Meta and Stanford’s Deliberative Democracy Lab Release Results from Second Community Forum on Generative AI
Meta and Stanford’s Deliberative Democracy Lab Release Results from Second Community Forum on Generative AI
Participants deliberated on ‘how should AI agents provide proactive, personalized experiences for users?’ and ‘how should AI agents and users interact?’
In October 2024, Meta, in collaboration with the Stanford Deliberative Democracy Lab, implemented the third Meta Community Forum. This Community Forum expanded on the October 2023 deliberations regarding Generative AI. For this Community Forum, the participants deliberated on ‘how should AI agents provide proactive, personalized experiences for users?’ and ‘how should AI agents and users interact?’
At a high level, Meta used this Forum to:
- Expand public input into AI development beyond the Global North, and into the Global South. This latest Forum involved roughly 1,000 people from India, Turkey, Nigeria, Saudi Arabia, and South Africa.
- Push the boundaries on the topics that the public will have input into. We moved from the foundational principles people wanted to see in GenAI towards addressing specific value and risk tradeoffs associated with issues like personalization and human-like AI.
The Forum resulted in several key findings on the principles that should underpin AI agents, including:
- Participants supported AI agents remembering their prior conversations to personalize their experience, especially if transparency and user controls are in place.
- Participants were more supportive of culturally/regionally-tailored AI agents compared to standardized AI agents.
- Participants were in favor of human-like AI agents that can respond to emotional cues.
- Across topics, participants consistently favored options for AI to include transparency and user control features.
Maturing our Community Forum Program
Beyond the findings of any one Forum, the Deliberative Democracy Lab and Meta have heard important feedback from stakeholders and have implemented several programmatic changes to mature our program. These include:
- More disclosure around the impact of results: Meta will share more information about how results are being actioned within the company on its Transparency Center page, which will be updated throughout the year.
- Following up with participants: We heard the importance of going back to participants to explain what we learned from their input and what we are doing with it. The Deliberative Democracy Lab will be hosting calls with participants from each of our past Community Forums, dating back to 2022, to update them on the findings from the Forum and Meta’s response.
- Supporting AI deliberation: A team of Meta AI experts has begun partnering with the Deliberative Democracy Lab to conduct research on how AI might further scale deliberation and optimize the Community Forum process. This includes, but is not limited to, using AI to aggregate themes that are emerging in discussions in real time and support engagement between participants and experts in plenary sessions.
- Supporting external research: Meta is supporting a consortium of independent researchers from around the world who will evaluate the data from its Forums and publish research papers on the deliberations and results. This will culminate in an academic conference later this year.
