hasbrouck: those.
Morteza: I mean um on character AI we do have uh you know as I was mentioning before kind of similar to that a new dynamic that has emerged since the company you know the service was created uh which there is like you know role playing happening between the user and the AI character but the character is also created by a creator and that kind of relationship kind of like exists. It’s not just the AI, it’s um it’s the it’s the creator that people have a mentality and perception of when they role play with the character. Um and maybe that’s a necessary mechanism in the democratization to get the scale. Maybe new things will emerge as the industry evolves. I don’t know. But but that relationship of being able to like role play with a character and tell a story with it um which I’ll double click on is a very um I think uh unique way to still maintain the relationship with the creator and the storyteller.
 
 

00:22:18

 
Morteza: And I think as we all in the in industry like build tools for expression for storytelling I think I I totally believe that that will evolve into being something that goes more towards democratization as you were saying and role playing right now is
Jay Hasbrouck: So
Morteza: kind of the first I guess like major use case of LLMs in that domain that exists. It’s uh where the user uh kind of like takes the story forward and the AI continues and people have been doing role playing with each other for a long time and now people can do it with AI. Um
Jay Hasbrouck: that’s more like a co-creation, right?
Morteza: yeah it’s collaborative
Jay Hasbrouck: That’s like you’re co-creating the narrative with the AI in that sense, right?
Morteza: storytelling.
Jay Hasbrouck: Yeah. Yeah.
Morteza: It’s and maybe quick distinction,
Jay Hasbrouck: Yeah.
Morteza: there is collaborative storytelling where like you’re just telling the story to enjoy it, to tell it, to create it and enjoy it at the same time. And then there is like interactive narrative games that have existed for a long time where like the goal is to like finish the game and win it.
 
 

00:23:18

 
Morteza: Uh maybe there’s like an explicit or implicit objective. Um um but that’s like done through text and like inherently like a narrative. These are a little bit different and in fact there are specific conferences for each. So like there’s a conference on interactive narrative technology and then there’s a conference for collaborative storytelling I kits uh both of them near and dear to my heart but but these are like inherently different experiences from a UX point of view and I guess what we’re talking about now is like a more collaborative way which you know the user is also part of telling the story they can they can tell the story they want and the creator’s creation has been mediated by this like AI technology but it’s still like the creator’s intent is reflected there so there is
Jay Hasbrouck: Mhm.
Morteza: this like give and take uh or collaboration that is happening both kind of like synchronous and asynchronous and AI has a role there your creator has a role there the user has a role there and I think that is a domain that will evolve to get us like more chores like the direction we’re pointing
 
 

00:24:18

 
Jay Hasbrouck: Yeah. Yeah. Interesting. So, and then what about the the other scenario? Like if we were in the sort of future facing scenario, I suppose if you’ve got the Well, maybe it’s it’s all I mean, you could argue that stories are always collaborative and co-created, right? like an effective story is, but um the idea of an AI leading a story, you know, and and facilitating the narrative formation. Um we’re not there, but like are have you have you put any thought into that and what that might look like?
Morteza: I think um and this goes back to like the research I did uh in school a little bit and which is a fancy way of like the stuff I learned halfway from cognitive science uh because I never studied it formally but I’m a I’m a fan. Um I think stories are like maybe many other forms of art or um or just expression. uh it really matters to people what they say about the world, about a certain population, about themselves, right? Like I think I have this hypothesis that the reason why there is this like AI slop negative reaction people get on Instagram or some other kind of social media, Tik Tok or what have you on like seeing like a fake bird.
 
 

00:25:35

 
Morteza: I mean if you look at the comments and do sentiment analysis on it is like very negative always once people figure out that it’s AI, right? And I think the reason isn’t that the aesthetics is bad. I think the reason isn’t that the bird is actually a bad looking fake bird. I think it’s that it feels like a lie. And the reason why it feels like a lie is probably because when people do see a real beautiful bird, it tells them something about the world that they’re also living in. And by proxy, maybe tells them something about who they are or they could be emotionally, right? Like it’s if I live in this world and this beautiful bird exists, my mind opens to new possibilities in a way, right? Like my world view changes in a very subtle inside mind butterfly effect way, right? Like and that is something that you will be taking away from them once they realize it’s actually not real and that feels like a lie and that negative rush of sentiment exists. With fiction, that’s different, right? And that’s the that’s the opportunity I think that for creative work work where like fiction uh once is you know presented as fiction it has an established paradigm and it has had throughout human history um for communication and enjoying and and I think that could be an interesting opportunity to double click on um how people could enjoy fiction uh more interactively collaboratively together and um and for it to not be seen as as AI
 
 

00:27:07

 
Jay Hasbrouck: Yeah.
Morteza: slob.
Jay Hasbrouck: Yeah. I I hear you. It’s like where is that I I wonder if it does come down to sort of like what is the initial spark for the story in some ways, right? Where is it the origination of it and what is it inspired, you know, what’s inspiring it. Um and certainly at least at this point AI is not capable of that. I mean, you could imagine a scenario where you’ve got an audience and you have the AI storyteller, whatever it is, a robot, um, and the audience is like, you know, shouting out, tell us a story about blah blah blah, you know, and then it takes that and runs with it. Spark comes from the audience,
Morteza: Yeah.
Jay Hasbrouck: I suppose. I don’t know. Just trying to think through like, you know,
Morteza: Yeah.
Jay Hasbrouck: possible
Morteza: I think the Yeah.
Jay Hasbrouck: flipping
Morteza: I think I don’t know the the honest answer is I don’t know. I think the principal components of that caring mechanism that like I the bird example that I said which is my hypothesis probably some psychologist is saying like there’s a book already on this that you don’t know I’m sure that I would love to read and learn but but that uh mechanism I think the principal components of that for storytelling um have to be kind of taxonomized and understood.
 
 

00:28:16

 
Morteza: So for example talking about the sh the lived experience of someone right like talking about something that uh is feels like a real experience someone has had people even in fiction I guess the plausibility of that in some way maybe matters u you know the world of Harry Potter isn’t real but the feelings that Harry had probably are real and relatable right like because you could imagine yourself in their shoes somehow uh I always and um and that I think is an interesting kind of maybe way to create a framework for what people care in a story and how it could be originated because under that hypothesis I don’t think it’s just the fidelity of the you know planting the seed of the story I think it’s more about these principal components that are experiential that people care about it having come from a real person or it being a relatable thing that you can put yourself in those shoes And um and I think that’s at least part of what they care about. Now there I should say there are like um maybe uh devil’s advocates or exceptions. One of them is comedy.
 
 

00:29:26

 
Morteza: Uh there’s research on comedy and like if it’s if something makes you laugh, it makes you laugh maybe you don’t care that much.
Jay Hasbrouck: H that’s interesting.
Morteza: It’s interesting to create a taxonomy of like you know what creates the perception of related human
Jay Hasbrouck: Yeah.
Morteza: relatedness I guess or something like that and how people can augment that with AI or delegated to AI or maybe we some of it we never
Jay Hasbrouck: That’s fascinating. I mean, think about comedy,
Morteza: can
Jay Hasbrouck: right? So, in some ways, the bar is slightly lower for comedy, but when you talk to writers, they always say comedy is the hardest thing to write,
Morteza: probably Yeah.
Jay Hasbrouck: you know,
Morteza: Yeah.
Jay Hasbrouck: because they they they talk about you have to be able to,
Morteza: Yeah.
Jay Hasbrouck: you know, really really craft the conflict to get the to generate the humor.
Morteza: Yeah.
Jay Hasbrouck: Um, yeah.
Morteza: And comedy always has like at least some parts of comedy has like a social layer that is harder to communicate, hard harder to get right. There’s dark comedy.
 
 

00:30:15

 
Morteza: So, uh I guess I had in mind like, you know, funny reels on Instagram and Tik Toks, but even for those there’s usually that. So, yeah, I uh I I agree.
Jay Hasbrouck: Yeah. Yeah. I I think some of what you’re getting at too comes back we in the one of the um studies that I led at Meta, we really kind of we decided we what are the kind of kernels that translate best for for storytelling in the in the digital space. And maybe there’s some of this is exposing some of what you’re saying about like the lie um and maybe that like gets in the way of resonating between AI and humans. um when it comes to narrative and we we boil it down to emotional connection, authenticity and interaction, right? So it has to feel interactive. It can’t just be a static um to be effective storytelling, right? I mean like like you said like I I think there’s a version of storytelling like mo movies can trigger very distinct emotions, right? But it’s not an interactive process.
 
 

00:31:09

 
Morteza: Yeah.
Jay Hasbrouck: You’re not contributing to that experience other than you bought the ticket,
Morteza: Yeah.
Jay Hasbrouck: right? Um so I think that you know that’s there’s that’s a distinction worth making you know in the digital space interaction can easily be facilitated right the interactions between people between you and your friends and
Morteza: Yeah.
Jay Hasbrouck: and it’s mediated through the platform um but authenticity is a big one right and and emotional connection are really big like those two are huge and I think we have humans
Morteza: Yeah.
Jay Hasbrouck: in general have a very very keen sense of what’s coming from an authentic place and what’s not. I think it’s I think we’ve we are we can hone in very very quickly on that.
Morteza: Yeah.
Jay Hasbrouck: Um and the same with the most emotional connection too.
Morteza: Yeah.
Jay Hasbrouck: So I mean there may be components I mean I suppose you could mirror emotional connection and get closer.
Morteza: Yeah.
Jay Hasbrouck: Um I was just thinking about like what are the if you were to like create some bars here on like what is easier and what is less easily accomplished between the human AI interaction interaction.
 
 

00:32:09

 
Jay Hasbrouck: Yeah, you could probably get there,
Morteza: Yeah.
Jay Hasbrouck: but authenticity really difficult. And then, you know, like I don’t know, emotional connection. I’d be curious if what your thoughts are on how those three
Morteza: Yeah.
Jay Hasbrouck: land.
Morteza: I mean, it’s hard to hard to know honestly um without experimentation, right?
Jay Hasbrouck: Yeah.
Morteza: Like it’s not it’s it’s easy to to think of a framework um that could
Jay Hasbrouck: Yeah.
Morteza: like give us the direction. I think gen definitely they sound true uh you know given observations I’ve had and studies I’ve um I’ve done since school and um but I think emotional connection is always like kind of with a character right like with a protagonist with someone else and one of the actually shared um experiential interests that was like taxonomized in the work I did and other researchers uh that I learned from um some of it is just like um story tropes, you know, something as simple as that. And tropes are always looked down on and like you know the underdog and um and stories like that that are like even you know sometimes romance originated or sometimes like just like you know themes of life I guess some people call them in in some papers and the reason why they’re they’re relatable is that they’re just situations we all find ourselves in right
 
 

00:33:30

 
Jay Hasbrouck: Yeah.
Morteza: like there’s situations of paradnamics of conflict of emotional connections in all different like you know uh balances of it um that we kind of find ourselves in and we all relate to and I think a lot the reason why tropes are popular is that they work and they have Russian forkl stories old Persian stories uh you know other um old uh you know parlor tricks from from from time before probably have connections to them and and I think like it’s interesting to study that um to I think there’s more more there more more to understand about like how they create the emotional responses they do even now that people are like you know they’re so explicit that in some websites you can see like drop down menus of them like listed like they explicitly in the UI are like list in the tropes and people still go click on them and they like willingly go and you know um explore them um and that’s I think one area of like maybe trying to understand more um yeah what do you
Jay Hasbrouck: I I like where you’re going with that.
 
 

00:34:37

 
Jay Hasbrouck: I feel like yeah, the emotional connection side could be pulled apart much more and thinking about yeah, its components and what are leverable or whatever if assuming that’s assuming we want to go in that direction. Um but yeah so maybe it maybe it is true like the authenticity part is really like what it you know like that is the highest bar you know like that is the point at which people become the most discerning in terms of the kind of narrative they are contributing to or consuming.
Morteza: Yeah, authenticity is interesting.
Jay Hasbrouck: Yeah.
Morteza: Authenticity is almost like the result, not the cause, right? Authenticity as a feeling. And like I I’m just like, you know, um thinking live here. But um but what generates authenticity in this or perception of authenticity in this domain is interesting, right? Like because there’s a there’s analoges to that where someone looks at art just even abstract art, right? like and they they perceive some level of a different level of authenticity maybe if it’s made by a person versus not and like as a proxy in fact I did some research on um perception of interestingness in uh generated um shapes moving around.
 
 

00:35:51


Morteza: There is a very famous 1940s studies by Haidider and Siml to butchering names always that you just you know put a triangle and a rectangle like and and a triangle and a circle and and a square moving around inside a rectangle that had a door on YouTube. It’s easy to find and and just show it to people. I believe that study was particularly study of apparent behavior like perceived behavior but perceived behavior through time is a narrative. So people in the study started like the triangle cheated on the re rectangle and and and the other one is mad and like the researchers like what the hell is happening? We just showed you some shapes. Of course they had the hypothesis to some extent but the results are astonishing to the depth of how much people the level of nuance. Uh I did some research on that too. There is more better research done by Andrew Gordon um and his lab at USC from
Jay Hasbrouck: Yeah.
Morteza: some time ago also using those shapes for perception of narrative and interestingness in narrative and we could use that for authenticity too.
 
