The first rule of etiquette on the social platform Reddit is “Remember the human.” This rule should apply to researchers, too.

It can be easy for people commenting online to forget that there are actual humans on the other end of the screen, and that online interactions can have an impact in the “real world.” Similarly, for the countless researchers collecting data from Reddit and other public websites, it can be easy to forget that this “data” comes from lived experiences of actual humans who might be impacted by its inclusion and dissemination in scientific research. We (myself, Michael Zimmer, Nicholas Proferes, Sarah Gilbert, and Naiyan Jones) have been thinking a lot about this problem.

In 2020, we read over 700 papers that used Reddit as a data source. Of those, less than 200 included any mention (generously defined) of research ethics. Based on inclusion criteria, we conducted additional qualitative, thematic analysis on 134 papers in order to understand what ethical considerations researchers are raising and ethical decisions they make. The result, published in early 2024, is a description and set of recommendations based on current practices.

First, we should note that for the collection of public data (which would be the case for most Reddit content), institutional ethics review boards such as IRBs at U.S. universities typically consider this research to be outside of their purview. This is because (in the U.S.) IRBs govern “human subjects research” which is defined by U.S. Code 45 CFR §46.102 as work that involves either direct interaction with humans or the collection of data that is both personally identifiable and private. For a number of the papers in our dataset, this was the extent of their ethical discussion: the data is public, and therefore we were not required to undergo ethics review. However, other papers engaged in a deeper discussion of their ethical obligations even if the research did not meet the criteria for formal review.

Other ethical considerations and decisions researchers highlighted related to issues such as consent, privacy, and risk mitigation. These included, for example: anonymization of usernames (e.g., use of pseudonyms), obfuscation or fabrication of quotes, analysis reported in the aggregate, ethical decisions based on the sensitive nature of the subreddit or content, content excluded on ethical grounds (for example, a researcher mentioning they explicitly did not collect certain types of content), possible misuse or unanticipated consequences of the research, risk versus benefit analysis of research, and discussion of whether user consent was obtained.

Drawing from our analysis of this corpus of papers, as well as existing literature on research ethics and best practices, we make the following recommendations:

• Regardless of how “public” data is, researchers should always consider the fundamental research principles of respect for persons, beneficence, and justice when making decisions about data collection and dissemination.
• Remember that ethical review is about more than just “checking the box” that the research protocol was screened by a required ethics review body, and it is important to document ethical decisions in the publication itself.
• “Publicness” is not the only ethical context that matters. Researchers should take a contextual approach to consider the ethics of research design beyond solely considering whether content is public (e.g., the context of the community, the sensitivity of the content, the goals of the research). Researchers should learn the norms of a community, which can help them understand community contextual expectations for information flow.
• Researchers should consider not just what potential risks and consequences are (including thinking beyond intended outcomes), but also what steps might be taken to mitigate negative outcomes.
• In considering the impact on and benefits to a community, researchers should share back with the community when they can — while also taking into account that some research findings might be harmful or put the researcher at risk.
• Researchers should holistically consider harms and benefits as they manifest to a community, and not just individuals.
• Researchers should consider explicitly including community members in their work, such as through methods like participatory/action research and collaborative ethnography, reaching out to moderators about data collection where appropriate, and reflecting on their own positionality relative to the population, even when conducting quantitative research.

Our work shows that many internet researchers have made positive steps to ensure attention to research ethics becomes more prominent in our practices and publications, and we offer these recommendations as a continuation of those efforts. We believe that by “remembering the human” Reddit researchers can continue to engage in important research in an ethical manner.

For more details on this study, our findings, and our recommendations, see our paper published in the most recent issue of Proceedings of the ACM on Human-Computer Interaction. This work will also be presented at the ACM SIGCHI Conference on Supporting Group Work in January 2025. It was conducted as part of the broader NSF-funded PERVADE project.

Casey Fiesler, Michael Zimmer, Nicholas Proferes, Sarah Gilbert, and Naiyan Jones. 2024. “Remember the Human: A Systematic Review of Ethical Considerations in Reddit Research.” Proceedings of the ACM on Human-Computer Interaction 8, GROUP, Article 5 (January 2024), 33 pages. https://doi.org/10.1145/3633070

What follows is a transcript of the above YouTube video, where I explain public domain and copyright terms in the U.S. in the context of our favorite corporate Mouse… while you watch the entirety of the now copyright-free short film Steamboat Willie.

It is January 1, 2024 as I am posting this video. And this is a day that many people thought might never come — the day that the character of Mickey Mouse officially enters the public domain and is no longer under copyright protection. Because Steamboat Willie, the Disney short that you are watching at this very moment, was officially released in 1928. 96 years ago. And in the United States — now, anyway — copyrights expire after 95 years. Or beyond corporate ownership, for works published in 1978 or later. the life of the author plus an additional 70 years. Which of course is bound to be even longer than that. So celebrate Mickey Mouse, because you’re not going to see public domain Harry Potter or the like for a very long time.

But why did it seem like this day would never come? Because Steamboat Willie’s copyright was originally supposed to expire in 1984, 56 years after it was originally released. Because that’s how long copyright used to last… but wait, let’s back up.

What is the public domain anyway? The public domain consists of all the creative works that no one owns the rights to. Which means that anyone can legally use them without any kind of permission. And one way that works make their way there is because copyrights expire. And this is by design! In the United States, this is even in the Constitution. To promote the progress of science and useful arts, by securing for limited times to authors and inventors the exclusive right to their respective writings and discoveries. We all know that patents expire — thank goodness for generic drugs, right? — but so do copyrights. The public domain helps us to build upon the creativity of the past and also preserve that past for future generations.

The public domain is how we have so many awesome and creative adaptations of, say, Jane Austen’s Pride and Prejudice. There are currently TWO musicals based on The Great Gatsby in production for Broadway, because its copyright expired one year ago. And of course, there are tons of works that are in the public domain because they were created before copyright even existed. Shakespeare, for example. IMDB lists over 1,500 films and TV shows with Shakespeare as a writing credit. And what about the works of the Brothers Grimm? Or Hans Christian Andersen?

Snow White. Cinderella. Rapunzel. The Little Mermaid.

… huh. It seems like Disney’s success was literally built off of the public domain. But Disney has always been quite protective of their copyrights. And of their Mouse in particular.

In the 1970s, copyright protection was 28 years, with an option for another 28 years upon renewal. And Disney was one of the corporations that lobbied hard for that term to be extended. As part of the Copyright Act of 1976 — which did other important things too, like codifying fair use — there was an extension and Mickey’s new expiration date was 2004. But in the mid-nineties, the lobbying started again. And the Sonny Bono Copyright Term Extension Act — yes, that Sonny Bono, who was a congressman at the time — extended everything by another 20 years, in 1998.

And by the way, the old copyright terms did not go quietly. A constitutional challenge to this law went all the way to the Supreme Court. Eldred v. Ashcroft. And the part of their argument was: hey, the constitution? It says ‘for limited times.’ If every time Mickey’s copyright is about to expire, the term is extended, that doesn’t seem very limited. But the court disagreed. They basically said that as long as the limit isn’t “forever,” that limit can change to whatever congress wants.

So, many people thought that Disney would keep up this fight. But I personally don’t find it that surprising that this day has finally come. It was inevitable. And it would have been both a hard fight and a PR nightmare to postpone this yet again. Plus — there are plenty of ways that Disney still has control over Mickey Mouse.

One important nuance to this situation is that only Steamboat Willie hits the public domain today. And the mouse that you’re seeing right now is not quite the same mouse that you would find most familiar. He’s black and white. He’s not wearing gloves. He’s got a tail. And there’s so much personality that can come out of a 7 minute short with no speaking lines. And later versions of Mickey — for example, the one in the very recognizable Fantasia — are still under copyright. Fantasia has twelve years left.

