Questions:
Democratic Inputs for OpenAI: AI Policython
Additional Members
How long has your team been working in the democratic / consensus-building space?
a. Lucas: Lucas has previous experience working on democratic data space and has assisted research in about a dozen different research projects, mostly at Harvard. He also launched the Coronavirus Visualization Team during the pandemic and worked on government interventions for COVID at Opportunity Insights and USA Facts. Additionally, he has organized hackathons focused on consensus building for COVID response and has worked with nonprofits like Voters Choose.
b. Jeremy: Jeremy was a national organizer for the Sunrise Movement, where he led citizen actions to push for the Green New Deal. He conducted public narrative trainings and organized citizen assemblies.
c. Scott: Scott worked on a science nonprofit called Retina with over 6,000 affiliates. He has also been involved in student research and startup ventures.
d. Shayan: Shayan is a student at Columbia University studying data science and public policy and the Founder & Executive Director of the tech nonprofit Reach4Help, operating in 38 countries.
Please tell us in a few sentences about prior projects done by each of the team members that they are most proud of.
Also, launching the Coronavirus visualization team during the pandemic and working on government interventions for COVID at Opportunity Insights and USA Facts. He has also organized hackathons focused on consensus building for COVID response and has worked with nonprofits like Voters Choose. Furthermore, raising over $2m for his Web3 startup, DAOHQ.
b. Jeremy: Jeremy takes pride in his role as a national organizer for the Sunrise Movement, where he led citizen actions to advocate for the Green New Deal. He conducted public narrative training and organized citizen assemblies, contributing to grassroots efforts to address climate change, and also worked with the office of Nancy Pelosi.
c. Scott: Scott's notable project involvement includes working with Retina, a science nonprofit with over 6,000 affiliates. Although specific details are not mentioned, his contribution to this organization demonstrates his dedication to advancing scientific research and collaboration.
d. Shayan: As the head of Reach4Help, Shayan leads efforts to coordinate disaster response efforts for 10K+ NGOs and charitable organizations worldwide in response to local crises—from running food distribution campaigns across 38 countries during the COVID-19 pandemic, to raising over €170K for medical supplies for Ukraine with the World Economic Forum's Global Shapers, to serving over 60K households affected by the recent floods in Bangladesh and Pakistan. For Reach4Help’s work, he has been invited to speak at Google, the UN Headquarters, and the Aspen Institute’s Future Leaders Climate Summit. Previously, he worked as a data analyst with the Bangladesh Ministry of Health, analyzing COVID-19 data to inform governmental policy on lockdowns, resource allocation, and public health communications.
Please list any other commitments (work/school) that each of the team members has until October 20, 2023
Are any team members covered by noncompetes or intellectual property agreements that overlap with this project? Will any team members be working as employees or consultants for anyone else?
As part of his work at Cerebral Valley, Lucas helps organize hackathons with AI companies monthly and demo days with venture capitalists weekly.
Question:
Why did you pick this question (or questions) to work on?
As a team, we selected this question because it aligns with our mission to establish a novel framework for answering questions and guiding decision making. Our primary focus lies in addressing politically significant inquiries, known for their divisive nature, and seeking out principles that resonate with a broad spectrum of individuals. We believe that by delving into these complex topics, we can identify approaches that work for the majority, fostering a sense of inclusivity and understanding.
Moreover, we hold a particular interest in questions posed by students. We recognize the vibrant energy and multitude of opinions that students possess, which we consider a valuable resource in our pursuit. By engaging with student questions, we anticipate a wealth of diverse perspectives and high-quality responses. Students, being representative of various backgrounds and experiences, bring forth a rich tapestry of ideas and insights that can significantly enrich the dialogue.
We also acknowledge the significance of policy in the lives of students. Recent events, such as the Supreme Court decision, have highlighted the influential role of policy in shaping not only the educational landscape but also the broader world, including the application of artificial intelligence. Understanding how policies impact learning environments and the utilization of emerging technologies is crucial for ensuring a well-rounded approach to problem-solving.