 

00:36:50

 
Morteza: So like if uh maybe shape those shapes and that framework and maybe something else. But I’m very interested in like if you look at an abstract art that a famous artist or a non-famous artist or maybe different classes of users and also an AI has made how people you know what are the what are the fundamentals of generating authenticity.
Jay Hasbrouck: Yeah. Yeah. Totally.
Morteza: I think I think
Jay Hasbrouck: I I used to work with a designer um she was awesome and she had a character in her mind for every this is how she learned the numbers. She had a character for every number and they all had personalities and she had completely like this world of one through 10 each of them and she she tell you about like who they were as people and she completely had assigned you know like she she built an authentic character around each each of the numbers. Um was really entertaining and fun.
Morteza: And I guess like anthropomorphism, one could say, you know, people see personalities or something in like clouds or shapes and, you know, we actively anthropomorphize thing. So maybe that has a that has a vector there
Jay Hasbrouck: Yeah. Yeah. Well, I it’s been a pleasure.
Morteza: too.
Jay Hasbrouck: We could probably go on for hours, Martasa, but I don’t know if you had any final thoughts you wanted to add to our conversation
Morteza: Uh no,
Jay Hasbrouck: here.
Morteza: I think I think well first of all thank you for for the chance to talk as um I’m lucky enough to to often do um chat with you Jay and and yeah, storytelling is I think the most underrated area of AI. Um I think it’s a it’s a very I guess fundamental tendency of humans in entertainment in other domains of life and
Jay Hasbrouck: Yeah,
Morteza: um how to understand it and replicate it and advance it really as a paradigm. It’s it’s really
Jay Hasbrouck: we’re at an exciting time.
Morteza: interesting.
Jay Hasbrouck: But thanks thanks for your time, Mortasa. I’m sure we’ll be talking again soon. Um, and I’ll end the recording. I think I will.
 
 

Transcription ended after 00:38:53

This editable transcript was computer generated and might contain errors.

Image

Robots: It’s Complicated

What do human-robot interactions tell us about our ourselves?

Let’s roll in the video cart for this post. We’ll start with this clip:

The official narrative that runs through this video is that robots are increasingly lifelike and capable of amplifying, or adding flourishes to, qualities we admire in living form: in this case ‘talent,’ but also efficiency, intelligence, agility, etc. It’s a story in which we are primed to marvel at their accomplishments and applaud their ever-amazing improvements. And we do — just look at the audience in this clip.

But look closer. The most critical part of the story in this video focuses our attention not on amazement with robotic antics, but on empathy, when one of the robots collapses. Do you really think this fail was accidental? It’s more likely a narrative device used to generate relatability and overcome the creepiness we humans sense when robots become ‘too perfect’ or too similar to living beings. What better way to enlist this device than through performative vulnerability, and Boston Dynamics serves it up steaming hot here.

Artists Sun Yuan and Peng Yu also leverage the device well in their piece titled Can’t Help Myself that was on display at the Guggenheim Museum in New York City. The work consisted of a glass encased giant robotic arm that was programmed to perform 32 unique dance moves, such as “ass shake”, “scratch an itch”, “jazz hands,” and “bow and shake,” to entertain and anthropomorphize the sculpture. However, it was also programmed to sweep up a dark red fluid that continually leaked from its inner core. As time went on the leak became larger and harder for the arm to manage, which shifted the behavior of the robot from joyful dancing to the punishing working of capturing and retaining its vital fluids. Eventually, the artists decided to turn the robot off.

Trigger alert: it’s sad, really sad. Seriously, the first time I saw this, it hit deep.

Let’s take a little diversion to think about what’s happening with audiences in both of these examples. We’ll start by borrowing from the Japanese concept of Wabi Sabi. It centers around an appreciation for the beauty found in imperfection. In practice, it highlights the aesthetic value of authenticity, including the passage of time, aging, and even decay. The ‘wabi’ half of the concept is expressed through rustic simplicity and understated elegance. The ‘sabi’ half refers to the beauty that comes with age and wear; and it celebrates this through appreciation for patinas, imperfections, and the stories they provoke.

Wabi Sabi can be seen as a deeply-rooted substructure for what appears to be happening in audiences that become emotionally engaged when a robot’s imperfections are surfaced. The Pratfall phenomenon from social psychology runs parallel to this. It suggests that a person’s perceived attractiveness increases after they make a mistake—provided they are generally competent. In addition to the ‘injured’ robots in the clips above, we see this phenomenon leveraged in widely popular works like Wall-E (we feel sympathy for him in his futile struggle to clean up infinite piles of garbage on a post-apocalyptic earth), or in the novel Hum (when humans feel sympathy for robots after witnessing them grow exhausted from perpetually delivering ads to humans). Their vulnerabilities and imperfections are designed to evoke an endearing audience response — and it seems to work (maybe too well).

When Things Get Personal

While narratives like the ones we see above can be very effective when robots are observed from a distance (behind glass, on a stage, in a movie or novel), they are at risk of disintegrating when robot-human interactions become more intimate. In those settings, where humans shift from audience to participants or cohabitants, we start to see forms of human resistance, and even destruction, as both the seamlessness of embodied technology and the manufactured vulnerability of its ‘imperfections,’ shift from human fascination to ruptures in our understanding of consciousness, and ourselves. This enchanting machine that was once surprising and charming is now ‘all up in my business.’ So, we resist; but our resistance is layered…and very telling. Let’s take a look at a few examples.

You may remember the hitchhiking robot from 2013. It was designed to be temporarily adopted as a traveling companion, and passed along, as part of an experiment to learn about how people interact with technology. From Wikipedia: 

It was small and had a look the team described as "yard-sale chic," to evoke trust and empathy, and had a child's car seat base to be easily and safely transportable.

The robot could not walk – it completed its "hitchhiking" journeys by "asking" to be carried by those who picked it up. The robot could engage in basic conversations, discuss facts, and function as a robotic companion during travels in the vehicle of the driver who picked it up.

[and then...]In 2015, its attempt to hitchhike across the United States ended when it was stripped, dismembered, and decapitated in Philadelphia.

In this CBS story covering the robot’s demise, one of the creators is quoted saying that she “was most concerned about children who loved hitchBOT and followed it on social media. Her team doesn’t plan to release the last photo of it to protect young fans who might be distraught.” (!)

Let’s move on to some more mundane examples to dive deeper into the contrast between the ‘official’ narrative of robotics and everyday human interactions with them. Take a moment to watch these robots folding laundry or peeling an orange.

How easy is it to imagine these robots performing these tasks in your home? And how comfortable is it to imagine them taking up the space equivalent of a human in your home to do so? Or, maybe they perform these tasks while you sleep. A bit disconcerting, right?

Maybe the robotics just “isn’t there yet.” Or, we humans still need more time to acclimate to their presence. Both are likely true, and I’m sure there are highly capable researchers addressing both.

However, even if those obstacles are overcome, the subtext of human-robot interaction narratives position humans as passive receivers of the wonders robots bestow upon us. Yet, when we see them function in the real world — in OUR world (not the factory floor or the stage) — they tend to frustrate and/or alienate us. Their presumed efficiencies often contrast with our expectations, and confound more than they delight us.

Second, if we presume that our role is that of commander in a future populated with everyday robots, do they not become an embodied and constant reminder of how inefficient we humans are? If we’re being told that the highest value of executing any task is sleek, seamless, and serene efficiency, our flaws and idiosyncrasies are magnified by contrast. We’re none of those things, at least not consistently. Their very presence may seriously and continually challenge the value of our agency in the world. What’s to become of our pride in ‘mastery’ (regardless of how mundane it may be)?

Further, even if robots are pre-programmed to express performative vulnerability and concocted imperfections to win us over emotionally, might we not become even more resistant to their presence once these strategies are exposed?

Taken together these considerations are signals that more intimate interactions with robots are likely to trigger or amplify animus within us, which may in part explain some of the destructive behavior we’ve seen above. We resist their presence to highlight, or punctuate, the difference between robots and humans; to create greater distinction between the two embodied, yet conflicted, forces sharing common space and familiar tasks. Take a moment to check out the Flesh Fair from the movie AI:

It would seem likely that this form of resistance is more likely to increase than decrease as robotics that enlist narratives of perfection and performative vulnerability advance. Might we not eventually devise screening tests, harbor resentments, demand constraints, or…worse? This is classic ‘Othering,’ something that our species can’t seem to shake.

But resisting and Othering are only part of the equation. It’s HOW robots go about their work that further exacerbates our troubling relationship with them when in close proximity. A closer look at the video of a robot peeling an orange reveals that what is intended by the makers to function as a pair of hands, starts to cross very quickly into the uncanny as it appears that each hand has a ‘mind’ of its own – a sort of hydra beast dissecting something that has no inherent meaning to it.

So, the simple task of peeling an orange starts to reveal an alien approach to the task that becomes unsettling for humans to witness. I wouldn’t be surprised if most people watching this are seized by the desire to grab the orange from the robot and peel it themselves — not to do it faster or more efficiently, but to put an end to witnessing the weird ‘logics’ embedded in each hand of the robot. This might be even more evident in the video showing a robot unlock a tool box on that same page.

Which leads us to the question of which is more disconcerting: robots that apply alienating logics as they interact with the familiar tasks around us, or robots that perform familiar tasks in increasing rates of efficiency that outstrip our capabilities? Or, perhaps worst, interacting with domestic robots that feign imperfection to position themselves as endearing? (lmagine the robot that peels an orange dropping a slice and then timidly apologizing for its ‘mistake.’ Eww.)

Perfection and Domesticity

In light of the unsettling relationships we may be forming with robots, let’s take a look at how our responses have already begun to take shape in domestic applications of robots when the narrative of efficient perfection alone is centered.

From my own research into technology in the home across many cultures, I can attest to a clear and constant tension between the goals we set for orderly households and messy business of living. ‘Magical’ tech solutions have compelling value propositions that suit our aspirations, but our experiences with them are inevitably disappointing. Again and again, it quickly becomes evident that tech perfection and efficiency are at odds with domesticity.

Take, for example, the humble robot vacuum. It actually has quite a long history, but ask anyone to tell you a story about their experiences, and they will more than likely share a story of failure and frustration. Their expectations for efficient, automated, and pristine cleanliness has more often than not been dashed and replaced by the new responsibility of tending and rescuing their annoying robot.

Image
Image

This contrast is perhaps captured most poignantly by this photo, in which cleaning staff at a robot vacuum showroom is seen using a good ole broom to clean the floors:

Image

Which leads me to ask: If narratives of robotics inherently conflict with the realities of domesticity, in what ways might they align more appropriately with our daily lives? I would argue that alignment is more likely when:

  • the consequences are low (we’re not expecting perfection or waiting for a robot to execute a critical task);
  • efficiency isn’t the point (we’re not comparing our capabilities to the robot’s);
  • opportunities for playful personification originate with humans (not from robots’ performative vulnerability);
  • the robot’s logics have metaphoric parallels (the robot either offers no clear indication of its own logic, or provides signals that reflect logics with which humans identify).

So, in the case of robot vacuums, even if the sales pitch (perfectly automated cleanliness) never fully materializes, it likely doesn’t matter that much if the floor isn’t as spotless as promised (low stakes). And, besides, its mishaps are sometimes amusing. Further, its ‘logics’ are largely cartographic and familiar (like mowing a lawn). What’s more, we seem especially keen to morph them into play things — a very telling hack, even if it is still just a vacuum in the end.

The Future is…Cute

So, where is all this going? If manufacturers are savvy, they will listen carefully to the signals consumers are sending and double down on autonomous robotic toys. As products positioned more for entertainment than for executing chores, they’re simply a better ‘fit’ for us than the promises of efficiency, feigned imperfections, or uncanny logics, that have triggered human resistance, Othering, and resentment. Instead, bots should enable low-stakes ‘delight’ functions, deprioritize efficiency, transparently facilitate personification, and avoid the discomfort of the uncanny by leveraging familiar human metaphors (the pet, the confidante, the class clown, etc.).

There are a few early entrants in this space now — chief among them Sony’s robotic puppy named Aibo and LivingAI’s desktop ‘pet,’ Emo. However, both are far from the emotional, personality-packed attributes we might imagine from an autonomous Wall-E who hangs out with us at home, for example.

And what of those humanoid robotic servants like Tesla’s Optimus? Perhaps their presence is better suited for factory floors and warehouses. In short, know your place, bots — cuz, well, we’re not always an amicable species.

Image

‘Humanizing’ AI

It’s Time to Revisit Universals

(Opinions are my own.)

Preamble

Most children start to take objects apart to see how they work at a very early age. In the field of human development, this type of curiosity is referred to as building a “connection schema,” and typically occurs around the age of two or three. Later, beginning at around four years old, children begin to engage in constructive play, in which they build structures that represent real-world objects, such as roads, furniture, or houses. This is a time when they are developing a sense of control and accomplishment by manipulating their environment to create something new. Toward five years of age, play becomes more complex and cooperative. Here, children begin working with others to build larger, more intricate structures. They also start planning their creations, thinking about the design process and the purpose of the structure, before they start building. This helps them develop good problem-solving skills, as they figure out how to make their creations stable and use their imagination to create more complex, detailed structures. I’d like you to hold this in the back of your mind as you read the rest of this post.

A Brief Diversion into Theory

As many readers of this blog know, the poststructuralist mode in social science argues for the rejection of human universals and objective truths in favor of an emphasis on social constructs, within which we all navigate interactions and ‘intersections.’ Within this view, we live in an ongoing stream of cultural patterns; ‘always already’ being influenced by cultural processes, and in turn, shaping them through our behaviors and interactions. Some of these patterns clash with others, some converge, some spiral around one another, etc.