A relevant piece of precedent for this situation might be Sherlock Holmes. Sherlock’s first appearance was in ‘A Study in Scarlet’ in 1887. And his last appearance was in 1927. 96 years ago today. So it wasn’t until last year that ALL of Arthur Conan Doyle’s Sherlock Holmes stories were in the public domain. And there were many years in which some stories were and some were not. Now most of the well known Sherlock Holmes adaptations — for example, BBC’s Sherlock — did the safe thing and got licenses from Conan Doyle’s estate. But in 2014 there was an appellate court decision in the U.S., Klinger v. Conan Doyle Estate, that clarified something interesting with respect to the public domain. The case was about a Holmes anthology. And the estate argued that because some Holmes stories were still under copyright protection, the character was off limits. But the court disagreed — with an important caveat. Characteristics of Sherlock Holmes that came from the later stories were STILL protected by copyright. So Sherlock Holmes was fair game… so long as he didn’t like dogs, and his sidekick Watson didn’t have a second marriage. This concept was also almost tested again in 2019 when the Conan Doyle estate sued Netflix over the Enola Holmes film, claiming that in part the character of Sherlock Holmes was too warm. That clearly their Sherlock, who was also more respectful of women, was clearly based on the later copyrighted character. That lawsuit fizzled out, perhaps settled, it was a little unclear.

But it’s likely that if Disney tries to guard Mickey’s copyright as much as possible, that they will make similar arguments about which versions of Mickey are being used. Will it matter if he’s wearing gloves? Maybe.

But perhaps even more importantly, Disney has a trademark in Mickey Mouse. And trademarks don’t expire, as long as they’re in use. They can absolutely still protect the mouse as a brand identifier. And you might have noticed that in recent years, Steamboat Willie has become more recognizable again. For example, Disney has been using a clip from the film as new intro logo for Disney animation. Now, courts in the U.S. have warned in the past about trying to use trademark as a sort of mutant copyright law. But how this is going to play out… could potentially be very interesting. I personally hope to see Mickey Mouse firmly in the public domain and built upon by others. But this is very likely not the end. In the meantime, enjoy this copy of Steamboat Willie, which I am now legally allowed to share with you.

Video link: https://youtu.be/lnGL55VOj_4

In the opening scene we meet our heroine: The Professor. (Some of you may recognize her.) As her final exam period ends, a classroom full of weary but excited students rush towards the doors. One of them shouts, “Enjoy your holiday break, professor!” and tosses a small candy cane to her. She tries to catch it but it hits the ground and breaks. Well. That’s about as much of a break as she feels like she’s going to get.

There is a montage of her leaving behind her big city, junior faculty, stressful life to spend three weeks in her hometown with a name that ends in “ville” and is in Georgia/Mississippi/Alabama or possibly a flyover state.

Her parents are thrilled to see her because “you work too hard!” Within seven minutes, her mother asks if she’s dating. Within nine, The Professor is trying to check her email on her phone — no cell reception. Fortunately the wifi password is her own birthday. She has 23 emails from students asking for extra credit opportunities. (As a reminder, the semester ended yesterday.)

When she opens her laptop, her father says “don’t you have three weeks off?” and as she opens the paper revision that is due to the journal days after Christmas, she just says “something like that” and five more emails come in from students complaining about grades.

The next day her mother suggests she go to a coffee shop to work. She even suggests which one. Our heroine isn’t sure why she is SURPRISED when she arrives only to find that her high school boyfriend owns the coffee shop and is a widower with a rosy-cheeked 8 year old.

It turns out that the coffee shop is in danger of closing. ON CHRISTMAS. Because, inexplicably, taxes are due on Christmas. He’s planning a fundraiser. Christmas cookies. He asks if she can help him bake the cookies. He has a red apron for her and everything. “You’re a teacher, right?” he says. “Teachers know how to bake.”

“I’m not going up for tenure on baking” she mutters, which is a joke he doesn’t understand, but she is already texting her awesome librarian sidekick/best friend to ask what she knows about tax law. Turns out she knows EVERYTHING because of some Boston Legal fanfiction she once wrote.

The Professor tells her ex boyfriend with the suspiciously muscular arms that she really needs to finish answering these emails, but that she’ll do some research for him. “But… baking?” he says, and she tells him that research is better. Really.

Outside, she runs into her former best friend whose life she knows literally every detail of because of Facebook. During some small talk, former best friend says “I can’t imagine what it must be like to have summers off AND not have kids! What do you do with all your free time?”

The Professor is so stunned by the question that she literally doesn’t remember later what she said in response (did she black out?), but she does tweet that comment verbatim to academic twitter and receives immediate validation for her outrage! Everyone reminds each other to actually take some time off over the holidays, and our heroine puts on an email auto-responder and feels better about herself.

There is a holiday party. To every person The Professor tells when asked that she is a professor, the next question is immediately “What do you teach?” and she eventually just gives up trying to explain why that is not the right question and answers vaguely, “Computer science.”

She gets tipsy on mulled wine and excitedly explains the premise of her recently accepted journal article to a group of high school friends. One of them politely asks, “How much do you get paid for one of those?” and she mumbles something and slinks off for more wine.

The next morning she emails the student who signed up to do an independent study on data science with her the next semester and had asked if there was anything he could do to prepare. She of course had told him at the time that he should relax and enjoy his break, but now says that if he REALLY wants to he can take a month off at the start of the semester if he does some work now, and outlines a project related to small town coffee shop revenues.

She also sends an email to her PhD student advisees reminding them that they should make sure to relax over the break, before opening and then closing an email reminding her that her review of a journal article is due December 26. She helps her mom bake.

In town, a white bearded man who is probably not ACTUALLY Santa tells her that instead of working so hard she should actually enjoy her life and celebrate Christmas! “WHY ARE THOSE THINGS MUTUALLY EXCLUSIVE” she shouts as she ducks into a shop to buy Star Wars wrapping paper.

In her childhood bedroom, The Professor video chats with her awesome librarian sidekick, who giggles uncontrollably at the Avril Lavigne poster positioned behind a cardboard cut-out of Han Solo. Then they discuss tax law for 45 minutes and librarian sidekick introduces her to her sister’s cat. (This is the point in the film in which you get really good ideas for Professor/librarian fanfiction.)

More family arrives. Her uncle tells her that he saw on Facebook that incoming freshmen at her university have to swear an oath of fealty to Karl Marx and that students pledge allegiance to the pride flag at the start of every class.

On Christmas Eve she receives an angry email from a student’s mother about his grade in her class and she spends an hour drafting an email that is FERPA compliant. [Author’s note: Based on a true story.] Her own mother tells her she’s getting frown lines.

The moment of truth: Christmas morning, the entire town gathers at the coffee shop for a bake sale to raise $50k in taxes that are due by midnight. The Professor swoops in with a powerpoint presentation, a jupyter notebook full of Yelp data, and a fanfiction-based knowledge of tax law.

What follows is a montage of regression analysis, trend lines, librarian sidekick piping up via FaceTime, cups of coffee being shoved at The Profsssor as she clicks to another slide, including bullet points of FINDINGS re: tax law loopholes & investment opportunities. More coffee.

Someone starts a slow clap. Someone else gives her a Christmas cookie. The coffee shop owner/her ex boyfriend looks longingly at her, as if he might be ready to propose on the spot. The rosy-cheeked 8-year-old tugs on her hand and asks if she’s going to be her new mom.

The Professor politely explains that, no, she will not be her new mom. But she hands her an entire stack of the Girls Who Code books and then helps her set up an Instagram account for her dad’s business. “You will have to help him from now on,” she says solemnly.