By selecting this question, we aim to facilitate greater public participation and input. We recognize the inherent value of this format in its ability to scale up the number of submissions and reach a wider audience. Leveraging our expertise and resources, we feel a responsibility to explore new avenues and push the boundaries of engagement. Students, who hail from diverse backgrounds and exhibit exceptional levels of engagement, have the potential to offer a diverse range of insights and perspectives. Through this collective effort, we anticipate generating a more comprehensive and robust set of responses that can drive meaningful change.
In conclusion, our team's decision to focus on this particular question stems from our commitment to building an inclusive and impactful framework. By embracing politically important and polarizing topics, particularly those posed by students, we aim to harness the power of collective intelligence to shape a better future. We believe that through this process, we can elicit a range of perspectives, drive policy discussions, and promote a deeper understanding of the issues at hand.
Why do you think that this question (or questions) are well suited to broader public input? What do you think labs, developers or others might change as a result of input on these questions?
First and foremost, we have observed the successful outcomes of policy hackathons, such as the ones organized by our team at Policy for the People. These events have demonstrated the power of engaging a wide range of individuals in the policymaking process and leveraging collective intelligence to crowdsource innovative solutions to local problems. Through events like the Vizathon, Pandemic Policython, Education Policython, Merge'22, and Researchathon, we have witnessed the immense potential of bringing together diverse voices and expertise to generate impactful policy suggestions.
The experiences we have gained from running these events, including my personal involvement in leading some of them, have reinforced our belief in the effectiveness of public input. For instance, one of our policy suggestions was taken to Opportunity Insights, a renowned research lab, resulting in a subsequent paper reaching President Biden's desk and leading to the reallocation of $20 billion in federal funding. These tangible outcomes demonstrate the power of harnessing the ideas and perspectives of the public, particularly students, in shaping significant policy decisions that affect billions of dollars, millions of individuals, and numerous corporations.
In the realm of developers, we have also organized the annual Merge: Hack+Policython, where participants are challenged to submit both technical and non-technical proposals. Despite the complexity of this task, we have consistently seen a high level of engagement, with over $25,000 in prizes awarded and around 200 participants involved. The submissions span a wide range of topics, from hormone therapy on native reservations to quantitative easing in Europe, highlighting the diversity of ideas and perspectives generated through this collaborative process.
Additionally, our participants represent a significant underrepresented demographic in politics, with a strong emphasis on inclusivity and diversity. Our Policython applicants have included 65% female participants and 73% identifying as BIPOC, while projects have spanned over 400 schools and an average of three countries per project.
The unique perspective of students, who form a vocal and active part of the public online, is particularly well-suited to broader public input. As exemplified by our involvement in designing The Residency Education Designathon, where dozens of tweet submissions were received on improving the education system and the winner had the opportunity to pitch their ideas to Sam Altman, we have witnessed the enthusiasm and valuable contributions students can bring to policy discussions.
While we plan to begin with mostly T20 university students, we plan to expand to colleges around the nation, world, and then beyond with non-profits and community-based organizations.
Process
Participant selection: How do you plan on obtaining a sample of participants for your experiment? How do you think about questions of representativeness and how they might matter for your question and method? Note: OpenAI can advise on methods or resources for obtaining a sample.
We aim to organize AI Policythons engaging diverse, global participants in developing policy proposals. Our goal is building sustainable solutions enabling continued impact. We will use AI to review and summarize submissions, identifying key insights and issues, and match proposals to stakeholders. We aim for outcomes influencing policymakers and companies.
Policy for the People has partnered with groups across the political spectrum, institutions, governments, nonprofits and student projects. Our open-sourced events welcome anyone. We aim for beginners and diverse participants, with tangible outcomes. Impact should benefit disadvantaged and underrepresented groups.
Inspired by Major League Hacking, we want to build a university network for policythons. We recently got a grant from Plurality Institute and pitched policythons to Harvard. We admire new think tanks aiming to find diverse student opinions. We've built a policython toolkit, advise other clubs, and worked with Rice/Columbia students on a health policy hackathon and the University of Toronto.
We've partnered with public policy groups advising on inclusion. We want to work with more minority-serving schools and community organizations. At the moment, participants span 400 schools.
We emphasize including all. We encourage participation at any level for good. Inspired by reports on public deliberation, we'll organize alignment assemblies to discuss AI questions. After we compile our policy recommendations and findings from our initial Policython involving college students, we propose to expand our cohort to involve people from various backgrounds and age groups, especially those who will be impacted by AI the most—from screenwriters to politicians to legal aids to artists to automotive industry workers.