There are no foundations in this view — just perpetual flows of interactions that host cycles of approximate replication and/or simulacra. Language is one medium through which these patterns transmit, and through it’s deconstruction we learn that there are no ‘truths’ either, only patterns that some of us agree are useful to define our experiences in the world. In a word, the best we hope for is a sort of “truthiness” that varies depending on each person’s own ‘lived experiences,’ affiliations, alliances, identities, etc.

The task most poststructuralists / postmodernists / critical theorists assign themselves is to expose how people are ‘situated’ within these cultural flows, and to deconstruct the institutions or social structures we’ve created within those flows to expose power imbalances. Often, the goal is simply that — to take apart institutions to reveal their underlying workings as ‘oppression.’ Sometimes it’s more political (e.g., “dismantle the patriarchy”), but we’ll leave it there for now, because I want to move on to something more interesting.

LLM’s as the Poststructuralist’s Dream

There are some fascinating similarities between the poststructuralist position and the ways in which large language models (LLM’s) work. For starters, LLM’s don’t naturally assign value to our cultural institutions or social structures, primarily because they don’t integrate the full historical, folk, or biological history of hundreds of thousands of years of human evolution. Instead, they’re trained on ever-changing flows of digitized languages — be they written, visual, symbolic, spoken word, code, or other forms. Of course, they’re particularly adept at identifying patterns within these flows, and they do so at unprecedented scale — but these patterns are not indicative of core human values or operating principles. A good example is ‘common sense,’ a set of valuable anchors humans use everyday to navigate the world, and a phenomenon that largely escapes the grasp of LLM’s.

When we interact with LLM’s, our human queries are used to determine a position for identifying patterns across huge data sets. These models then generate an ‘answer’ in the form of ‘new’ content derived from the patterns they identify. They aren’t vested in any truths per se. Instead, their ‘truthiness’ is formed from an amalgamation of patterns. They present their outputs as facts because they were designed to present their findings that way. For them, all of their outputs are equally factual (a pattern or string of relationships shows up often enough to be ‘true’) — even when our real world experience may prove otherwise. This is why their absurd hallucinations are often delivered with such confidence; and it’s also why you hear some people argue that LLM hallucinations are actually a feature, not a bug.

So, here’s what I’d like to ask – If, from the postmodern perspective, our social constructs are all ever-changing amalgamations mediated through language, what does it matter if an LLM taps these patterns and ends up hallucinating? If it makes up quotes, references books that don’t actually exist, cites legal cases that never happened, or advises people to eat glue, aren’t these all just extrapolations of the pattern ‘logics’ within the data it taps? Aren’t all these outputs equally valid, in the same way that everyone’s ‘lived experience’ represents their ‘truth’ within the postmodern condition?

If all of this is so, it would seem to me that LLM’s are in some ways a poststructuralist’s dream; and indeed, we see evidence of fascination with their ‘truthiness’ among both AI-doomers and AI-utopians. From the doomers we hear about nightmares involving AI take-overs of our infrastructures, runaway fraud, and strings of falsehoods built on top of falsehoods driven by those savvy enough to ‘flood the zone’ with their own ‘truths.’ From the utopians, we see product peddlers hype our collective magical / mystical associations with technology, playing into the propensity for human fascination with tech itself — the wondrous tech wizardry that ‘knows’ you and sells pattern identification veiled as sage, magician, or prognosticator. (Did you know that when the telegraph first came onto the scene, people would hold seances around them?)

To techno-utopians in particular, we need to pose a few specific questions. First, if LLM’s postmodern truthiness is, in fact, a virtue, or at the very least, a legitimate reflection of the ‘truths’ these models encounter in their pattern-finding schema, why strive for accuracy, ‘responsible’ AI, or ‘alignment,’ at all? Don’t these attempts at ‘correction’ run counter to their deep faith in pattern identification across increasingly large models and data sets? Even if our utopian friends manage to hold together the contradictions of an appreciation for model truthiness and the prioritization of model responsibility, WHOSE version of responsibility are we talking about within this relativist view? WHOSE ethical standards will hold these models accountable (if that’s even possible)?

Cynics might argue that AI-utopians ultimately don’t care about whether these models are responsible, and even more, whether or not they risk dissolving our cultural institutions with a never-ending plethora of ‘truthiness’ / AI slop. They might also say that some AI-utopians could very well be quiet advocates for deconstructing our cultural institutions anyway (aligned with the postmodern assertion that we’ve falsely attributed power and importance to those institutions); and further, that LLM’s could be instrumental in tearing them down. So, a poststructuralist might argue that the way LLM’s work merely reflects the fluid truthiness of language, and that our assumptions about authenticity or the cultural cornerstones of language were always delusional. Besides, LLM’s ultimately pose little threat to humanity if we’re all hurling toward some form of postmodern trans-humanism that’s free from the burden of cultural institutions, history, folk knowledge, and even biology.

Let’s pause here. I get it — I was in grad school at the turn of the century when poststructuralism was saturating the social sciences — Frankfurt School, Foucault, and all that. But, despite the magnetic pull it has had within academia, I’m asserting that we, as a species, need anchors far more than postmodern ‘fluidity’ or deconstruction — even if those anchors may be imperfect, or if they’re ultimately constructs within ever-changing flows of idiomatic chains.

The Case for Human Universals

The case I want to make here is that current LLM’s inherently undervalue the cumulative, undocumented, emotional, and interdependent expressions we humans bring to our experiences in the world in two important ways. First, they mistakenly privilege language as primary (just as many postmodernists do). Second, that they train on data that’s heavily over-indexed on contemporary linguistic expressions (social media, forums, etc.).

To address the second point, let me introduce you to two people: Donald Brown, an anthropologist; and Manfred Max-Neef, an economist. Although motivated differently, both focused on the value of human universals as foundations, or lenses, through which we reach understanding our place in the world.

For Donald Brown, his drive was to challenge contemporary tendencies within the social sciences to over-index on social constructionism, which he felt exaggerated the range of variance across cultures. Instead, he argued that cultural and linguistic differences are actually products of highly structured human universals. By referencing cross-cultural evidence of hundreds of features (grammar, kinship terms, moral codes, etc.) he demonstrated that our shared cognitive architecture actually sets a finite limit in the range of cultural variations. In fact, despite potentially infinite stylistic elaborations, his research points toward a remarkable convergence on the same essential set of solutions to the human condition.

Max-Neef’s focus on human universals was motivated by his drive to correct the many failures of programs designed to support economic development in the global south. Instead of imposing a system on cultures, he wanted to tie economic development to core human universals that would ensure lasting and deeply rooted success. He argued that conventional development efforts suffered from a conflation of human needs with their satisfiers. Instead, he positioned fundamental human needs (like subsistence, protection, affection, participation, etc.) as finite, few, and constant characteristics of the human experience across time and cultures. How those needs were satisfied (satisfiers) is where he saw variance. Without that critical distinction, he asserted, we run the risk of denying human dignity rooted in our shared fundamental needs, and reducing the person to a malleable object.

LLM’s and Human Universals: What Tools Should we Use?

If we agree to limit ourselves to language in the near term, I want to advance a position on whose version of responsibility we should prioritize as LLM usage spreads, and whose ethics we should use to calibrate our interactions with LLM models. It’s not going to come from Reddit posts; or Wikipedia, or the political whims of the latest US executive branch. While these are all arguably valid training sources, they should be positioned as layers of expression, not core indicators of shared fundamental, and enduring, human needs and values.

We can start with Brown and Max Neef as foundations for training models that are rooted in enduring, human-centered, essentials, perhaps as a merger of their two approaches. To this, we might add insights from key evolutionary scientists that explore the links between human foundations and expressions of them:

  • Cultural Group Selection, posits that competition between groups with different cultural traits (norms, institutions, technologies) drives the evolution of large-scale human cooperation and altruism (Joseph Henrich).
  • Memetics, a model for a self-replicating cultural unit (a meme), that creates a simple, powerful analogy for cultural transmission in which ideas, behaviors, and styles are replicators that undergo a form of Darwinian selection (Richard Dawkins).
  • Punctuated Equilibrium (a paleontological pattern consisting of long stasis punctuated by rapid change), has also been used to highlight the role of historical contingencies and the concept of spandrels (non-adaptive byproducts) in human evolution and culture (Stephen Jay Gould).
  • Dual Inheritance Theory, models how genetic evolution shaped the human capacity for culture, and how cultural traits, in turn, influence the selective pressures on genes (e.g., lactose tolerance in dairy-farming cultures), (Robert Boyd & Peter Richerson).
  • Kin Selection theory, based on the concept of Inclusive Fitness, explains that an individual can increase their fitness by helping relatives who share their genes, providing a direct evolutionary basis for family bonds and cooperation (W. D. Hamilton).
  • Sociobiology, systematic application of evolutionary principles (like kin selection and cost-benefit analysis) to explain social behavior, including the development of biological explanations for human social structures, aggression, and gender roles (Edward O. Wilson).

Finally, we could add data sources that encompass an array of ‘satisfiers’ / cultural expressions. The annals of anthropology include diligently recorded, rich archives of human cultures and practices that allow us to identify similarities, differences, and the ways our foundational needs are addressed across time and cultures. Some particularly comprehensive sources we might tap include:

  • Human Relations Area Files (HRAF) – an international consortium of institutions dedicated to promoting comparative study of human culture, society, and language. (Database here).
  • Standard Cross-Cultural Sample (SCCS) & Ethnographic Atlas – datasets developed for quantitative cross-cultural analysis, a subset of the larger Ethnographic Atlas, now housed as part of the World Cultures Journal.
  • D-Place (Database of Places, Language, Culture, and Environment) – an open-access resource that integrates the cultural, linguistic, and environmental data for over 1,400 societies
  • Aarne-Thompson-Uther (ATU) Index – a comprehensive classification system for folktales – related indices and sources can be found here.

While some of these theories may be controversial in certain quarters, they can collectively help us lay a foundation of human universals that’s far more grounded and systematic than online echo chambers or social media pissing matches. If we enlist them in conjunction with a wide range of similar sources, we can provide a training foundation for LLM model pattern identification that prioritizes human interests rooted in deep understandings of human behavior.

The objective here is to establish an initial framework to anchor these models in what I’ll call the moral foundations of human universals. This might be achieved by using the resources above as key components of model context protocols (MCP’s) for consumer-facing LLM’s, or perhaps as content for reinforcement learning from human feedback (RLHF), but there are people much more qualified than me that could best leverage these resources to train LLM’s.

Is Gaming The Next Frontier?

Returning to the first point above, let’s dig a bit deeper, and think about future alternatives in which AI doesn’t treat language as primary (particularly text).

To get started, consider religion. Religion is about far more than scriptures. These are only partial relics of a set of practices, interactions, and emotions that compel us to come together for spiritual fulfillment and bonding. Other critical relics of religion also include the institutions and artifacts that mediate practices: temples, churches, kneeling, bowing, the cloth of priests, the smell of incense burning, etc. Practices and relics together form what I’m calling anchors that are far more rich, layered, historical, and critical to the human experience than LLM’s currently capture by focusing on written language. In fact, there’s MUCH more to be gained by understanding what qualities have made these anchors and related institutions ‘work’ for humans across cultures and time than in deconstructing and re-assembling approximations of language related to them.

So how might that look? There are some pioneers in this space that are beginning to flesh out different approaches that show promise. The most likely near term traction for this approach is within the fields of gaming and entertainment, where world creation has always played a central role. For example, Michael Libby’s Worldbuildr creates digital ‘twins’ of guest experiences (games, theme parks) to help simulate flows, interactions, and physical requirements for players and guests. Or, take Fei-Fei Li‘s World Labs, a “spatial intelligence company, building frontier models that can perceive, generate, reason, and interact with the 3D world.” Their mission to: “transform seeing into doing, understanding into reasoning, and imagining into creating” has recently produced Marble, a multimodal world-building model that allows users to create and edit 3D environments from various inputs like text prompts, images, and videos. Efforts like these, in which the experiential is moved closer to the center of focus appear to leave more room for integrating human universals within the rich range of expressions they take. Perhaps even more promising is the direction Yann LeCun appears to be headed. His departure from LLM’s centers around a vision for building “world models,” or systems that perceive their environments and grasp physical concepts.

Let’s hope these visionaries take the time to build models that don’t stop at integrating principles of physics or human factors, but treat our shared human values as a critical foundation for their work.

Photo: Examination of Easy Listener, by Nick Veasey

Image

Meaningful vs. Compelling

Are we headed toward a post-meaning online world?

Think back to the most recent movie you saw that you really liked. What did you like about it? What sort of impression did it leave?

Now, think back to one of your all time favorite movies — one you always like — a movie you’d watch again and again, anytime. What keeps you coming back? What makes this movie so special to you?

Now let’s consider some key differences between each of these movies.

Which one pops up in your mind’s eye when you’re just moving through your day? Which one resonates more deeply on an emotional level? Which one have you recommended to friends more?

The difference between the ways you responded to these questions surfaces some of the qualities that distinguish meaningful from compelling content; and this difference is becoming increasingly critical in what we now seem to be calling the ‘attention economy.’ Most importantly, this difference is a core part of determining how algorithms are shaped, which plays a critical role in establishing and reinforcing what constitutes ‘common’ knowledge (think content-scraping for training AI models), as well as which content becomes resilient (since the role of memory is an essential part of this difference — more on that later).