The ex boyfriend approaches and, beaming, says that he has arranged a job for her, teaching math at the local high school. (“Thtat’s basically the same thing, right?”) Don’t worry, you don’t have to go back to your big city, tenure-track, stressful life! You can teach here instead! And marry me! And bake on the weekends!

Behind them, the white-bearded man who is probably not ACTUALLY Santa, holds out a red apron with The Professor’s name embroidered on it. It says “Mrs.” instead of “Dr.” Her mother is near tears. The rosy-cheeked child is wide-eyed. Everyone waits with bated breath for her answer.

The Professor looks down at her phone, where her awesome librarian sidekick is still on FaceTime and thankfully is muted, because she is DYING with laughter. “SAY YES TO THE APRON,” she snorts, tears streaming down her eyes.

Our heroine clears her throat, and politely says, “I have a journal review due tomorrow, and also a paper revision the day after, and also my tenure materials go in this summer, and also have you all lost your damn minds, so I’m afraid I’ll have to be going now.”

“By the way,” The Professor adds as she backs out of the coffee shop, “I am not married to my job! Believe it or not, you can move away from your hometown AND have a successful career AND date people who support that career AND be happy! In the big city!”

“I want to be on the tenure track when I grow up!” squeaks the rosy-cheeked child, followed by the quiet sound of the librarian cheering over FaceTime. Not-Santa looks scandalized.

With the warm glow of having saved Christmas AND avoided an implausible romantic entanglement, our hero joins her family for a lovely holiday, and afterwards finishes the paper revision over a glass of mulled wine and some cookies.

Our story closes with a montage of the rest of break: revising her syllabus, ice skating with her nephew, doing a review, tweeting advice for prospective PhD students, teaching the rosy-cheeked child how to use Scratch, reading a sci-fi novel, binge-watching Nailed It, outlining a paper.

She says goodbye to her small town family and returns to her big city, tenure-track, stressful life — and her other family. We close on the first lab meeting of the semester where her newest PhD advisee beams over news of her first paper acceptance. Academia bless us, every one.

Call me, Hallmark.

This was originally shared in 2019 as a thread on the platform formerly known as Twitter. It was a sequel to the academic murder mystery that I should also archive somewhere else. :)

Also, enjoy my 2020 TikTok skit version of this story! And if you’re still feeling festive, check out The AI Bias Before Christmas.

OpenAI’s ChatGPT launched officially on November 30, 2022, and over the past year has often dominated news cycles for days at a time. In addition to my frequently updated spreadsheet of AI ethics and policy news, I have also been keeping a more systematic eye on ChatGPT. This spreadsheet contains five news stories (via a Google News search) about ChatGPT for each week for the past year, starting December 1, 2022.[1] What follows is a non-comprehensive look at major stories over the past year about ChatGPT, with an eye towards the evolution of the technology and its criticisms. (And what a season finale!)

Week 1: Enter chatbot.Within 24 hours of ChatGPT’s launch we were seeing viral Twitter threads of experiments that might seem quaint now (explain zero point energy in the style of a cat) but that at the time made people understand quite quickly what it could do, and that it could seem both brilliant and weird. But people were also immediately talking about how easily it could go from entertaining (explain how to remove a peanut butter sandwich from a VCR in the style of the King James Bible) to problematic (explain how to make a molotov cocktail).

Week 2: What is this thing? In addition to countless explainers, we began to see a whole lot of “ChatGPT wrote this!” (and similar local news segments), and even ChatGPT writing the explainers. But the “what is this?” was combined with a lot of “what does this mean?” with immediate think pieces about the potential for disruption.

Weeks 3–4: But wait… With time for some of the initial novelty to begin to wear off, we saw more deep dives into ChatGPT’s limitations. Was it “smart enough” to justify the immediate concerns around job loss? And sure, it could do some of your schoolwork, but would it be a good student? (And with final exam season upon us, there was a lot of reflection about what this meant for education.) We also started to hear about the inevitable oncoming AI tech race.

Weeks 5–7: Mitigation and misinformation. In the first weeks of 2023, the conversations about education picked up steam as new school terms started and schools had to decide what to do. This resulted in a number of outright bans of the technology in schools. School bans were an attempt to mitigate cheating, but also reflected broader recognition for misinformation and inaccuracy from the chatbot (and the difficulty in spotting it). Academic venues also began to address use of ChatGPT. But even a midst warnings about other types of potential risks, there was also still a great deal of enthusiasm for the usefulness and possibilities of the technology as more people began to use it.

Weeks 8–11: Microsoft and Google enter the chat. In late January 2023, Microsoft announced a multi billion dollar investment in OpenAI, and shortly after, integrated OpenAI’s tech into their Bing search engine. Meanwhile, Google unveiled Bard, its rival to ChatGPT. ChatGPT also hit 100 million users, making it the fastest growing userbase ever. But as the AI race sped up, we also continued to hear about the ethical implications, particularly when both the Bing and Bard launches included hallucinations and factual errors.

Weeks 12–14: Picking up speed. ChatGPT continued to “blow minds” while we heard about many specific use cases of the technology — from the good (learning assistance for visually impaired students) to the bad (church sermons that “lack soul”) to the “what do we think about this?” (the boom of AI-written books). Chinese apps removing ChatGPT also pointed to the potential for a global AI race, and the release of ChatGPT’s API opened the chatbot floodgates.

Weeks 15–17: Growing up. At three months old, ChatGPT hit a growth milestone with the announcement of GPT4, which many found immediately impressive. Though we continued to hear about some significant growing pains, like security flaws and a negative impact on the workforce. Meanwhile, Google officially released Bard and the head-to-heads began, with many users putting ChatGPT on top.

Weeks 17–21: The need for rules. Regulatory pressure heated up considerably, as did overall pushback against AI. ChatGPT was banned in Italy over privacy concerns, the EU created a privacy taskforce, and in the U.S., President Biden and Senator Schumer both made calls for AI regulation. Also OpenAI faced its first defamation lawsuit and criticism over security risks. The “AI Pause” letter suggested a temporary pause to development of more powerful models, citing the potential future risk to humanity, which also sparked criticism over a lack of attention to current AI harms.

Weeks 22–25: No seriously, rules. Regulatory pressure continued, with OpenAI attempting to assuage concerns. CEO Sam Altman testified before Congress. Italy allowed ChatGPT to return, with promises of new privacy protections. There was an arrest in China over a use of ChatGPT that violated their law around deepfakes. And as we neared the end of the spring school term, we heard that ChatGPT was killing homework assistance site Chegg, and also many teachers were trying hard (sometimes inappropriately) to catch students using ChatGPT to cheat. Also OpenAI released a ChatGPT iPhone app and a subscription service.

Weeks 26–30: Risks near and far. Over the month of June 2023, we saw recognition of current ethical issues with ChatGPT and similar AI systems, as well as discussion of long-term existential risks. A lawyer was caught using ChatGPT to write legal filings. There was another security breach and another defamation lawsuit. People were losing jobs due to companies using ChatGPT to automate work. Microsoft and OpenAI were hit with a huge privacy lawsuit. At the same time, there was another open letter, this one warning about the “risk of extinction” from AI. CEO Sam Altman weighed in on risks versus benefits of his technology, focusing on long-term threats rather than current harms. Also at the end of June we saw more movement globally as China’s Baidu claimed that its rival chatbot Ernie was outperforming ChatGPT.

Weeks 31–32: Copycat. Multiple lawsuits related to ChatGPT’s training data emerged: one as a proposed class action from authors against OpenAI for copyright infringement, another led by Sarah Silverman against both OpenAI and Meta, and another from anonymous plaintiffs against OpenAI for “unprecedented theft of private and copyrighted information.” ChatGPT’s growth also showed a decline for the first time at the beginning of July.

