Peitho

Inspiration

In today’s polarized world, real dialogue across ideological divides has become rare. We were inspired by the idea that technology—particularly AI—could help create safe, structured, and meaningful conversations between people of different political and social beliefs. Peitho is designed to bridge those divides by allowing users to step into a conversation with avatars that embody contrasting perspectives, helping them better understand other viewpoints and refine their own.

What it does

Peitho allows users to choose a region from the political compass (e.g., authoritarian-left, libertarian-right) and then select an AI opponent that represents the typical views and personality associated with that quadrant. The user is then immersed in a live debate-style conversational interface with a talking avatar powered by AI. Through voice and text interactions, users can challenge the avatar, ask questions, and receive articulated counterarguments. This creates a dynamic space for civic education, empathy-building, and critical thinking around complex issues.

How we built it

We used a combination of technologies to build Peitho:

• A political ideology classifier and persona system to model and generate coherent opposing viewpoints.
• OpenAI's language models for generating dialogue based on the selected ideological stance.
• Text-to-speech (TTS) and speech-to-text (STT) APIs to enable fluid spoken conversations.
• A web-based front end with WebRTC and animation layers to present the AI avatars in real-time.
• Flask and WebSockets for managing state and handling real-time interactions.

Challenges we ran into

• Persona fidelity: Ensuring that each AI persona represented a nuanced, internally consistent ideological viewpoint rather than a caricature.
• Speech synchronization: Making avatar lip movements realistically sync with real-time generated audio.
• Bias and safety: Avoiding offensive or extremist content while maintaining authentic and challenging dialogues.
• Latency: Optimizing the voice interaction pipeline to feel conversational, not robotic or delayed.

Accomplishments that we're proud of

• Built a fully interactive, real-time debate experience with live avatars and multi-modal communication.
• Developed a modular political persona engine that scales across ideologies and topics.
• Created a space where people can safely engage in contentious issues without feeling attacked or silenced.
• Received positive feedback for the system's ability to increase user empathy and perspective-taking.

What we learned

• People are more open to challenging views when the conversation is framed as exploratory rather than adversarial.
• Users appreciated being able to "talk it out" instead of only reading opposing opinions.
• Modeling ideological positions requires careful curation and guardrails to prevent AI hallucination or radicalization.
• Designing for empathy and disagreement simultaneously is both powerful and difficult—but worth the effort.

What's next for Peitho

• Mobile and VR support for immersive and accessible conversations on the go.
• User-uploaded questions and scenarios to drive debates around locally relevant issues.
• Community moderation and feedback loops to continually improve persona accuracy and safety.
• Educational integration to help schools and civic organizations use Peitho as a tool for teaching critical thinking and civic discourse.
• Localized political models for different countries and cultural contexts to expand the platform’s reach globally.

Built With

• code