To get us started, I’ll explore how both meaningful and compelling content have played distinctly different roles as the web matured. Then, I’ll consider how we interact with each, the value ascribed to them, and how they’ve been (and could be) reconfigured as we head into a new era of interactions in the age of AI.

Breaking it Down

A close look at the ways the tech industry traditionally related to content helps set the stage. Although the history is diverse, the industry originally focused on the informational and the factual. Content was, simply, data. This made sense, given the core nature of digital technology itself, and the way programmers and engineers drove most of how tech products were developed. From strings of zeros and ones to coding logics, these systems were ultimately rooted in binary ‘call-and-response’ interaction models: inputs and outputs. Content functioned technically as a resource that could be referenced or processed. Eventually, the internet took this model to another level. It made information available globally through a robust and decentralized distribution system that had never existed before. Yet, the interaction models within it remained relatively constrained by utilitarian transactions.

Then came social media. While you could argue that it was originally conceived in primarily transactional terms — finding and ‘friending’ others — it quickly became evident that many users found these interactions to be valuable beyond the transactional. They valued the connection itself. As these platforms matured, there were clear indications that people wanted to include a much broader range of expressions, interactions, and co-modifications of content that included qualities like nostalgia, commemoration, fondness, flirting, and humor (to name a few).

It was perhaps no accident that the rapid growth of social media nested comfortably within a parallel rise of reality entertainment (and the decline of traditional media institutions). There’s much more to consider here, including the long (and deep) tradition of seeing ourselves as sources of entertainment (parlor tricks, seances, folk dance, ice bucket challenges, etc.) — more on that in another post. The point I want to make here is that transactional relationships with digital content only went so far. People were seeking deeper engagement, with content and with others. Their impulses were driving them to look for much more than just getting an answer. They were seeking exchanges with other humans, channeled through content that was emotional and authentic. We’re still seeing this play out in our digital lives everyday, of course.

Whether you perceive these emotions and connections as positive influences on our collective social fabric or not, it’s clear that the range of our interactions has expanded, and our shift from transactional to social revolutionized the way we think about digital content. A techno-centric view might posit that technologies like social media ushered in a new wave of behaviors. An anthropological one might argue that those behaviors always existed, and the technology caught up to (and exploited?) them.

“Huh” or “Hell, Yes!”

Let’s return to the differences between meaningful and compelling. While meaningful experiences are typically tangible, useful, and timely, they don’t necessarily fill us with the urge to reach out and share with others. Their value comes in the form of knowledge and understanding. In contrast, compelling interactions are magnetic, personal, and tap shared cultural narratives. Even more importantly, compelling interactions are catalyzing. We’re drawn to them again and again because they resonate so deeply for us. This depth propels ownership, action, and agency.

The table below summarizes some of the core differences I’d like to highlight.

MeaningfulCompelling
Transactional / Informational
Tangible
Useful
Timely		Engaging / Magnetic
Deeply Personal
Culturally Relevant
Catalyzing
Outcome: Knowledge & UnderstandingOutcome: Ownership, Action & Agency

In other words, meaningful content makes you go “huh,” or maybe “I see.” Whereas compelling content makes you go “OMG, did you see THIS?!” Meaningful content helps you feel informed, appreciative, or even satiated, but its function is ultimately utilitarian — then you move on. With compelling content, you feel driven to share, to build on the momentum of your enthusiasm by adding your own energy to it (sometimes repeatedly).

Value and Risk

In terms of our daily lives and interactions with content, we can think of meaningful content as the something that’s often used to resolve small disagreements, to navigate, to make decisions, or to learn something new. Overall, it’s a tool, not a propellant. Compelling content, on the other hand, has a distinctly emotional component. Interacting with it is embedded with qualities we don’t associate with the transactional nature of meaningful content.

Psychiatrist Dr. Goulston argues that when people are engaged in compelling interactions, they feel respected, engaged, and invited. They feel talked with (instead of talked at, or over), and then behave as if they’re choosing to do something. Perhaps most importantly, compelling interactions catalyze the initiation of new interactions from people exposed to content they find deeply engaging and highly resonant. They’re driven to share their experiences with others (remember your all time favorite movie?).

In this light, it’s easy to see how compelling content can set the stage for increased customer loyalty and brand resonance in marketplace settings. Examples include engagements with content that grow organically within a culture or community of customers. Emotional connection, authenticity, and active participation are some of its core characteristics. Here we’ll find superfans (not paid ‘brand ambassadors’) who enthusiastically share content that’s compelling to them. This can be exceptionally powerful and empowering when organizations are savvy enough to include them. They co-create, co-opt, and co-evolve brand identities (think IKEA hacks, cosplay, or fan fiction).

It’s also easy how to see how compelling content can be misused. In Jonathan Haight’s now classic analysis of moral judgments, he uses the metaphor of a rider sitting atop an elephant to illustrate how we reach moral decisions. His argument is that intuition (the elephant moving along) tends to come first, dominating our moral compass. Its nature is to go where it likes (what compels it). Rational thinking, represented by the rider, has some control, but is secondary to the often overpowering drive of the elephant.

Hacking this dynamic by ‘luring’ the elephant with compelling content is the risk I’m referring to here. There’s now ample evidence that conflict draws clicks, as does other content that one could argue has questionable social value. But the intent isn’t always nefarious.

Motivational science also plays on this dynamic to optimize for positive behavior outcomes. For example, in the early days of recycling in the US, it was initially difficult to persuade people to change well-engrained behaviors around waste disposal. Gradually, however, municipalities began to realize that if they tethered their recycling goals to something people cared deeply about — their children — they could influence household habits. Many launched recycling education campaigns in schools, which included both take-home materials and a new environmental consciousness students brought home to their parents, who were then ‘compelled’ to recycle at least in part out of their deeply rooted love and care for their children. Ownership, action, agency.

Beyond Binary: AI

What does this difference between meaningful and compelling mean for the digital world, and especially the web, moving forward? One view is that meaningful content may increasingly be subject to automated and endless reinterpretations via AI’s ability to scrape, summarize, re-word, and translate content. This could lead to commonplace repackaging of meaningful content that personalizes it in ways that give it ‘voice’ (both literally and figuratively). Could this transform meaningful content into compelling content? Will AI essentially merge the two in the form of ‘companions’ (or agents) that time, contextualize, and find personal points of relation and activation of meaningful content for us?

This potential merger (dissolution’?) is something our collective imaginations have only recently entertained. Even in the fairly recent past, our interpretation of how we might interact with AI was constrained well within the bounds of traditional meaningful-content conceptualizations. Take this clip from the film AI: Artificial Intelligence, where AI is portrayed as a font of knowledge largely conceived of as a giant vending machine for human queries. Here we see a future in which meaningful content is simply scaled up, but not reinvented as compelling content. In fact, while Dr. Know is all-knowing, his services are provided purely on his terms — and he’s quite literal!

We’ve come a long way in thinking about how we might interact with computers, in a short amount of time.

If we take the potential dissolution of distinctions between meaningful and compelling interactions and extrapolate it, it’s not hard to conceive of a post-meaning AI-dominant online world, where compelling content and interactions are ALL that matter — and meaning (or accuracy) is served up as a secondary priority embedded within compelling forms, tailored to our interests and preferences. Some would say this is already happening. They point toward common hallucinations and inaccuracies that ‘sound right,’ and are positioned by AI as correct, but are completely fabricated based on familiar patterns the tools have identified. In short, AI models may inherently index toward compelling at the expense of meaning (or accuracy). Another risk is that institutions or organizations generating AI content may use this to leverage our ‘elephants’ in ways that obfuscate the reasoning of our ‘riders.’

In a recent piece that explores the impact of AI hallucinations on our interactions with content, Matteo Wong takes this a step further: “AI products could settle into a liminal zone. They may not be wrong frequently enough to be jettisoned, but they also may not be wrong rarely enough to ever be fully trusted. For now, the technology’s flaws are readily detected and corrected. But as people become more and more accustomed to AI in their life—at school, at work, at home—they may cease to notice. Already, a growing body of research correlates persistent use of AI with a drop in critical thinking; humans become reliant on AI and unwilling, perhaps unable, to verify its work.”

Maybe. Or we could see an erosion of trust in online content (including suspicion that its purveyors are toying with our ‘elephants’) that leads to a rise in reliance on verifiable real life experiences. Events, gatherings, and any other interactions not mediated digitally may grow in importance, as a sort of radical empiricism rises. These gatherings (be they salons, conferences, clubs, debates, or other) may become the only way we feel we can find what we consider sincere, faithful, meaningful interactions, and authentically-expressed enthusiasm for compelling content. This is not to say that AI couldn’t play a useful role in organizing such events, or identifying alignments between them and potential attendees, but that ultimately the value of real life gatherings may lie in our collective appreciation for face-to-face experiences in which meaning is generated through conversations, and compelling interactions are ascertained by looking directly into some one’s eyes.

There’s an important take-away here for Silicon Valley: it’s probably not a good idea to throw AI at every possible interaction. Product-market fit isn’t rooted in what’s technically feasible, but in the value and trust a product builds with its users. At a minimum, considerations for using AI in any product should include the degree to which users are seeking meaningful versus compelling content and interactions, as well as the potential risk for inaccuracies to erode trust. This may not matter much in creative tasks, or conversational interactions; but when we need to rely on content and value its meaningfulness for productivity, for example, the stakes are much higher.

In Closing: A Break from it All

I’ll leave you with the clip below from a recent stay in New York — a glimpse of people engaging in playful acts of joy, scooting around in bumper cars in a rink of giant pink inflated balls. No phones. Probably no meaning. Just deeply compelling fun. And, perhaps most notably, no AI.

Image

What Works? Storytelling.

The enduring value of storytelling in the age of AI

(Part four in a series of four)

Let’s start with a story.

I was once part of a sizable team working on what’s sometimes called a Phase Zero project — the type of project with a brief that begins “The Future of…” The team included a few designers, a historian, an account manager, me, and a crew of other advisors who came in and out as needed.

About halfway through the project, we reached a significant milestone and decided to treat ourselves to some wine to celebrate. We cracked a bottle from a case we’d bought, and sat there in our project room feeling accomplished, welcoming our hard-earned decompression.

The room was situated in a row next to others in a huge loft-like space that was constantly abuzz with activity. It had three walls on which we leaned stacks of foam core boards filled with post-its. There was no door, so our colleagues could easily see the entire contents of the room just by passing by—which got us thinking: anyone could come by and help themselves to our wine! There wasn’t really a problem with theft, but we wanted to preserve our stash for future milestones.

As we sat there and sipped, the conversation drifted toward where we were going to hide our case of wine. The company kitchen was obviously a poor choice; so was the locker and bike rack area—way too much traffic. As we considered various options, the lead designer on the project suddenly stood up, grabbed one of the many Sharpies scattered about, walked over to the case, and scrawled the phrase USER VIDEOS on top of the box.

“There,” he said, “No one will ever touch it!”

As a relatively young researcher, I’m pretty sure I didn’t find his joke very humorous at the time. But in hindsight it was perfectly timed and well-delivered. And, he had a point. The reason video clips of interviews with participants can fall short is that all too often they aren’t situated within a compelling narrative. Talking heads aren’t a story, and simply sharing them as ‘the voice of the consumer’ risks random interpretations of their meaning among your stakeholders. The larger narrative we construct about participants’ experiences IS the insight. It should convey the interpretive value we bring through the positioning and sequencing of stories from ‘the field’ that help our message resonate.

“The difference between giving an example and telling a story is the addition of emotional content and added sensory details in the telling. A story weaves detail, character, and events into a whole that is greater than the sum of its parts.” —Annette Simmons, The Story Factor

With this in mind, I thought it useful to take a deep dive into storytelling from an anthropological perspective, with an emphasis on what gives it lasting value, as well as where AI fits within our long shared history of storytelling. So, let’s start with the basics.

Why we Tell Stories

“Our species thinks in metaphors and learns through stories.” — Mary Catherine Bateson

Image

Anthropologist Mary Catherine Bateson, daughter of Margaret Mead and Gregory Bateson, is recognized for having enriched the discipline in many ways, especially her contributions to intergenerational communication. And like many anthropologists, she leveraged the power of storytelling in two primary ways. First, she saw the value of storytelling among her participants as a means of understanding how they embodied and shared cultural norms in personal ways. Second, she deployed storytelling in her own work to help her audiences see how their experiences within their own culture are echoed in, or contrasted with, those of another. In both cases, the value of storytelling centered much less on conveying facts or convincing others of a particular position, than on sharing narratives that evoked, inspired, connected, and transported people beyond the current moment and their constrained spheres of concern.

“Information has value only for the moment it is new, but storytelling is capable of releasing information even when the story is very old…” —Anne Pellowski, The World of Storytelling

On a personal level, we tell stories for many different reasons: to seek emotional clarity, process feelings, connect with others, achieve affirmation, fill a void, or make sense of the unknown. But above all, we tell stories to share and; if the conditions are right, transform ourselves and others in some way. Social research into the dynamics of live storytelling demonstrates that the process is primarily a co-adaptive interaction, where the storyteller offers up a personal narrative that they entrust with their audience. In return, invested audiences demonstrate empathy and engagement by showing that they care enough to take the time to listen and respond.