Weeks 33–35: Investigation and competition. The U.S. Federal Trade Commission launched an investigation over potential consumer harms and OpenAI’s security practices, and meanwhile, researchers experimented with bypassing ChatGPT’s safety guardrails. Elon Musk announced a new company, xAI, intended to compete with ChatGPT; Apple’s stock rose in response to reports that it was developing a ChatGPT competitor; and Facebook announced Llama 2, the latest version of its LLM.

Weeks 36–38: School and work. As a new school year began, schools continued to grapple with how to best handle ChatGPT. But this time, there were more reports of schools deciding repeal bans or even actively embrace it, experimenting with teaching students to use AI systems. Meanwhile, in the workforce, some companies considered banning ChatGPT due to concerns over data security, but others were making increasing use of AI continued to sound alarms over job loss. But also, listings for AI-related jobs increased dramatically.

Weeks 40–41: Useage. As OpenAI released its enterprise version of ChatGPT, promising expansion, Baidu’s chatbot Ernie jumped to the top of the Chinese app store. And though ChatGPT’s traffic dropped for the third month in a row, we also started to hear about the impacts of resource consumption in the community near OpenAI’s data centers.

Weeks 42–45: Seeing and hearing. A late September update to ChatGPT gave the system the ability to analyze images (“see”), understand voice commands (“hear”), and use text-to-voice (“speak”). And in the wake of these advancements in the tech, we also saw more conversation speculating about where the tech might be going next. In the meantime, concerns over security continued and prompted the US Space Force to halt use of AI tools, and OpenAI was hit with its most high-profile copyright lawsuit yet, from the Authors Guild featuring writers like George R.R. Martin, Jodi Picoult, and John Grisham.

Weeks 46–48: Research findings. Over time we’ve been seeing more published research on the impacts and capabilities of large language models. One study warned that chatbots could be tricked into producing malicious code. Another found that ChatGPT and other LLMs propagate debunked race-based medical information. Though yet another suggested the potential for ChatGPT to produce less biased insights for doctors (while acknowledging ethical concerns). Meanwhile, media organizations released a scathing indictment of large language models as “illegal rip-offs.”

Weeks 49–50: New stuff and downtime. One of the announcements during OpenAI’s Dev Day in early November was a new service called GPTs and a GPT store — the idea being that anyone can create custom chatbots (and commercialize them) without coding experience. Around this same time, Elon Musk also announced his own chatbot, intended to compete with “politically correct” AI systems (and not everyone was impressed by its humor). And shortly after Dev Day, ChatGPT went down for a couple of hours due to a DDOS attack.

Weeks 51–52: ChatGPT’s season one finale looked like an episode of Succession. On November 17, the OpenAI board announced that Sam Altman had been removed as CEO due to his being “not consistently candid in his communications.” CTO Mira Murati became interim CEO. Former Twitch CEO Emmett Shear then became OpenAI’s third CEO in as many days, and Microsoft announced that it had hired Altman to lead his own AI division, and almost the entire workforce of OpenAI threatened to quit. There was rampant speculation about the reasons behind Altman’s removal, many of them focusing on the tension between AI commercialization and speed, and AI safety. By November 22, Altman had returned as CEO of OpenAI, with most board members being replaced as well. There have since been reports that this shake-up was due to internal concern over a rumored AI breakthrough at OpenAI. In a letter to the company published the day before ChatGPT’s first birthday, Altman touted the company’s resilience and stated priorities towards “advancing our research plan and further investing in our full-stack safety efforts” and “building out a board of diverse perspectives.” As of November 30, the new OpenAI Board consists entirely of white men.

I will not be updating the spreadsheet of ChatGPT articles any further (consider this single year “wrapped!”) but if you are interested in my broader news round-ups, I maintain a much more comprehensive spreadsheet of AI ethics and policy news, and post periodic video news round-ups on Instagram and TikTok, and ocasionally other places. You can find all of my social media here: https://casey.prof/ My public education this academic year is supported by the Notre Dame IBM Tech Ethics Lab!

[1] A note on methodology. The database of ChatGPT articles is based on a Google News search (for “chatgpt”) in incognito mode for each seven day period. I added the top articles that appeared (in whatever way Google is ranking relevance), leaving out things like press releases and blogs. The dataset itself is obviously non-comprehensive (both because of what came up in search and because of what the media reports and how they report it), so the narrative above is as well — I used the spreadsheet to construct narrative chunks, though occasionally supplemented with related links from my broader AI ethics news spreadsheet. And of course, though I was attempting to be relatively neutral in describing what the media covered, the narrative above is colored by my own positionality as an ethics researcher and therefore how I perceived the most salient events.

As public concern about the ethical and social implications of artificial intelligence keeps growing, it might seem like it’s time to slow down. But inside tech companies themselves, the sentiment is quite the opposite. As Big Tech’s AI race heats up, it would be an “absolutely fatal error in this moment to worry about things that can be fixed later,” a Microsoft executive wrote in an internal email about generative AI, as The New York Times reported.

In other words, it’s time to “move fast and break things,” to quote Mark Zuckerberg’s old motto. Of course, when you break things, you might have to fix them later — at a cost.

In software development, the term “technical debt” refers to the implied cost of making future fixes as a consequence of choosing faster, less careful solutions now. Rushing to market can mean releasing software that isn’t ready, knowing that once it does hit the market, you’ll find out what the bugs are and can hopefully fix them then.

However, negative news stories about generative AI tend not to be about these kinds of bugs. Instead, much of the concern is about AI systems amplifying harmful biases and stereotypes and students using AI deceptively. We hear about privacy concerns, people being fooled by misinformation, labor exploitation and fears about how quickly human jobs may be replaced, to name a few. These problems are not software glitches. Realizing that a technology reinforces oppression or bias is very different from learning that a button on a website doesn’t work.

As a technology ethics educator and researcher, I have thought a lot about these kinds of “bugs.” What’s accruing here is not just technical debt, but ethical debt. Just as technical debt can result from limited testing during the development process, ethical debt results from not considering possible negative consequences or societal harms. And with ethical debt in particular, the people who incur it are rarely the people who pay for it in the end.

Off to the races

As soon as OpenAI’s ChatGPT was released in November 2022, the starter pistol for today’s AI race, I imagined the debt ledger starting to fill.

Within months, Google and Microsoft released their own generative AI programs, which seemed rushed to market in an effort to keep up. Google’s stock prices fell when its chatbot Bard confidently supplied a wrong answer during the company’s own demo. One might expect Microsoft to be particularly cautious when it comes to chatbots, considering Tay, its Twitter-based bot that was almost immediately shut down in 2016 after spouting misogynist and white supremacist talking points. Yet early conversations with the AI-powered Bing left some users unsettled, and it has repeated known misinformation.

When the social debt of these rushed releases comes due, I expect that we will hear mention of unintended or unanticipated consequences. After all, even with ethical guidelines in place, it’s not as if OpenAI, Microsoft or Google can see the future. How can someone know what societal problems might emerge before the technology is even fully developed?

The root of this dilemma is uncertainty, which is a common side effect of many technological revolutions, but magnified in the case of artificial intelligence. After all, part of the point of AI is that its actions are not known in advance. AI may not be designed to produce negative consequences, but it is designed to produce the unforeseen.

However, it is disingenuous to suggest that technologists cannot accurately speculate about what many of these consequences might be. By now, there have been countless examples of how AI can reproduce bias and exacerbate social inequities, but these problems are rarely publicly identified by tech companies themselves. It was external researchers who found racial bias in widely used commercial facial analysis systems, for example, and in a medical risk prediction algorithm that was being applied to around 200 million Americans. Academics and advocacy or research organizations like the Algorithmic Justice League and the Distributed AI Research Institute are doing much of this work: identifying harms after the fact. And this pattern doesn’t seem likely to change if companies keep firing ethicists.