A key first step will be Citizen Future Telling events with 10-20 diverse participants discussing scenarios like how AI should handle children's questions on controversial topics. We'll ask for their viewpoints and underlying principles, have others challenge them, and repeat this process. We'll input responses into GPT-4, identify novel ideas, vote on favorites and share with Policython participants to guide their work. We want to empower regular people in policymaking.
Thus, the key elements of our Policython participant selection are: gathering diverse input through in-person and online discussions; using AI to analyze, supplement and match this input to create outcomes; building partnerships and toolkits enabling broader efforts; focusing on inclusion, minority and beginner participation; and recognizing the value of student work and public deliberation on these issues. The vision is a sustained, global and scalable process for shaping policy around advanced technologies in a collaborative, transparent and human-centered manner.
Limitations: What do you expect to be the biggest limitations of your approach? (e.g., potential for process gaming, types of questions your process would be unable to help answer)
The main limitation is time
We can anticipate several limitations associated with our approach:
For the tooling or infrastructure required for our experiment, we have identified several areas where tools can be useful:
Marketing: We have a distribution list that is 5000 members strong. To support our marketing efforts, we can leverage existing tools and platforms such as social media management tools (e.g., Hootsuite, Buffer) for scheduling and analyzing social media posts, email marketing platforms (e.g., Mailchimp, Sendinblue) for managing email campaigns, and analytics tools (e.g., Google Analytics, Facebook Insights) for tracking and analyzing marketing performance.
Submissions: We’ve previously used Google Suite for submissions. Everything can be done on a free plan, although we’ve stored over 100 TB of data across our initiatives. Then we share the results using Airtable, which allows for easy search and filtering based on different criteria.
Portal: We want to build a portal for storing, following up, and voting of submissions. We’ve previously shared submissions to the world for feedback. Given that the paper length median is 10 pages, we plan to use GPT-4 to summarize each submission to make it easier for people to judge other submissions in stages. Ideally, we hope to use Democratic Inputs to AI grantees in this process.
Policymaking Research/implementation pipeline: For streamlining our policymaking research and implementation pipeline, we plan to utilize project management tools such as Notion.
Quadratic voting: Implementing quadratic voting can be facilitated by specialized tools designed for collective decision-making. Platforms like Quadratic Voting by RadicalxChange Foundation or Qvorum can provide the necessary infrastructure to enable participants to allocate voting weights and express preferences using quadratic voting mechanisms.
Existing d-lists and partners: While managing existing d-lists and partnerships, we can employ customer relationship management (CRM) software such as Salesforce, HubSpot, or Zoho CRM.
In the case of existing tools, the compelling features for our project lie in their established functionalities, user-friendly interfaces, and compatibility with standard practices in marketing, data organization, project management, and decision-making mechanisms.
At past policythons, we had multiple judges rate each submission to determine the finalists, which were presented to a second set of judges, along with a public vote.
Selecting people from non-technical backgrounds and educating them on AI: Most of our initial cohort will be college students who are either well-versed in AI or technical matters or have a high degree of interest in them. However, since AI will likely affect those without much of a technical background the most, we need to ensure we include those people in the decision-making process. Since these participants will have limited familiarity with technical concepts, we’ll need to provide educational materials or conduct pre-workshops to enhance participants' understanding before engaging in discussions or decision-making processes.
Contextualization to global South countries: When contextualizing our findings and proposals to global South countries, we may encounter challenges due to differences in cultural, social, economic, and political contexts. The solutions or recommendations generated through our approach may need to be adapted or tailored to these regions' specific needs and realities. Engaging local experts and stakeholders during the process can help address this limitation and ensure a more accurate and relevant understanding of the context.
Turning the winning proposal into something robustly implementable: While our approach may identify a winning proposal or solution, transforming it into something robustly implementable may present challenges. Practical implementation often involves navigating regulatory frameworks, resource allocation, technical feasibility, and stakeholder buy-in, among other factors. It is essential to recognize that the implementation phase may require additional efforts beyond the scope of our approach, such as further research, collaboration with policymakers and industry experts, and iterative development to refine and adapt the proposal for real-world implementation.