This creates a form of trust-building that’s critical to the shared meaning-making that storytelling provides. In short, good storytelling connects. It’s the medium that holds the shared experiences that build empathy through the dynamics between storyteller and audience.

“Stories provide the context to understand the awakening of transformation…This experience is at its most powerful when it occurs at the cusp of an era, a transitional turning point in a person’s life, or in current events, as we experience the world changing before our eyes.” — Bobette Buster, Author of Do/Story

On the biological level, neuroscientists have identified ‘mirror neurons’ that are triggered inside our brains when we watch the actions of others. These neurons, which are also associated with empathy, have been shown to activate when we immerse ourselves in stories. So, when a storyteller expresses an emotion, the neurons that fire in their brains are mirrored in the brain of the listener.

Even more intriguing, neurologist Dr. Uri Hassan‘s experiments have revealed that not only do these neurons fire in the brains of the listener when triggered by a storyteller, but that listeners often anticipate upcoming emotional stimuli, as though they know what will happen next. Even with different stories and different listeners, the pattern persists. Whether or not this can be construed as evidence for storytelling as a hardwired, primal, trait remains up for debate. But clearly it is critical to how we interact with one another and navigate the world around us.

Storytelling Basics

Many social scientists, behaviorists, and even neuroscientists have long argued that all humans are inherent storytellers and that we are continually engaging in different forms of narrative every day. This is evidenced by much of the research into storytelling in the field of human development.

If you’ve spent any time with kids, you know that they have very unique ways of telling stories. The plot (if there is one) often wanders; they tend to ramble on, end abruptly, suddenly shift topics, or connect unrelated ideas. Although the absurdity (and cuteness) of these stories can be entertaining, they can also be challenging to follow. What you’re witnessing in these tales are little experiments with language that eventually build toward full stories.

Human development researchers have identified five main stages of narrative development that map roughly to different age groups. These range from early ‘heaps’ of content that tend to form at around age two, through to ‘true narratives’ that emerge when children are five to seven years old. At each stage, narratives build greater context, reflect increasing awareness of cause and effect, and eventually delve into resolving a problem and connecting motivations of characters to a plot.

Image

Below, listen to a 4 year old share a primitive narrative…

…and here Claire (advanced for her age) shares a focused chain.

Each of these stages are layers we all build upon as we grow; and any layer is accessible to us as we share stories in everyday settings. For example, early stages of brainstorming could be categorized as ‘heaps’ or ‘sequences.’ Some memes might be considered focused chains, where additional knowledge is required to interpret an ending (a story fragment embedded with a ‘wink’ shared among the like-minded). The same can be true of small group conversations between people familiar with one another, where participants tell just enough to be understood by others, who then complete the story on their own based on shared views of a topic. In short, we’re all telling some sort of stories, all the time. They may not always be full-fledged ‘true’ narratives, but they tap different stages of how we learn to tell stories, depending on our audience and contexts.

Putting Storytelling into Practice

Hopefully by now I’ve convinced you of the value and underlying drivers of storytelling. But you may be thinking “I don’t have the time to master this skill.” Well, we don’t all have to be world class storytellers to leverage the connections that storytelling brings. In fact, it can sometimes be enough to engage one of the subsets of storytelling we all learned as children.

For example, if your audience consists of people with very similar shared experiences or close ties of some sort, it may be enough to offer focused chains, in which events follow a theme, but the additional knowledge of the audience is required to interpret an ending. In these cases, you might enlist focused chains to reinforce bonds between the storyteller and people in the audience, as well as between audience members themselves.

Similarly, primitive narratives, in which events follow a theme and there is some indication of cause and effect, could be used as short cuts for storytellers to identify with their audience, without the necessity of constructing a logical sequence of events. Memes often function as very effective primitive narratives. Neither of these require developing full true narratives, but can be very effective ways of leveraging our shared storytelling integrants.

Regardless of your strategy for leveraging different storytelling stages, you’ll want to prioritize a few critical techniques that will always help you tell stories effectively: emotional connection, authenticity, interaction.

  • EMOTIONAL CONNECTION: Expressing your emotions evokes empathy between you and your audience. Center your story on emotions that will trigger connections with each audience. You don’t necessarily need to cry or shout, but letting people know how you feel is incredibly powerful.
  • AUTHENTICITY: Variation, imperfections, and vulnerability help stories feel real and accessible. It’s a myth that you need to “find” your storytelling voice, just speak as if you’re talking to one very understanding friend.
  • INTERACTION: Good stories evolve and adapt based on exchanges between the storyteller and their audience. These interactions draw storytellers and audiences into a stimulating exchange that invites the audience into the creative process. Provide ways to help your audience become part of a co-constructed narrative – ask questions, pose hypotheticals, and leave room for pauses where the audience can digest and reflect.

In addition to these core techniques, Bobette Buster has written an invaluable guide that offers an enlightening set of ‘rules’ for impactful storytelling. Everything from making your story personal, to highlighting one gleaming detail, and emphasizing the senses, are all baked in to tiny book that serves as a critical foundation for constructing stories. She also shows us exactly how storytellers leverage metaphor and passion to move their story from personal to shared meaning. I won’t attempt to summarize her work here, mostly because it’s worth its weight in gold to any student of storytelling (just go buy it, the link is right up there!).

However, it’s worth thinking about how some of her rules map to the core techniques listed above. Below is a rough sketch of how you can use some of Buster’s rules to accomplish this.

Image

Finally, I’d like to expand on a key component that Buster and other storytelling scholars emphasize. Nearly every storytelling expert argues that one of the most critical components of effective storytelling is letting go so that audiences have the space to identify personally with the content by engaging their own imagination and personal references.

Letting go often unfolds in stages as storytellers seek out new audiences in ever-widening circles of safety. This helps them adapt their narrative and find new ways to visualize more satisfactory alternatives. Experts describe the process as one of claiming ownership (comfort and confidence in telling the story), building trust (prioritizing honesty & transparency), and gaining permission (understanding when your audience is ready and willing to hear different parts of your story). They also emphasize that these techniques should be used only after the storyteller has fully processed the experience, and understands its meaning for themselves in order to usher others into it.

The Moth, a storytelling organization that hosts live events, does a great job of helping storytellers both actualize their stories and “let go” to make room for the audience. They not only encourage storytellers to “have some stakes,” in which the teller stands to gain or lose in the telling, but to also “…Play with the details. Enjoy yourself. Imagine you are at a dinner party, not a deposition.”

Storytelling in the Workplace

In workplace settings the goals of storytelling shift a bit. Instead of connecting with a broad audience, you often need to connect with — and influence — a wide variety of stakeholders. While this does make things more complex, tailoring your story to them and their circumstances will go a long way toward connecting with them on a personal level. Your mission, using the tools described above, is to get beyond convincing, so that you can change hearts and minds. In short, you want to evoke, inspire, and transport them (and maybe you); and to do that, it helps to start with an understanding of who they are and what motivates them.

Before heading into any room of stakeholders, remind yourself that each person there is coming from a different experience that day, and more importantly, few of them have likely considered the the role of research in their work. What’s more, many of your stakeholders likely exist in a world of continual context switching, which increases the importance of ‘meeting them where they are’ as you begin to share your stories. This means that the first minute or so of your story is incredibly important, since it sets the tone for the interactions that follow and provides an ‘on ramp’ for them to acclimate and immerse themselves in what you have to say.

Ask yourself where the team is within their development process. Are they early stage? Mid stream? Post-launch? Each of these will have bearing on the how they listen and the degree to which they can absorb additional cognitive load. You’ll often find greater receptivity for stories focused on foundational insight at early stages, or during a de-brief at the end of a project, when their appetite for expanding their thinking is greater. Take the time to assess what forms might work best at these stages as well. Are they looking for a North Star? Do they need a hero’s journey? A common enemy (aka competitive landscape)?

Shorter stories that emphasize action tend to resonate best when the team is fully immersed mid-stream in a project. These shorter stories may later add up to a larger POV, but ask yourself whether or not the team has the capacity or interest to soak in a ‘master’ narrative at this point. Consider collecting these shorter insights to pull together in a larger story that emphasizes sense making and reflection when the team is better situated to receive it.

Finally, ask yourself where else your POV might be useful. Pull your head up and look around, there may be multiple places to land your work. Are there groups who are in a better position to benefit from your insights right now? Are there people who are well-positioned to amplify the story you’d like to tell? Who could advance (and challenge) your thinking? To whom might you reach out for feedback? Where is there mutual benefit for others to champion your work?

AI and Storytelling

By now you might be thinking, “OMG, when is he going to get to AI?!” Your time has come, my friend.

It’s now common to hear that AI models can generate stories. While it is true that they can use predictive language modeling to manufacture texts that include the basic structures of stories, they are inherently limited to patterning narratives after the data on which they’ve been trained. Some have described these models’ attempts at constructing stories as a ‘race to mediocrity’ to describe the quality of the narratives they produce, since they’re inevitably constructed from content that’s been ‘averaged out’ across all of the stories they’ve been fed. They can iterate ad nauseam on a theme, but can’t produce unique or novel ideas (or any actual ideas at all). More importantly, they can’t offer the kinds of connections from one human to another that storytelling provides.

That’s because a huge part of storytelling’s value lies in how it’s experienced. Storytellers learn as they share by experimenting with, and building upon, layers and sequences of metaphor — discovering where they find resonance between their passion and their audience’s engagement each time they share. This interactive exchange between the storyteller and audience responses not only helps the storyteller evolve and refine the narrative, but also (and more importantly) builds the scaffolding for trust between them and their audience.

While large language models may iterate and refine their stories based on human feedback, their results are always already focused on pleasing the user. Real-time feedback in the form of an audience’s laughter, gasps of shock, tears, boos, poignant silences – none of these are inputs that large language models integrate in their process of refinement. And, even if they eventually do, it’s difficult to see how these models might find value in connecting with, and transforming alongside, their audiences.

You could imagine a scenario in which a human prompt engineer develops a storytelling system designed to connect and transform audiences. Maybe it’s a story that adapts itself each time it’s told to a different audience. Maybe it’s delivered by an animatronic robot. Even then, the relationship between the audience and the model is predicated upon the priorities and dispositions laid out by the prompt engineer, who is ultimately (presumably) driven to connect with the audience. The model itself doesn’t care. It isn’t passionate about sharing any story. But a storyteller does care; and, good storytellers are full of passion.

All this is not to say that AI isn’t a valuable tool for honing stories that storytellers craft. I’ve heard of storytellers that use AI to help them bring order random thoughts, to ‘interview’ them about an experience, or even to better express themselves in a voice they’ve established as a signature form of their unique expression. Perhaps we’ll see storytelling evolve over time as these tools are increasingly leveraged by storytellers. In the end, AI’s greatest value won’t come from taking the place of storytellers but to enrich the worlds storytellers co-imagine with their audiences.

Man…is the storytelling animal. Wherever he goes he wants to leave behind not a chaotic wake, not an empty space, but the comforting marker buoys and trail signs of stories. He has to keep on making them up. As long as there’s a story, it’s all right. Even in his last moments, it’s said, in the split second of a fatal fall—or when he’s about to drown—he sees, passing rapidly before him, the story of his whole life. — Graham Swift, Waterland

Sources

Image

What Works? Adaptation.

The enduring value of adaptation in the age of AI

(Part three in a series of four)

Growing up in Florida meant many long days at the beach. We’d spend the whole morning riding waves, playing ‘Marco Polo,’ and chasing down errant frisbees. Then, after a snack-gorging break, we’d shift to more sedate activities like collecting shells and, eventually, settling down to build sand castles. As we grew older, our castle construction methods graduated from standard bucket-and-shovel to more elaborate drip castles and sophisticated molding techniques. Once finished with our creations, we would pause to admire them from different angles, and then force our parents to do the same. We might add a few finishing touches or maybe a tunnel or two, based on ‘feedback.’

Eventually, our attention would stray and we’d stroll down to the pier, or wander off to watch people fish as the tide came in. Oh, wait! The tide’s coming in! A moment of panic would set in as we realized that we’d built our castles too close to the water’s edge (again). Filled with urgency, we raced back to check on our creations. The first order of business was to dig a moat as fast as possible to establish a barrier. Then we’d hastily construct a ‘wall’ to block the incoming force. It would work for a while, but the water would inevitably power through, slowing shifting from broad thin skims to more forceful foamy gushes. But this only boosted our valor. Sometimes ‘re-enforcements’ were called in. But we knew all along that the tide would win. In the final moments, we’d sit and watch each wave slowly take down our creation. We’d collapse, out of breath, and gaze at the destruction with a mix of loss, awe, and fascination as we surrendered our work to forces much larger than us.

Inherent Adaptation

There are many ways we embody and embrace adaptation as a species. In fact, we willfully and regularly construct contraptions far more substantial than sand castles with full knowledge that time will pass, paradigms will shift, and pillars will fall. Even when we don’t consciously do this, our enduring fascination with both historical shifts and science fiction are there to remind us that what may feel is permanent, is anything but.

In her recent article “Can Animals Tell Time?,” evolutionary biologist Heather Heying highlights this point in her consideration of the different ways humans have devised to measure time.