Speculating — responsibly

I sometimes describe myself as a technology optimist who thinks and prepares like a pessimist. The only way to decrease ethical debt is to take the time to think ahead about things that might go wrong — but this is not something that technologists are necessarily taught to do.

Scientist and iconic science fiction writer Isaac Asimov once said that sci-fi authors “foresee the inevitable, and although problems and catastrophes may be inevitable, solutions are not.” Of course, science fiction writers do not tend to be tasked with developing these solutions — but right now, the technologists developing AI are.

So how can AI designers learn to think more like science fiction writers? One of my current research projects focuses on developing ways to support this process of ethical speculation. I don’t mean designing with far-off robot wars in mind; I mean the ability to consider future consequences at all, including in the very near future.

This is a topic I’ve been exploring in my teaching for some time, encouraging students to think through the ethical implications of sci-fi technology in order to prepare them to do the same with technology they might create. One exercise I developed is called the Black Mirror Writers Room, where students speculate about possible negative consequences of technology like social media algorithms and self-driving cars. Often these discussions are based on patterns from the past or the potential for bad actors.

Ph.D. candidate Shamika Klassen and I evaluated this teaching exercise in a research study and found that there are pedagogical benefits to encouraging computing students to imagine what might go wrong in the future — and then brainstorm about how we might avoid that future in the first place.

However, the purpose isn’t to prepare students for those far-flung futures; it is to teach speculation as a skill that can be applied immediately. This skill is especially important for helping students imagine harm to other people, since technological harms often disproportionately impact marginalized groups that are underrepresented in computing professions. The next steps for my research are to translate these ethical speculation strategies for real-world technology design teams.

Time to hit pause?

In March 2023, an open letter with thousands of signatures advocated for pausing training AI systems more powerful than GPT-4. Unchecked, AI development “might eventually outnumber, outsmart, obsolete and replace us,” or even cause a “loss of control of our civilization,” its writers warned.

As critiques of the letter point out, this focus on hypothetical risks ignores actual harms happening today. Nevertheless, I think there is little disagreement among AI ethicists that AI development needs to slow down — that developers throwing up their hands and citing “unintended consequences” is not going to cut it.

We are only a few months into the “AI race” picking up significant speed, and I think it’s already clear that ethical considerations are being left in the dust. But the debt will come due eventually — and history suggests that Big Tech executives and investors may not be the ones paying for it.

This article is from The Conversation, an independent nonprofit news organization dedicated to sharing the knowledge of academic experts. Find out more about them.

AI has social consequences, but who pays the price? Tech companies’ problem with ‘ethical debt’ was originally published in The Conversation on Medium, where people are continuing the conversation by highlighting and responding to this story.

Twas the weeks before Christmas
And Santa was busy
His list was so long
It was making him dizzy

There are too many children
List checking can’t scale
If I do this by hand
I will certainly fail!

But one of his elves
Had a brilliant idea
What if we automate
Christmas this year!

There always are kids
Who are naughty or nice
Every year for millennia
You’ve checked that list twice

Every child there’s been
From Aaron to Zeta
We can build an AI
Because we have the data

Now, Santa is old
Didn’t quite understand
But he trusts his elves
So he okayed the plan

First the elves cleaned the data
They were very precise
To determine the patterns
In the naughty and nice

Gender and parents
And address and race
Grades and faves
Really, every trace

Billions of children
Made into numbers
AI makes predictions
While each child slumbers

The goal of the model
(Just to be clear)
Given these patterns
Who’s naughty this year?

They held back a sample
So that they could test
And the model? Not bad.
Ninety percent, at their best

So they brought it to Santa
And said, never fear!
Machine learning
Will save Christmas this year

Santa pulled out his list
For his own kind of test
Hours later he went
From impressed to distressed

We can’t use this, he said
You must realize the stakes
There also are patterns
In errors this makes

Like take little Aaron
Do you think it’s fair
That in numbers he’s ‘like’
Naughty kids over there?

We’re also assuming
Our own perfect past
Have we made mistakes?
We at least have to ask

Because look at this insight
The data supplies us
Some elves on these shelves
Might have unconscious bias!

And you all made choices
Which features to add
Are you sure this was fair?
And the elves looked quite sad

They had to admit
They’d bought into the hype
But those errors did matter
And Santa was right

For now, Santa said
I’ll go back to my list
Some things need a human
To ensure they’re not missed

I do think tech could help us
You didn’t mean to abuse it
But with any great power
Be careful how you use it

So next Santa hired
An AI ethics team
We moved too fast, he said
We don’t want to break things.

A few years ago someone asked me to explain AI bias to them like they were five, and so I thought of something five-year-olds care about: Santa. It also happens that the idea of using AI to predict who will land on the naughty or nice list also serves as a metaphor for biased recidivism algorithms used in the criminal justice system. (And I’ve also always thought that elf on the shelf is a good way to teach your kids about surveillance…)

Merry Christmas to all and to all a good night, where you can dream of real people learning the same lesson that Santa did. :)

When I applied to PhD programs the first time around, I was rejected from every single one of them. I’m sure there are a lot of reasons for this, and in the end though it was a long road it’s worked out fine for me; I tried again years later, got a PhD, and now I have tenure. But in retrospect I realize that my application also probably wasn’t very good!

When I started my faculty job in a new department and we accepted PhD applications for the first time, I immediately volunteered to lead the charge. That’s been my role for seven years now, and over that time I’ve read a LOT of applications, fielded a lot of questions, and given a lot of advice. And every year, I read a large number of applications where I think: Oh no, no one told you how to do this.

PhD admissions provide an example of hidden curriculum, unspoken rules of academia where people have an advantage if they are taught things explicitly, or have mentoring. For example, something as simple as “you should refer to specific faculty members in your statement of purpose” (particularly in STEM disciplines) is not obvious to everyone — unless someone tells you. And a lot of prospective students might not even know to ask — something that disproportionately impacts first generation students, and students from groups underrepresented in their fields. I think that demystifying these things can go a long way towards making admissions more equitable and graduate programs more diverse.

During the summer of 2020 I was bored and missed talking to people, so I decided to start a YouTube channel. One of my good friends has a popular writing advice channel and she once told me that she just likes “telling people what to do — not in a bossy way, but in a helpful way.” And I thought, that’s kind of how I feel about helping graduate students!

Over the past two years I’ve made dozens of YouTube videos, many of them focused on PhD applications/admissions advice, from deciding whether you should get a PhD in the first place to writing a statement or purpose to what to expect in an admissions interview to choosing where to go if you have multiple offers.

The rest of this post is mostly links to those YouTube videos, with some of the advice distilled into very small chunks. I may write some of these out in prose someday, but in the meantime, I hope you find this helpful!

Disclaimer: PhD programs and admissions varies a lot by discipline, and by country. My number one piece of advice is to get advice from someone in your field if you can. I also try to mention in my videos when I think that something may be discipline-specific — but in general you should know that I am in a department of information science, in a subfield that is structured very similarly to computer science, at an R1 institution in the United States. I think that this advice is most applicable to STEM, social science, and other lab-based disciplines in the U.S., but much of it is more broadly applicable as well. Also every professor or department would not give you this same advice — so please do not take it as gospel, just advice that you might find nuggets of usefulness in.

Step 1: Decide if you really want a PhD.

In this video I lay out what I think are four bad reasons and four good reasons for deciding to pursue a PhD — and the advice part is both helping you think about whether this is a good path for you, but also, maybe don’t use the bad reasons in an application. (Note you don’t have to have all of the good reasons, but at least some!

Bad Reasons: #1 You’re good at school. #2 You don’t know what else to do. #3 You want to make money. #4 You want prestige.

Good Reasons: #1 You want to be a professor. #2 You love academic writing. #3 You want to add new knowledge to the world. #4 You want a research career.

If you’ve decided that you’re interested in a PhD then you might be thinking of how to improve your chance of getting into the program of your dreams — so if you’re still at this point, you should think about what you can do as an undergrad if you want a PhD. You also might be thinking about how pursuing a PhD will affect your finances, so you should make sure you understand PhD funding models in your discipline. (For example, typically for sciences and often in other fields as well, you do not pay tuition for a PhD and receive a stipend.)

Step 2: Apply to PhD programs.

Putting together a PhD application can be a lot of work, and it has a lot of moving pieces! I made the video above first with my general advice, and after receiving tons of questions about nuances and details, I made more on specific topics. So here is my very high level advice, with links to some other videos with more details.

1. Do your research. This includes deciding where to apply, finding out about the research areas and faculty in that department/program, maybe contacting potential faculty advisors/supervisors before you apply, and making sure you understand the application requirements. For example, you may or may not need test scores, writing samples, etc., and in general you should just always make sure you’re following all the requirements for each application. Also don’t miss deadlines!
2. Choose appropriate letter of recommendation writers. The best letter writers can speak to your potential as a graduate student or researcher. There’s a lot of nuance here, and I also have some recommendations in the video for how to ask, what kind of information to give them, and how to help keep everything organized.
3. Write a strong statement of purpose. This video highlights five things that a good SOP should do: Convince them you want a PhD, describe research experience & motivations, show passion for a topic, explain why that program (including faculty/research match), and communicate well. In terms of other things you might need to prepare, you may or may not need an academic CV.
4. Don’t stress about single metrics. Applicants constantly send me questions like “here are my test scores / GPA / research experience” and ask if they have a chance of being admitted into a PhD program. My answer is always “yes.” Strengths always make up for weaknesses. (Also I’m so glad that so many programs are doing away with GRE requirements!)
5. Know that sometimes it’s not you, it’s them. There is a huge amount of randomness in the PhD application process (similar to the academic job market), and sometimes it’s honestly just bad timing. Particularly in programs where admitted students have to be matched to advisors, not being admitted might mean nothing other than that the only advisor match for you is out of funding or at capacity with PhD advisees. So sometimes there are reasons you might be rejected from a PhD program that have nothing to do with how great you are! This is why I always suggest applying to multiple programs — and also that if you try one year and aren’t successful, you might have much better luck if you try again! There’s also a bunch of other things that I think don’t matter as much as people think they do, and I cover some of those in this FAQ video.

And then of course, once you apply… all you can do is wait.

Step 3: (Maybe) Interview/Visit.

Interviews or pre-admissions campus visits are not standard to every program. But the above video gives you a sense for the kinds of questions you might be asked in a formal interview or even an informal discussion with a prospective advisor, and how you might want to think about answers: 1. Why do you want to get a PhD? 2. What drew you to this program? 3. What kind of research do you want to do? 4. What are your strengths and weaknesses as a researcher? 5. Tell me a little about yourself…

You might also have an interview combined with a campus visit, or an open house prior to admissions, in which case you should know what to expect and how to make the most of it!

Step 4: Make a choice.

The best position you can be in is to have multiple offers of PhD admission so that you can decide on the absolute best fit for you! It’s important to know that this is a very personal decision, people have very different priorities, and whatever is important to you should be what you most take into account. But here are some things you might consider: Who will your advisor be? (This one is so important it gets its own video!) What do you think of the faculty overall? What are the structure/requirements of the program? How is the culture/community? How much do you care about prestige (and should you)? Does location matter to you? How is the funding package?

And… that’s it. A million easy steps to starting a PhD program.

It’s also very possible that, like me the first time around, you aren’t successful, and do not receive any offers of admission. This might be some time for soul searching, but as I said before, it honestly could just be bad timing. I do have some suggestions as well for what to do next if you’re rejected.

You can find my entire PhD admissions playlist here, and also — once you’re in a PhD program I have videos about graduate school and academia in general as well, including advice for new PhD students! And I do livestreams on YouTube occasionally, answering questions; you’ll find them announced on my channel (at the time of this writing my next one is scheduled for 11/13/22).

Also if you prefer short-form content, music, and memes, I have a playlist of PhD application advice videos on my TikTok, @ professorcasey.

I hope that you find all this helpful! Though I can’t provide individual advice on application materials (and you’d be better off asking someone in your specific field anyway!), I’m really glad that these videos can continue to help people every year. So if you’re somewhere in the PhD admissions process: Good luck to you! See you on the other side. :)

When the AI conference NeurIPS began requiring papers to include a broader impacts statement that includes “ethical aspects and future societal consequences,” there was a fair amount of pushback, and one social media comment I distinctly remember was along the lines of: How do they expect us to see the future?

Isaac Asimov once said, “Science fiction writers foresee the inevitable, and though problems and catastrophes may be inevitable, solutions are not.” It is true that science fiction fiction writers are very good at seeing the future, particularly when it comes to things like the unintended consequences of technology. However, it also doesn’t tend to be the job of science fiction writers to come up with solutions — that is a task that often falls to technologists. So something I’ve been thinking a lot about lately is: How can we help technologists think more like science fiction writers?

Now to be clear, when I say “science fiction,” I don’t mean that technologists should be designing towards sentient AI or the robot wars. I mean the ability to consider future consequences at all — including (very) near future. Speculation is a skill that can be practiced, and science fiction can be an engaging way to do that. But: if technologists can get excited about imagining some harm caused by technology someone might create in 50 years, why can’t they do the same thing with what they create tomorrow?

In 2018 I wrote about two teaching exercises that I’d been using in my tech ethics and policy class, including the Black Mirror Writers Room, which I subsequently turned into a slide deck that others could use to run the exercise. Over the next few years I heard from many instructors who had used this exercise with students from high school to PhD and in many different kinds of computing-related classes. Hoping to learn more about the benefits and challenges of this exercise, as well as using ethical speculation in the computing classroom generally, PhD student Shamika Klassen interviewed 12 of these instructors, and our paper based on this study was just published at the ACM SIGCSE 2022 computer science education conference: “Run Wild a Little With Your Imagination”: Ethical Speculation in Computing Education with Black Mirror.

Overall, the instructors we interviewed were extremely positive about the benefits of this teaching exercise for their classrooms. They spoke about the positive responses students had to the opportunity to be creative and collaborative, and that it was a fun way to make ethical topics engaging and salient. As one instructor said:

“It was a really great way for them to express their creativity, plus their knowledge of our domain space that we had been talking about… It just shows this synthesis of the knowledge that they’ve built up over the quarter about these types of technologies in this context.”

For the past few years my students and I have also been conducting research about pedagogical strategies for teaching ethics and related topics in computing. Our interviews also revealed how this exercise could work well as a way to introduce ethical thinking in technical computing classes. One instructor also noted how students were “shocked to be asked to think about these issues and be creative,” and another elaborated:

“They know how to fix your computers, they understand how to code, they understand everything that there is to know about software, but they don’t get any ethics most of them and they just really liked it. They really liked just thinking that way.”

However, despite the overall positive sentiment towards this pedagogical strategy, instructors also pointed out a number of challenges, including the lack of scaffolding around the ethics instruction for them, and the lack of cultural context (around Black Mirror or science fiction generally) for some students. Based on this feedback, students and I are working on providing some additional resources for specific domain areas, as well as ethical speculation exercises that do not rely on Black Mirror. For example, as part of a research project on AI+cybersecurity education, PhD student Samantha Dalal and I created this exercise (please feel free to copy and adapt this slide deck for yourself!) for a computer vision course that introduces types of threats and then has students write future news headlines.