There are, we humans have proclaimed, 24 hours in a day. And an hour is now of fixed length, a length that is split into 60 minutes. But in the Middle Ages, this was not always the case. In some parts of Europe during the Middle Ages, it was asserted that those 24 hour were evenly split between day and night, no matter the season. Twelve hours of day. 12 hours of night, all year long. Thus, in the long sunlit days simmer in medieval Europe, the 12 hours of daylight were long hours, far longer than today's standard 60 minutes, and the brief nights had short hours, 12 short hours each of which was shorter than the modern hour.

“[Humans have] a tendency to create things, and then let them change.” – Heather Heying

In daily life, adaptation can surface as we watch the ‘anchors’ we collectively create shift from feeling like permanent structures to reminders that what’s around them has changed. Shifts in government, holidays, or commemorative objects — and the social functions they perform — are all prime examples.

At the genetic level, our bodies evolve through ongoing adaptations that take the form of mutations. There needs to be enough ‘error’ in the ‘printing’ of our DNA to allow our bodies to change in response to different environmental conditions. In that sense, adaptation is truly hard-wired into our physical form and how we exist in the world.

Unsurprisingly, adaptation is equally prevalent in the insights industry. The frameworks, strategies, and mappings we create may be perceived as fixed representations of a system; but, in the grander scheme of adaptive change, they function more as snapshots or perhaps catalysts within the flywheels of change. Essentially, they’re tools to drive more informed iteration (not solutions in and of themselves). Ultimately, the value of our work is in guiding iterations, and reminding our colleagues that the ‘right’ solution is rarely the first one, and likely won’t be the last either.

Leveraging Adaptation Models

So how do we leverage our species’ inherent forms adaptation? How do we facilitate them within cross-functional teams? To follow is an example from work I led that leveraged an adaptive process to inform product development.

As part of the central team within a product incubator for social apps, my role was to provide insight into human behavior and culture to help founders and their teams develop products that aligned with the communities they intended to serve. One of the biggest challenges most of the startup teams faced was how to connect people in meaningful ways while also providing ongoing value. Most of the founders tended to approach this challenge by focusing intently on tweaking different features and then testing them with a small group of beta users in tight feedback loops. While critical, what this approach often missed was a deep understanding of what consistently brings people together to form communities with lasting deep bonds. In short, they needed to zoom out, not in.

Borrowing from some previous work I led on collective achievement and the optimal ways it can be facilitated on digital platforms, I leveraged Van Gennep’s rite of passage as an inspirational model for how humans form lasting bonds. This was instrumental for the start-up teams within the incubator because it helped them understand the critical, and universal, stages of experience that drive people to form deep associations with a collective.

Depending on the team and where they were in their development process, I would start by sharing a mild adaptation of the model itself (below).

Image

Then, I added a layer over the model that translated it into the needs of the teams. Part of this involved breaking the model down into three basic stages: trust, bonding, and new identity. I created a goal for each, and then broke the goal down into actions they could take to reach the goal. The actions were rooted in research that ran parallel to rites of passage, most of it focusing on community development practices.

Image

These two frameworks were part of a workshop I hosted, which included a set of examples from start-ups and companies that excelled at building products that helped people form bonds. It also included exercises to help teams actualize the steps they would take to apply the principles.

While many factors contributed to the success and failure of each team, this application of one way in which humans adapt was instrumental for many of them to create more compelling and enduring experiences for their users. Many of them used the model to frame their product growth metrics across the stages of trust, bonding, and new identity, during their reviews.

AI Doesn’t Wander and it Isn’t Embodied

It’s worth noting that, even when mediated through an app or other digital experiences like gaming, people experience rites of passage in highly personalized, and often emotional and physical ways. They form trust with others through experiments with vulnerability and humility; they bond through shared struggles in which the give and take of collaboration and support are navigated through relationships within a cohort; and they take on new layers of identity that internalize their own unique mix of the personal, the cohort, and the cultural.

While AI models can help us identify patterns across adaptations, and even adapt themselves (listen to two AI agents ‘talk’ to one another here), their capabilities don’t include the forms of embodied experimentation essential to our adaptation. Nor do they involve the emotional connections between humans that might be considered the ‘lubrication’ for that embodied experimentation. In short, we humans see inherent value in the freedom to wander (even aimlessly) through new and different environments with each other; to form bonds through those experiences; and, eventually, reach new embodied understandings (even new physical manifestations) of ourselves.

Image

The French situationiste’s understood this well. Wandering explorations of urban environments they called derivé demonstrated how our bodies serve as collective devices through shared experiences across changing conditions. This practice was designed to generate collective meaning-making through “aimless, random drifting through a place, guided by whim and an awareness of how different spaces draw you in or repel you” — to sense the ‘psycho-geographic’ conditions of a location. Applications from derives led to new approaches to architecture and urban planning that integrated what they called the “discovery of unities of ambiance…their principal axes of passage, their exits and their defenses.”

The ‘topoanalysis’ of phenomenologists like Gaston Bachelard, brought similar values to light. His examination of intimate spaces such as the house, cellar, attic, drawers, nests, shells, and corners, illustrates how these spaces evoke deep emotional human responses, and even reveries, within us. He argues that these spaces, and our relationship to them, serve as holding bodies for our memories and dreams; and that they play a crucial role in shaping our sense of self.

The forms of human adaptation explored in this post, their relationship to our physical environments, and our interactions embedded within them, are largely physical, sensorial, and emotional. They demonstrate how we’re propelled to adapt and interpret and bond, through forces that are uniquely human. Call it intuition, or maybe instinct blended with milieu; but whatever it is, human adaptation doesn’t follow set protocols or adhere wholly to patterns of predictive modeling.

As human insight practitioners, we’re in the unique position to leverage the human propensity to wander, to identify symbolic outliers, to sense layers of emotional meaning in our surroundings and interactions. And, we’re experts at sharing these realizations, struggling through interpreting them, bonding through their mutual resonance, and ushering in change (both emotional and physical).

I’d like to close with a few prompts:

  • Are you limiting yourself to responding to questions from your team about behavior, or are you helping them see how behaviors sit within human adaptive responses?
  • Are you caught up in solving for an endless string of particulars or are you zooming out to see how they’re connected, and under what paradigms?
  • Are you checking boxes or interrogating the unexpected?
  • Are you shifting focus between process and context?
  • Are you showing up with answers or hosting new understandings?
  • Are you focused on reactions or adaptations?
Image

What Works? Curiosity.

The Enduring Value of Curiosity in the Age of AI

(Part two in a series of four)

Let’s start in the backstreets of Tokyo, with a story from Ethnographic Thinking.

My team and I were in the discovery phase of our project, exploring different models for retail experiences. More specifically, we were looking for new and innovative ways that retailers were combining products and services, and Tokyo fit the bill.

Our focus was on businesses that had similar offerings to our client: home care, nutrition, and beauty; so our field research included visits to many of Tokyo’s retail hubs and major department stores. In one, we bought a set of tea cups, and observed the intricate packaging practices and customer care rituals of Japanese retail. In another, some team members had skin tests performed at a beauty counter, and received personalized products that matched their skin type. We also visited tea shops, nutritional centers, and took a cooking class.

At the end of one particularly long day, having duly completed our research agenda, we wound our way back to the hotel through Tokyo's labyrinth of tiny alleys...and promptly got lost. In our haze, we spotted a shop that looked like it had something to do with shoes, although it didn't look like any cobbler we'd ever seen. In fact, it looked more like a small hotel lobby, with shoes prominently featured in the window. Although this clearly wasn’t on our research plan, we decided to venture in.
At the back of the shop, the shopkeeper was putting the final touches on a repair she’d recently completed. We spent some time talking with her about the business — how long she'd been there, what types of repairs she was making, etc. While we were chatting, a customer came in. 

A clerk appeared out of nowhere to serve the customer, and after some back and forth, pulled out a pair of shoes and showed them to the customer. Then he put them back into a cabinet behind the workspace. They chatted a bit more, and he pulled out another pair — discussed them at length with the customer, and then put those back too. As we stood and watched, we eventually asked the shopkeeper what was happening. 

After a bit of back and forth, we learned that this was not only a shoe repair shop, but also a shoe hotel, where customers with small apartments could store their shoes when not in use. None of us had ever heard of a shoe hotel before, so we continued to probe. It turned out that in addition to storage, the shop offered ongoing high quality care, and perhaps more importantly, insight and casual banter about the latest trends in footwear — a hub for all things shoe related.

This unexpected experience drove whole new directions for our team to innovate. We used it as inspiration for a new concierge model in which experts (not salespeople) offered both product and tailored services or experiences related to it.

None of this would have happened had we not stopped at that shop. Having the genuine curiosity to move beyond our plan, and stretch our thinking to be more than just inquisitive, showed us all the value of fully embracing an open mindset.

The Universal Appeal of Curiosity

You might be thinking that I’m going to call for more curious and exploratory research. Yes, expanding our pool of perspectives and remaining open to unexpected connections in our work clearly has advantages. For one, continually applying a curious mindset increases the odds that new ideas will inspire our work. More importantly, curiosity drives an influx of ideas that eventually cross-pollinate and build on one another. This exposes our organizations to a wider range of perspectives, which stimulates creativity and reduces stifling groupthink — all valuable contributions.

But the point I want to make in this post goes beyond that. Across many projects over the years I’ve noticed that curiosity generates the most value when we create opportunities for our cross-functional partners to engage and embody their own curiosity as well. For example, our accidental visit to the Tokyo shoe hotel had such wide-ranging impact in part because the Vice President of Innovation was with us that day in the field, where his own curiosity was activated. In fact, after we got the conversation rolling with the shop owner, he was just as engaged in asking questions as the research team.

Facilitating the curiosity of others requires us to shift focus from thinking of ourselves as the only instrument for exploration to serving as hosts for activating the curiosity of others. Your cross-functional partners may be curious about different things than you, but they are curious. This is because the core drivers of human curiosity are deeply intertwined with our evolutionary history and neurological structures. The literature on this topic runs deep, but I think it’s worth highlighting a few insights here.

First, like all humans, our collaborators have an intrinsic motivation for novel information and exploration. This fundamental aspect of curiosity has significant evolutionary value. The drive to explore the environment, even without immediate rewards, allows us (and many other animals) to learn about resources, dangers, and opportunities, which enhance our adaptability and chances of survival in challenging environments. Neurologically, curiosity also triggers our dopaminergic system and prefrontal cortex as intrinsic rewards associated with novelty-seeking behaviors.

Our responses to surprise are particularly telling in this regard. From an evolutionary perspective, responding to unexpected events is crucial for learning about changes in the environment and avoiding potential threats. But neurologically, experiments suggest that there’s actually a neural link between surprise, memory, and the drive to learn more.

Enabling curiosity works because we’re inherently curious as a species.

If you can find ways to ignite, engage, and frame your collaborators’ inherent curiosity, they will often become more active contributors to strategic insights and — even better — advocates for those insights across the organization, since their own process of discovery will make those insights more memorable and useful to them.

So how exactly do we facilitate and enable curiosity within cross-functional teams? In the following sections, I’ll share three examples from projects in which we leveraged curiosity to accelerate strategic insight and cross-functional ownership. In each, I’ll emphasize a key principle of human curiosity — novelty, information gain, or embodiment — although they each included all three to varying degrees.

Immersive Experiences: Hosting Curious Embodiment

Let’s start with the last of these — embodiment. In my partnership with Ethnoworks co-founders Soo-Young Chin and Yoon Cho, we developed a series of immersive experiences that were an incredibly powerful means of helping healthcare stakeholders engage their curiosity by experiencing what it’s like to go through life as one of the patients they served. Rooted in ethnography and inspired by street theater, the projects began with ethnographic research into the daily experiences of patients (in one case, uninsured patients, in another, those managing complex health records). Our insights from this work helped us develop a set of personas, each of which served as a role for our stakeholders to assume through a series of scripted scenarios set in real-world locations.

Each scenario was designed to directly reflect the challenges of patients from our ethnographic research—and some were quite challenging! One stakeholder fielded calls from his ex-wife (played by one of our researchers) from a golf course as she fretted about managing their daughter’s diabetes diagnosis in light of his recent unemployment. In another, a stakeholder suffered long waits in multiple waiting rooms struggling to track down paperwork to cover breast cancer treatments. In still another, a stakeholder stood on a street corner looking for work alongside day laborers (spoiler alert: a real knife fight broke out!).

Image

At key times throughout each immersion, stakeholders were presented with a set of in-the-moment choices to make in response to scenarios and prompts. For example, the stakeholder taking on the role of a day-laborer suddenly ‘suffered a fall’ from a ladder and had to choose between visiting a botanica, navigating care at a local clinic, exploring acupuncture, or ignoring his pain and continuing to send money back to his family in Mexico.

After a day spent at various locations, facing often difficult choices and tradeoffs, participants shared their experiences with a broader set of stakeholders across the healthcare industry the following day. This is where curiosity payed off even further. Those who hadn’t played roles in the immersion had the opportunity to question those who did, and, more importantly, to explore their choice logics. They also had the opportunity to view video clips taken directly from our ethnographic research. Breakout groups then served as the venue for stakeholders to brainstorm solutions tailored to (and across) each of the roles and scenarios. In one immersion this brainstorm generated eight new ideas to meet the needs of patients; in another, ten. Results from one included the launch of a healthcare access phone system, and in both cases, solidified new private / non-profit partnerships to meet the needs of patients.