Finally, an important insight from our interviews confirmed a caution around ethical speculation as an approach that had already begun to take shape for me. We asked instructors about justice-related themes in the Black Mirror episodes students came up with, and some instructors noticed such themes — most often around class inequality or gender discrimination, but none of the examples instructors recalled involved race. We also asked instructors about the demographics of the classes in which they used the activity, and most noted that there were few BIPOC students present in these classes. This led us to consider the limitation of how different types of diversity in the classroom might result in different kinds of stories, and one of our interview participants had this critical insight about the limitations of speculation:

“I tend to be cautious about speculation in part because people tend to lean on their own experiences pretty heavily in speculation, and don’t, unless, they’re very carefully prompted, consider broader context. I worked as a user experience designer. . . and I would see use of like personas or other kinds of vaguely speculative exercises where you’re asked to imagine the ideal user or imagine ideal kind of use cases. And almost universally people would be like, the user is a middle-class white person and here’s what they’re going to do. . . you are just reifying all the stereotypes. And so I think that’s the danger with speculation done well: it’s a great opportunity to disrupt the stereotypes, but I think it takes a lot of work to do that right. . . In more diverse classes, you tend to get a lot of different viewpoints in a lot of discussions, although there is a tendency for people who are in kind of minority populations to suppress their viewpoints to some degree in order to not stick out in the majority. So that’s always a risk and that’s a discussion we have on the first day of class in my class.”

In other words: It is inevitably easier for people to imagine futures for and harms to people like them. So given our diversity problems in the computer science classroom, there is a real risk of a student-created season of Black Mirror that speculates about technological harms to middle class white people. And in fact I realized my own blind spot here. The Black Mirror episodes I was using as examples were indeed all about middle to upper class white people:

We are working on continuing to create more scaffolding for this and related speculation exercises to bring in e.g. principles of design justice for learning from and involving people who are different than you. In the short term I have added slides to the current exercise to point out this issue along with the reminder that technological harms disproportionately impact marginalized groups. But in general in using speculation in the classroom, we caution instructors to ensure that they avoid reinforcing stereotypes and consider ways to encourage students to think broadly about injustice.

I have written before about the importance of touching on ethics throughout computing curriculum in addition to students learning in standalone classes from experts, and I hope that this exercise helps computing students understand that thinking through potential consequences and harms is part of what it means to design and build technology. I am also thinking about the potential power of this approach beyond the classroom, and my recently funded NSF CAREER grant has as an outcome an “ethical speculation toolkit” that can help practitioners reduce ethical debt in the design process. And again, the point here is not to encourage students or practitioners to actually think like they’re writing for Black Mirror — but instead to pull back and use the same skills to think about tomorrow just as easily as fifty years from now.

Finally, I want to point out that beyond commentary about the substance of the exercise itself, a number of our participants also offered (unprompted) how helpful it was to have come across a teaching exercise that was openly shared, and could be “plug and play” or adapted to fit a particular instructor’s needs. The majority of participants said that they heard about the exercise from Twitter, either as originally shared by me or by someone else who had used it. Multiple participants mentioned how useful it was to have instructions, examples, and a template, both in terms of running the exercise and in students using the slide template to create artifacts. I hope that this is encouraging for instructors who might consider sharing teaching materials so that others can use and build upon them!

Thank you to the Responsible Computer Science Challenge for partially funding this research, to NSF for supporting my work in this space moving forward, and especially to the instructors who generously gave us their time and insights. You can read Shamika’s and my full paper here:

Klassen, Shamika, and Casey Fiesler. “Run Wild a Little With Your Imagination”: Ethical Speculation in Computing Education with Black Mirror. In Proceedings of the 53rd ACM Technical Symposium on Computer Science Education, pp. 836–842. 2022.

At this link you will also find more resources on the Black Mirror Writers Room, and for more on my work related to ethics in computing education, check out my lab’s webpage of writings and resources or the video below:

And for more examples of the amazing things that students in my classes have come up with over the years, some TikToks. :)

The Black Mirror Writers Room: The Case (and Caution) for Ethical Speculation in CS Education was originally published in CUInfoScience on Medium, where people are continuing the conversation by highlighting and responding to this story.

When I graduated with my PhD in the summer of 2015, I’d written a 200-page dissertation and also accumulated almost 30,000 meaningless internet points known as yakarma. Anonymously, on the location-based app Yik Yak, I complained about my university’s wifi, shared photos of campus squirrels and cookies I’d baked, argued about feminism and politics, gave relationship advice to twenty-year-olds, consoled students upset about failing grades, asked for recommendations for coffee shops when I traveled, and ruthlessly downvoted racist and homophobic jokes. Of course, at the same time, many Yik Yak users were engaging in cyberbullying, harassment, and even bomb threats — and well, the racism and homophobia was getting plenty of upvotes, too.

In 2015, Yik Yak was one of the most popular social networking platforms on college campuses, and it was valued at close to $400 million. By the end of 2017, it was gone. Many blamed its demise on significant design changes — notably, doing away with anonymity by moving from first optional and then required user handles, which arguably destroyed a differentiator that had drawn many users to the platform. However, it also had persistent problems with toxic content that led to some universities banning its use.

Now, quietly and without much fanfare, Yik Yak is back — and its reputation precedes it. At least one university has already banned the app, and I’ve been thinking about what the problem with the app was and whether we’ll still see those problems today. Or maybe part of the issue is that we’ve got all new ones.

Like many online platforms, Yik Yak wasn’t strictly good or bad. It was complicated. In 2015, in the wake of calls for universities to ban the platform due to harassment against marginalized groups, YikYak was also being used for discussions of stigmatized topics, exploring identity, finding support, and giving marginalized students a voice through organization of activism. This tension, in which the same features that allow harassment or hate speech to thrive can also allow for critical anonymous social support, is true for pretty much every social platform. I could spend fifteen minutes on Reddit, Tumblr, or TikTok, and find examples of both the best and worst of what humanity has to offer. Anonymous online interaction absolutely leads to the kind of awful people being awful you might imagine, but it also provides ways for people to connect when they might be fearful of being outed, or to open up about stigmatized topics. In other words: anonymity isn’t the problem, people are.

So with stars still in my eyes from spending five years on a dissertation about one of the more positive and supportive parts of the internet, I wrote a piece for Slate about what was wonderful about Yik Yak. I wrote about how many parts of Yik Yak were doing a good job policing themselves, and how the platform design around upvoting and downvoting allowed for formation of positive social norms. “It isn’t a given that people will be terrible,” I wrote. “Sometimes they might surprise you.”

Six years later upon hearing that Yik Yak was back, my first impulse was dread. So I asked myself: what changed? One of the answers might be that six years as a professor researching technology ethics has changed my own perspective. But I also think that many of us feel worse about the internet than we did in 2015. Maybe we’ve seen even more evidence that, given the opportunity, people will be terrible — and that our platforms aren’t as well equipped to handle it as we’d like.

The Wall Street Journal recently released a barrage of investigative reports about the inner workings of Facebook and specific, concrete harms related to the platform: how Instagram can be damaging to mental health, especially for young girls; how algorithms can fuel outrage; persistent challenges for content moderation; and how vaccine misinformation thrives. Misinformation on social media especially is objectively a much more profound problem than it was five or six years ago — not only because of how easily it spreads, but because of the stakes. You thought misinformation’s impact on democracy was a problem? Well, just wait until we’re in a global pandemic!

These kinds of big problems with other platforms, particularly as related to content curation and recommendation algorithms, are indicative of how we currently interact most with social media. We’re entirely accustomed to algorithms deciding what content we should see. And YikYak is in a way an example of what people say they want instead — a content feed curated by people. Content floats to the top if upvoted, and disappears if downvoted. But given the problems that YikYak faced with respect to negative content, I don’t know that there’s reasons to think humans are necessarily better than algorithms in this respect.

The relaunch of Yik Yak came with promises for “community guardrails” and their guidelines do clearly state that posts with bullying and harassment, bigotry, violence and threats are not allowed on the platform, as well as posts that “knowingly share fake news, unless it’s obvious satire.” However, moderation is challenging especially at scale, and my own previous optimism about the benefits of algorithmic content moderation on Yik Yak has definitely faded thanks to my research on content moderation, most recently on TikTok. The problem with algorithmic flagging of content like hate speech is that algorithms will never be perfect, and they are known to be biased. Platforms like TikTok are struggling to find an appropriate calibration between false positives and false negatives — and either way, the errors tend to disproportionately harm people from marginalized groups. So just as Yik Yak previously could serve to both help and harm marginalized students on college campuses, depending on very nuanced aspects of their moderation practices, their voices could be either uplifted or suppressed.

You might be wondering if my dread was well-founded. What is Yik Yak like in this new digital moment? Honestly… so far I’m not impressed, and polling my colleagues at other universities suggested they’re seeing much of the same. A lot of tasteless jokes with underpinnings of sexism, racism, and homophobia. Mean-spirited call-outs of individuals and groups. And unfortunately, not as much supportive content to even it out. Of course, it’s still early days and perhaps early adopters like to push boundaries and see what they can get away with. On the bright side, I personally haven’t seen COVID-related misinformation on the app. Also I hope the 20 people who upvoted my supportive message felt slightly better because of it.

Though I also hear that uptake of YikYak simply isn’t that high — a lot of the college students I’ve asked haven’t heard of it at all, or folks my age are surprised to hear it’s back. Others told me they downloaded it briefly and then didn’t go back.

So maybe the ways we expect to consume content and interact on social media really has changed since 2015. This brings with it challenges for platforms to meet current engagement norms (see e.g. how Facebook and Instagram are constantly adding new features that look a lot like features from other platforms), but also challenges for new kinds of ethical issues and new forms of harm. And whatever comes next, it’s almost certainly not going to be strictly good or bad — it’s complicated.

This post touches in part on two papers out of my research lab being presented this week at SIGCSE 2021: “Integrating Ethics into Introductory Programming Classes” and “‘You don’t do your hobby as a job’: Stereotypes of Computational Labor and their Implications for CS Education.” Post hoc, I realize that something at the heart of both of these projects is the potential benefits of expanding our notion of what “counts” as computer science.

As reported by Dianne Martin in a 1997 paper, a computer science department chair wrote a letter to the editor of Communications of the ACM in response to a proposal that social and ethical impact be integrated as an explicit strand into CS education in 1996:

… the most glaring problem is that proposed subject matter is not computer science…the content of the ‘strand’ has no algorithms, no data structures, no mathematical analysis, neither software development nor software design, no computer science theory. In short the content is devoid of every standard element present in computer science research and education. … It’s hard to imagine a computer scientist teaching these things. … Ethical and social concerns may be important, but as debating the morality of nuclear weapons is not doing physics, discussing the social and ethical impact of computing is not doing computer science.

The idea that ethics is not “doing computer science” persists over two decades later. We see evidence of this attitude every time we hear some variation of “I’m just the engineer” from a tech professional when questioned about ethics, and when ethics is barely covered at all in computer science curriculum. (Which yes, is still the case in many departments despite ABET requirements — I hear from college students all the time who say they’re not hearing about ethics at all in their CS program.)

A couple of years ago I suggested (“What Our Tech Ethics Crisis Says About the State of Computer Science Education”) that if we want to shift computing culture such that thinking about the ethical impact of computing is part of the role of every technologist, we have to change the way we teach computer science — not as a specialization or an add-on, but integrated into technical classes throughout the curriculum. Since then, the Responsible Computer Science Challenge provided funding for myself and many others to put this idea into practice. In thinking about the importance of this cultural shift, my collaborators and I decided to start at the very beginning: with introductory programming classes.

So all of this is to say, I think that if we want to thread critical ethical inquiry deep into the tech industry we need to start with education, and that means showing future technologists every step of the way that thinking about the ethical and social impact of computing is not only “doing computer science” but something that must be done in computer science. I also think that computer science departments should be hiring tenure-track faculty for whom computing ethics is their area of research. And I think that coding interviews at tech companies should include questions about ethical implications. (I have a lot of thoughts about the role of interviews but that’s a whole other post. :) )

It’s also worth noting that as we continue to grapple with the lack of diversity in computer science departments and in Silicon Valley, a huge number of the thought leaders in computing ethics are women and people of color. As someone who has degrees with “human-centered computing” and “human-computer interaction” in the title lines, and who currently writes more tweets about ethics than I write lines of code, I’ve certainly had enough people tell me that these “human” aspects are the “soft” part of CS, not “real” CS, etc. Which brings me to the connection I drew to our other paper at SIGCSE this year.

My PhD advisee Brianna Dym led an interview study with fan creators (majority women and/or queer and/or BIPOC) who do computational work — that is, making websites, modding video games, doing data science about their own communities, or even building massively successful online platforms. In my prior work on computational learning in fandom, participants also described the learning environment as “empowering” and “loving” and contrasted it to open source communities. This work motivated the current project (which is NSF-funded) — to explore how people from groups traditionally underrepresented in computing are still “doing” computing that might not look like our traditional notions of computer science.

The stories we heard from asking participants about their computational projects and how they saw them as part of computing tracked to a lot of what we already know about damaging stereotypes in computing: that computer science isn’t fun or creative, that it’s not for them. And these ideas were reinforced by experiences with CS education. A refrain we heard was: I took a computer science class, and it sucked, so this awesome thing I’m doing can’t be that. For example, one participant described dropping out of a computing camp because she felt condescended, talked down, and mansplained to; she contrasted this to learning things in fandom, which felt “empowering.” And though sometimes this alternate learning space led to participants pursuing careers in tech, other participants expressed not being “good” with computing even as they talked through the computational skills they learned in fandom. In undervaluing their skills (just modding video games, just writing HTML, it’s not “real” code), someone might not see computing as something they had already started and might continue to pursue. Overall, our findings suggest that it is important (e.g., in the computer science classroom) to break down not only what a computer scientist is but also what counts as computer science — particularly in terms of encouraging participation from traditionally underrepresented groups. And also, to potentially meet them where they are.

And as one participant put it: “If you want to broaden your pool for computing, you have to look in unconventional spaces. If you stay in the conventional space, you’re going to get a lot of white dudes.”

Or, as my favorite coding drag queen on TikTok says, “Computer science professors are like, women don’t want to learn how to code. Maybe they don’t want to learn how to code from you.” In fact, I liked this sentiment so much that I build on it to explain our findings!

In sum, I think there are a lot of reasons why reducing gatekeeping in computer science would be a positive thing. And when computer science departments are considering what to teach, what kinds of research is encouraged, and who they hire, broadening that definition might also broaden the pool of participation.

Fiesler, Casey, Mikhaila Friske, Natalie Garrett, Felix Muzny, Jessie J. Smith, and Jason Zietz. “Integrating Ethics into Introductory Programming Classes.” In Proceedings of the 52nd ACM Technical Symposium on Computer Science Education (SIGCSE’21). New York, NY, USA: ACM. 2021.

Dym, Brianna, Namita Pasupuleti, Cole Rockwood, and Casey Fiesler. “‘You don’t do your hobby as a job’: Stereotypes of Computational Labor and their Implications for CS Education.” In Proceedings of the 52nd ACM Technical Symposium on Computer Science Education, pp. 823–829. 2021.