Live Model Tests: Getting Curious about Information Gain

The second example I’d like to share focuses on ‘live model tests’ – experiments designed to try out a new offering in real-world conditions. I led research for two of these tests, one in Moscow and the other in Toulouse. Both were pop-up experience centers our team built that combined retail sales and sustainable living activities (e.g., cooking classes, yoga, candle making, etc.). The idea was to learn as much as possible in a pre-defined time frame under real-world conditions in markets that had been traditionally challenging for the company.

Image

These tests were a fantastic opportunity to gather data about real-world responses to our experimental offerings. We tracked everything from types of products sold, foot traffic, attendance at events, sales agent productivity, and even time and motion studies of customer movements within the stores. So. Much. Data.

In fact, the challenge was often to determine what not to track since there were so many opportunities to gather data and derive insights. We spent many long nights after the shops closed, reviewing not only the data we’d gathered, but what we might gather, and what to eliminate. Ultimately, it was our team’s curiosity that helped us all determine the types of information that would be most valuable. Sales results were, of course, critical; but what factors influenced them? What inter-dependencies across various measures were contributing to our understanding of successes and failures for our test?

Genuine curiosity helped get us past data overload, and focus on what was most informative and compelling. Overall, the pop-ups served as a vehicles for curiosity, so that we could engage and understand the value of serendipity, unexpected interactions, unanticipated causal chains, and unforeseen interdependencies that drove sales.

Signal Scanning: Leveraging the Power of Novelty

Much of the work I’ve led over the years has involved pathfinding — a practice that identifies emerging needs and works backwards to develop products that fit, or can adapt toward, those needs. From my days at Intel’s Digital Home Group, where we were tasked with understanding the emerging role of technology in Egypt, Brazil, South Korea, Germany, to more recent focus on generative AI, I’ve deployed a mix of foresight, foundational, exploratory, and secondary research to ensure organizations are investing in products that meet both evolving and enduring customer needs.

Early stages of pathfinding often involve a practice known as signal scanning (or horizon scanning) — where researchers explore the margins of a given topic to identify signals of change that demonstrate strong momentum or early signs of growth. The goal is to identify patterns across those signals, and determine if there are strong enough themes that point toward marketplace shifts. A clear understanding of those shifts can help organizations determine whether it makes sense for them to invest in developing products aimed at serving emerging customer needs.

While this process includes careful analysis and interpretation of signals, patterns, themes, and shifts, I’ve found that signal scanning itself often offers one the greatest opportunities to engage a team’s curiosity. In fact, more recently I’ve found that gathering and sharing signals of change are in many ways one of the most compelling, participatory, and valuable stages of the pathfinding process. Cross-functional partners with deep expertise in their field are always thirsty for what’s new in their industry. They enjoy the challenge of considering how they might respond to those signals and leverage change in the marketplace. They also enjoy sharing signals of change they find themselves — a way for them to take on the role of curious explorers.

Their curiosity about those signals, the different scenarios they may portend, and how the organization might respond, is often more engaging for them than passively receiving reports focused on researchers’ interpretations alone. In short, the novelty of signal discovery, exploration, meaning-making, and debate for our cross-functional partners is often one of the most valuable assets of this work. We just need to provide the platform for them to engage their naturally-occurring curiosity, offer frameworks for interpretation, and get out of the way!

AI is Not Curious

Returning to the question of the enduring value of our practice for this series of posts; curiosity works as a form of engagement and organizational growth precisely because it’s uniquely human. It’s directly tied to evolutionary benefits like adaptation and survival, as well as neurological reward for learning and remembering. What’s more, across our fellow insights practitioners, among our stakeholders, and throughout the populations and participants who partner with us, curiosity is not only a trait we share, but a bonding force that emerges when we recognize it in each other. A curious colleague or participant is signaling openness, engagement, generosity, and sometimes even empathy. In short, we enjoy learning together.

AI models may simulate curiosity by asking questions based on patterns they identify, but ultimately these are prompts that lack the demonstration of commitment and camaraderie that curiosity builds between humans. This contrast is especially evident when you consider how AI tools ask questions of their users. If you’ve ever engaged with a conversational AI, it’s plain to see that they’re mapping patterns of question-making to the contexts of the inputs they receive. In short, they function in ways that might be best characterized as inquisitive, not genuinely curious.

Similar to the way we initially held fast to our research plan in Tokyo, AI models never stray from the constraints of the context they’re given — and likely never will, since that’s how they’re designed. In fact, LLM’s, and their prioritization of pattern recognition and reproduction, operate within the realm of the known, even when they are tasked with identifying outliers.

Curious humans, however, pick up on odd signs, stray interruptions, absurdities, serendipitous encounters, and other stimuli that trigger new lines of inquiry and increased engagement with each other. We will ask the stupid questions; and the responses have the potential to generate unexpected resonances and collisions with other humans. We will wander and explore without direction; and the result will create alliances and trigger conflicts with other humans. Thank God.

So, if we simply want to entertain additional lines of inquiry about a topic, these models can ask productive questions. They can even stimulate, propel, and enrich our innate curiosity. (In fact, the opportunity to surprise and delight here is incredibly promising for education applications.) But, if you’re designing for truly novel, embodied, human-centered offerings, human curiosity (not AI inquisitiveness) rules.

More to explore:

Image

Staying Power

Identifying our Enduring Value in an Age of Change

The recent uptick in conversations about the future of applied research has driven many in the industry to pause and reflect. What role will AI play in our practice? How have organizational priorities shifted post-COVID? What counts as ‘actionable insight’ in world where smaller teams can scale rapidly in ‘founder mode’? Which practices and processes can be streamlined? Which ones shouldn’t?

These are all valid questions; and while I think it’s important to address them, I’d like to free up some of our collective mind share to get us past the hand-wringing to focus on our foundations. To that end, this post is the first of four that will consider our enduring value in light of relatively recent shifts in the industry.

I thought I’d kick it off with a few personal reflections.

Timing

First, I’ve grown to realize that timing is critical in our industry. Perhaps that’s true in most professions, but it seems particularly pronounced in the world of insights provision. More specifically, the ‘in-the-moment’ value of insights work is indispensable, and I’ve seen it repeatedly accelerate our impact if we take the time to read the signs. It might be tempting to dismiss this as opportunistic. It’s not. It’s about knowing your audience, and tailoring your insights to sync with their momentum, reward structures, and priorities.

This shouldn’t be confused with blanket appeasement, which erodes our credibility. Instead, it’s about optimizing our value. I’ve written much more about this recently, but the point I want to make here is that adjusting our deliverables to the context in which our work is received doesn’t mean compromising rigor, which should always be at the foundation of what we do anyway.

In fact, this approach allows us to maintain that foundation, while simultaneously freeing us. More specifically it frees us from taking on the weight of thinking we need to solve EVERYTHING, and instead positions our work as more valuable because it is both insightful and contextually-aware — it’s work that’s critically instrumental in lifting everyone up, not just researchers.

Facilitation

Beyond insights, this also applies critically to the interpretation of data. As research practitioners, we are often trained to think that our interpretations of data are predominant. The same might be said of design thinking, and the inherent ‘it-can-solve-everything’ presumptions it advanced during its heyday. While both are often more informed by theory, years of practice, etc., I’ve got news for you: our interpretive value isn’t always top of mind among our collaborators.

If we open our lens a bit, we can see that a huge part of our interpretive value may actually lie in facilitating, guiding, and setting the boundaries for a wide range of interpretations to flourish. If you’re willing to accept this, it too can be remarkably freeing — and powerful. We don’t need to feel responsible for solving “everything, everywhere, all at once.” In fact, it’s often more powerful, and useful, to define the playing field itself.

Resilience

Finally, a deeper dive into context. Nothing in our training as ethnographers, designers, writers, strategists, or insights practitioners of other breeds, can prepare us for some of the friction we face in the workplace. Your professors weren’t a good gauge for dealing with this; and maybe your coach isn’t either. Interacting (and maybe sometimes ‘wrestling’) with cross-functional peers, over and over again, and being consciously and sincerely empathic about where they’re coming from, gives you two things: a broader lens and thicker skin. We all face a mix of headwinds and tailwinds. It’s up to us to know when to ignore the cynics and when to pause and consider the critical concerns of others; when to build incrementally on institutional orthodoxies and when to advocate for entirely new perspectives.

Enduring Value: Intro to the Mini-Series

Once we’ve freed ourselves from the burden of solving for everything, and established solid groundwork for facilitating well-timed and thoughtfully-contextualized insights, we’re in a position to work relationally, rather than directionally. That seemingly small shift opens the door to a broader surface for our work; and it is the context in which I want to consider three attributes that serve as what I consider critical foundations for our practice: curiosity, adaptation, and storytelling. In each of the next three posts, I’ll dig deeper into the layers of value that each of these attributes provides, and try to position them in the context of the capabilities of AI. I hope you’ll join me.

Links to series installments: Curiosity, Adaptation, Storytelling

In the meantime…

Image

Three Questions

Leadership in the Age of AI

I’ve had the good fortune to work with a few exceptionally skilled managers over the years. Like a great editor, or a skilled coach, the best managers help you gain perspective and build on your strengths. But more than that, the best managers are authentic, sincere, and invested in ways that make interactions with them feel inspiring, yet challenging. A big part of their job is to get you OUT of your head to frame and direct your energy, which can be a bit destabilizing — and that’s a good thing.

Recently, I’ve been thinking about what I appreciate most from exceptional managers, how it plays out in day-to-day interactions, and how both have influenced my own leadership approach. What I see as impactful leadership often arises from a combination of well-timed prompts, moments of insight, and (perhaps more importantly) critical prioritizations, well-informed actions, and keen decision making. At a time when the tools at our disposal are evolving rapidly and becoming much more powerful, these skills are even more essential in the insights industry.

In that light, I’ve synthesized my thinking in the form of three driving questions I see good leaders ask as part of productive contributions to teams.

What do we need to learn?

Our value as insight providers is shifting. Many of us are increasingly asked to go beyond providing insights, recommendations, or developing design principles, to prescribe learning priorities for a team. Leadership that excels in this climate frames conversations with teams around tradeoffs and the respective value of pursuing different lines of inquiry.

Operationally, instead of asking “What do we want to learn?” or “What would be interesting to learn?” or especially “What do we need to know that would prove we’re right?” (eek!) these conversations start with organizational priorities, then determine the most critical knowledge gaps, and then understand how team needs fit into both. This is the point at which we can distill strategic research questions, the answers to which satisfy needs across these layers.

Strategically, good leadership of this variety helps move beyond simply answering questions for a team and aims toward building core foundational knowledge that helps the organization thrive. It takes deep understandings of the organization’s customer experiences, and situates that knowledge within an outside-in perspective of the organization’s position in the marketplace, to inform a strategy for identifying opportunities the organization is situated best to pursue (both physically and culturally). This is sometimes framed as identifying what’s desirable (do consumers want it?), then determining what’s viable (is it possible to meet those needs in the marketplace?), and finally, defining what’s feasible (is our organization set up to meet those needs in that market?). Each requires a different learning ‘prescription’ offered at the right time and place.

What’s the best way learn?

It may not be so coincidental that we’re seeing increased value placed in the leadership skills described above. Two major changes are happening in our industry: pace and scale; both driven in large part by large language models (LLM’s).

For pace, we’re learning that LLM’s can accelerate many processes in our workflows. Give these models a data set from which to ‘learn,’ and they can outline research or project plans, transcribe interviews, summarize vast amounts of text, identify themes and outliers, jumpstart brainstorms, create prototypes and mockups…all almost instantly. Although not without the risk of hallucination, these models are remarkably effective at pattern recognition and eliminating rote processes, which frees us to spend more time on strategic questions.

As for scale, in the very recent past many of us in the industry developed insights through methods that identified patterns from data at scales contained within discrete data sets. We might later draw connections between those insights and other broader themes, but this was most often done by association or inference. While much of this work has long been assisted by computing power, LLM’s now provide the ability to dramatically increase the scale of our work beyond a dataset or association. The scope of LLM’s can span across entire bodies of knowledge, and interpret them via highly tailored prompts, allowing us to mine for insight and identify patterns at previously unimaginable scope and scale.

Given these two changes, considerations for how our insights align with organizational strategy become ‘weightier’ in many ways. Where do we go deep? Where do we stretch across? What methods and models make the most sense? Why? With the accelerated pace and unprecedented scope and scale LLM’s offer, the risk of rapidly veering off course is now greater; and, while you could argue that we can now recover from mishaps more rapidly with these tools, we also risk flailing haphazardly without the insightful steer of solid and well-connected leaders that deeply understand and advocate for the right focus for our work.

How do we engage?

It’s easy to assume that, as insights provisioners, we’re responsible for ‘solving’ team challenges by answering questions. Yet, in the vast majority of cases, our value is far more effective when we sync and parse our insights according to the needs of our teams and the broader organization. There’s much more to consider here, but in this post, I’d like to focus on leadership. As our roles shift within organizations, our strategic value is also changing course. An engagement strategy, with a focus on timing, team alignment, and — above all — a deep integration of organizational priorities, is becoming increasingly important. This means that ‘research reports’ occupy only part of engaging with teams.

Instead, we need a mix of deliverables and outreach, applied strategically based on an understanding of the organization and its institutional culture(s). So, there may be times when it’s far more engaging and impactful to present only key pieces of raw data, or conversely, to use LLM’s to generate insights across massive knowledge bases to identify broad patterns. There may also be times when you want to gradually build on the momentum of positive relationships with close colleagues, or, make a big splash with unfamiliar coworkers by introducing a new perspective or deliverable across the organization. A leader brings value by continually gauging the environment and adapting accordingly.

When executed well, an engagement strategy can create a North Star that goes beyond ‘solving,’ to accumulate a vision for teams, de-risk bets for the organization, and build momentum around promising new opportunities. It’s rarely perfect, and it often consumes MUCH more time than anticipated.

Final Thoughts

All of this may feel a bit ‘out of scope’ for some, but without it, we risk operating rudderless (and runaway) projects with new ‘power tools’ we fail to use responsibly. By focusing on the three questions above, skilled leadership helps organizations maintain focus and framing. Are you hearing your leaders ask these questions? Are you finding yourself asking them?

Image

Read. The. Room.

Leveraging ethnographic thinking to increase resonance with stakeholders

It would be an understatement to say that a lot has changed since 2018, when the first edition of Ethnographic Thinking was published. As the second edition hits the shelves, I’ve been thinking about the context in which this new release is landing. While I believe strongly that ethnographic thinking is more valuable than ever, we’re at a critical moment where strategically (and creatively) positioning that value is essential.

This post is a personal reflection on what I’ve seen shifting in the broader world of research / insight since the first edition, and how I’ve tried to leverage ethnographic thinking to respond. As always, your mileage may vary. Let’s start with a couple of high level observations, and then focus on a three-pronged strategy I’ve been using to address them.

Hold on a Sec

The first observation is simply this: Nobody wants your strategy. While this may be a bit hyperbolic, it’s a healthy premise for framing your approach. Across our stakeholders, many think of themselves as providing strategic insight in one form or another through the lens of their own practice. From their perspective, there’s no reason for them to prioritize your strategic insights simply because you’re a researcher. If you enter into engagements assuming that your insights should take priority, you risk coming off as arrogant — not a very effective strategy for influence (especially in an unsettled work environment). It’s our responsibility as practitioners to build a case and connect the dots to strategy, and research insights are not inherently strategic for many of our stakeholders.

So, you may be thinking that if positioning insights to inform strategy is an uphill battle, then you can lean into the role of messenger who conveys customer sentiment. Which brings me to the second observation: The ‘voice of the customer’ is not enough.

For this, I’d like to share a brief story about a failure that helps bring the point home. I was working on project focused on understanding the needs of young entrepreneurs. Our team had recently wrapped up initial analysis from a set of field visits across the US, and the company’s leadership was eager to hear what we’d learned. Because we hadn’t fully developed a set of actionable insights, we decided to pull together a set of video clips from field visits with some early theme statements for a screening with the leadership team.

On the day of the meeting, we briefly reviewed the background of the project, set the context for the video as a ‘first look’ into lives of our target demographic, and hit play. When the lights went back on, the room was deadly silent. The leadership team’s reaction can be summed up perfectly with just one gif:

Image

Clearly, we hadn’t provided either enough of an on-ramp to help them recognize the state and intent of the deliverable, or offered enough interpretation to give them an invitation to engage with the work. We were so steeped in our own process that we failed to recognize where they were in theirs.

The takeaway from both of these observations is that your insights won’t resonate with stakeholders if they don’t have a way to understand them from their own perspective—and it’s your job to determine what that is. The reality is that any of your stakeholders who aren’t researchers are working in a completely different headspace most of the day. Your deliverables need to help usher them to your POV. Give them compelling points of entry and fresh interpretive approaches that will lighten their cognitive load and seed their engagement.

How? You need to tell a story that resonates with each person in that room so that they internalize your insights. There’s much more that can be said about the craft of storytelling and tailoring narratives to different listener dispositions, but I’ll reserve that for another post. For now, I want to highlight the question of both when stories have the greatest influence with stakeholders and what types of deliverables could help.

I bring these two considerations up because I think they’re essential for effective engagement; and, also because I see a professional landscape in which many organizations are more idiosyncratic about the ways they integrate research insights than they were back in 2018. Once companies began building internal teams to bring insight work closer to product work, they also started to understand how that relationship did or did not jive with their approach to innovation, their organizational practices, and their interpretation of what impact means. As insights provisioners operating within this context (in-house or as consultants), navigating how and when orgs ‘digest’ insights is increasingly important. The good news is that doing so involves tapping ethnographic thinking in ways that are very familiar to many who operate in the insights industry.

The When and the What

So back to the when and what. A while ago, I started to reflect more on these two questions in the context of research deliverables. I decided to go back through key projects and really dissect what worked, what didn’t, and why. I focused on the context of deliverables in particular, so it wasn’t just an assessment of rigor or research quality, but also a consideration of reception, audience, engagement, and demonstrated momentum. What I found overall was that in the vast majority of cases, success hinged on instances where I consciously considered the timing of insights, and de-emphasized efforts to convince my audiences.

To follow is a three part strategy I devised in response to this assessment. As I mentioned before, it may not apply to your set of circumstances or work style, but it has proven effective in many of my settings. The strategy relies heavily on turning the ethnographic lens toward your teams or org so that you can understand their values, priorities, behaviors, norms, etc., and adapting your contributions accordingly. It also relies heavily on collaboration with your core team to pull together the right insights, in the right form, at the right time.

The starting point is to consider where your stakeholders are in their workflows, and tailor what you have to offer based on their current set of priorities. I’ve always found the catch phrase “meet them where they are” a bit cringey since it leaves little room to challenge the status quo, stimulate new thinking, or evolve ideas; but it comes close to conveying the general approach here, at least initially. So, to begin, ask yourself: Is your team heads-down in their current work streams, operating in get-it-done mode? Or, are they at a pivot point where they need to pause and assess or prioritize? Or, maybe they’re starting something new and are just beginning to give shape to the project goals. Each of these should trigger a different response in the way you frame and engage stakeholders.

Just in Time

Let’s begin with what is often the most common scenario — a team that’s mid-way through their product development process and is fully immersed in execution. The circumstances here usually include a lot intense, heads-down concentration on building and revising features. Weekly, or sometimes even daily stand-ups, prototype testing, and breakout squads are a common part of the team’s workflow. Do they have time for an extensive research read out, or a mentally-taxing workshop? Likely not. Instead of trying to force the team to disrupt their workflow and adapt to your process, step back, and ask yourself where they are in theirs. Attend their stand-ups, get to know their priorities, their short term and long term goals, try to understand their pain points, their promising bright spots, their motivators, what energized them and provides them with momentum, etc. Then, revisit your insights, break them down into subsets, and determine which ones you can bring back to the team in the form of a ‘just in time’ deliverable (or deliverables) that will help them most. Ask yourself what would help accelerate or fine tune their work best right now? How can you deliver this in ways that dovetail with their workflows so that your work offers added value.

This may delay some of the plans you had in place, but guess what? That happens all the time to your stakeholders too. Adaptation equals success in these circumstances, all the way around. Effective research insights that I’ve seen fit into the ‘just in time’ deliverables bucket have been adapted to the needs of the team in a few core ways: they are succinct, directional, and tied to the team’s current priorities and processes. You’re shooting for small-scale adjustments that have potential for outsized impact. Keep your deliverables simple and brief, prioritized for immediate relevance.

A useful conceptual model here is something like a newspaper brief — with a tight headline and just enough information for relevance and action. It’s helpful to craft phrases and generate labels for concepts that help people remember key points. Similarly, it’s often useful to illustrate the value of critical takeaways by highlighting one or two key images that help reinforce your message. And, most importantly, keep it short. Outputs that I’ve seen work well in these cases include resources that allow the team to quickly absorb insights, apply their learnings, and then revisit the deliverable when needed. Some examples include:

  • A simple visual diagram / framework that helps define an ecosystem, the flows within it, and current opportunities for action.
  • Brief walkthroughs of competitor products and features to illustrate marketplace context and steer shifts in development that offer product differentiation.
  • Co-creation sessions with stakeholders to quickly iterate on product specifics and prepare ideas for testing.

Insight Ignitors

It’s not uncommon for researchers to regularly have our own ‘aha moments’ in which we identify valuable connections across our insights and others. As insiders these moments are often exciting realizations, but they aren’t readily apparent for stakeholders who aren’t immersed in research workflows. If you want to highlight the value of these broader themes, you need to bring them on a journey and help them reach the same conclusions you did, and hopefully ignite the same enthusiasm you have along the way. If successful, igniting insight with stakeholders helps them internalize the connections and higher order insights you’re highlighting, and gives them ownership of both so that they can leverage them in their own work.

Timing insight ignitors includes two considerations. First, do you have a set of insights along a theme that can lead to a larger POV? This doesn’t have to be a grand thesis. It can be as simple as pattern identification across four or five projects, paired with secondary foundational insights, and possibly a framework that ties them together. What’s most important is that these insights collectively lead to a higher order POV that has direct relevance to product decisions. Second, where are your stakeholders in their own process? Are they open to pausing a bit and joining you on this brief journey? I’ve found that these deliverables are best shared at key milestones within a project’s lifespan — moments where project pivots might be necessary, or where priorities need to be set (or re-set).

As for crafting these type of deliverables, you want to guide your stakeholders through a clear and compelling story, with actionable results. You might imagine yourself an attorney who’s building a case for a jury. Set the stage, use (and carefully time) multiple data sources, alternate dense and lightweight content, use visuals to drive home key points, and edit down to just the essentials. Your job is to help them reach the same realization you did. Some forms I’ve seen this take include:

  • A short video to illustrate key themes across research insights and their connection to current product priorities;
  • Competitive landscape analysis that situates your team’s product within the context of current offerings.
  • Residual, Dominant, Emergent analysis that positions your team’s product within industry trajectories spanning from the past, through the present, and into emergent paradigms in the marketplace.

Host an Exploration

For these deliverables, your goal is to help expand the team’s thinking and guide them toward a ‘North Star.’ You’re not there to convince them of the importance of your work, but to provide the fodder and insights that can stimulate creative, generative thinking. Ask yourself what’s most compelling about the insights you have that can accomplish this. And, further, what sorts of provocations might you offer that spark curiosity and engagement?

While there are generally smaller windows of opportunity for these deliverables, they can have outsized impact if they resonate well with stakeholders. In most cases, the ideal setting is when a team has either just finished a project, is starting a new initiative, or is in a strategic planning stage of some sort. You want to catch them while they’re in a reflective mindset, and are pausing to consider critical direction or set key priorities. Sometimes this takes the form of adding to a de-brief or strategy session, or maybe contributing to a vision initiative. Explorations are typically welcome components at of any of these.

A useful model here might be think of yourself as a director staging a play. What are the key touch points that help your team connect personally with your insights? How might you offer a variety of content that helps people gravitate toward what interests them most? Have you shared what captivates you most about your insights? Have you left ample room for the team to engage, contribute, and interpret their own views? I’ve found that the following deliverables have worked well when hosting an exploration is the goal:

  • A podcast series or ‘fireside chats,’ followed by an open Q&A, where you invite stakeholders to explore and discuss different facets and considerations on a key strategic topic;
  • A consequences wheel exercise to frame and stimulate long-term thinking and actively integrate stakeholders insights;
  • A microsite or ‘museum’ accompanied by an open forum that collects and curate different POV’s, and gives people the chance to roam, interact with insights, contribute their own perspectives, or bounce ideas off of one another.

A Summary

To encapsulate the key points of this strategy, and help readers quickly decide where they might leverage these strategies best, I’ve summarized some prompts, models, and approaches for each below.

Just in Time — Is your team midstream in product development?

  • What would help accelerate their work best right now?
  • How can you deliver key insights in small ‘bites’ that dovetail with the team’s workflow?
  • Model: Newspaper headline writer
  • Approach: Offer simple statements, images, or summaries that emphasize action and direct connections to current team goals.

Ignite Insight — Is your team at a key milestone in their work?

  • What connections across your insights would be most valuable to help your team make better strategic decisions?
  • How can you bring them along to reach the same conclusions you found valuable?
  • Model: Attorney making a case
  • Approach: Usher your team through a clear and compelling series of insights that build on one another.

Host an Exploration — Is your team pausing to expand their thinking or find a North Star?

  • Do you have a set of insights that could be configured in ways that stimulate creative thinking?
  • How can you convey engaging, compelling insights that set the stage for generative work?
  • Model: Director staging a play
  • Approach: Curate themes across insights and offer easy ways for the team to explore and dive deeper where interested.

Some Final Thoughts

Although most of us aren’t operating in the entertainment industry, there are some interesting parallels embedded within this strategy. The following quote has always resonated for me:

“You have to have the talent for the art–the music, the acting, the writing, the art–but you also have to have the talent for being in the right place at the right time with the right people with the right approach. I had to become a certain physical person and I had to place myself in certain places in front of, beneath, and around the right people. It’s an art to be noticed; to be necessary; to be needed and desired. Develop it, if you can. If you can’t, then I don’t think any amount of talent will be of any use to you. Talent has to move. Talent has to walk up to people and ask to sit down and talk a bit. Most talent stays at home, and it remains a gift, but it doesn’t get out enough. Someone has to see it in the right context. The context is entirely your job.”

Marlon Brando / Interview with James Grissom
Image
Photograph of Brando and Marilyn Monroe at the Actors’ Studio benefit screening of Tennessee Williams’ “The Rose Tattoo”

Main photo credit: Jay Hasbrouck 2016, Independence Palace, Ho Chi Minh City