Peace Polls
AI - Artificial Intelligence
Using AI with the United Nations Innovation Cell in the Department of Political and Peacebuilding Affairs (UN DPPA)
Abstract
Inclusive, transparent engagement with the people caught up in a conflict greatly enhances the potential for a successful peace process, giving the people a ‘seat at the negotiating table’ in both the age of paper (1) and the age of digital media and AI (2). In collaboration with the UN DPPA Innovation Cell at the UN Mission to Libya five 1000 participant capacity AI dialogues were run through Remesh’s computerised system in New York. In addition to demonstrating the technical success of this approach to conflict analysis some practical advances were also achieved where AI was used as an outreach tool increasing the participation of the Libyan population in their peace process, culminating in the election of a new Government of National Unity (GNU) in Geneva in 2021 (3). This paper reviews:
- The AI methodologies of Large Scale Digital Dialogues as practiced on the Remesh AI platform and how it is used to track and analyse consensus.
- The top line results for the five AI Dialogues completed in Libya.
- The political context in which these AI Dialogues were undertaken and how they were used to help advance the peace process.
- Lessons drawn from this experience and how these new AI tools will be developed in the future.
(1) Irwin, C., (2020) The People’s Peace, Second Edition: Public Opinion, Public Diplomacy and World Peace, CreateSpace, Scotts Valley, CA. Available at: http://www.peacepolls.org/peacepolls/documents/008880.pdf
(2) Bilich, J., Konya, A., Masood, D., and Varga, M., (2019) Faster Peace via Inclusivity: An Effective Paradigm to Understand Populations in Conflict Zones, AI for Social Good workshop at NeurIPS, Vancouver, Canada.
(3) Williams, S., and Feltman, J., (2021) Can a political breakthrough mend a broken Libya? Brooking’s Initiative on Nonstate Armed Actors, 17 February. Available at: https://www.brookings.edu/blog/order- from- chaos/2021/02/17/can- a- political- breakthrough- mend- a- broken- libya/
Read the full paper Using Artificial Intelligence for Peacebuilding
In 2023 Open AI launched a program to award ten $100,000 grants to fund experiments in setting up a democratic process for deciding what rules AI systems should follow, within the bounds defined by law. The RemeshAI team (Andrew Konya founder and CEO of Remesh, Lisa Schirch at the University of Notre Dame, Colin Irwin at the University of Liverpool, Aviv Ovary at the AI and Democracy Foundation) were awarded one of these grants and that produced the following results.
Abstract
We design and test an efficient democratic process for developing policies that reflect informed public will. The process combines AI-enabled collective dialogues that make deliberation democratically viable at scale with bridging-based ranking for automated consensus discovery. A GPT4-powered pipeline translates points of consensus into representative policy clauses from which an initial policy is assembled. The initial policy is iteratively refined with the input of experts and the public before a final vote and evaluation. We test the process three times with the US public, developing policy guidelines for AI assistants related to medical advice, vaccine information, and wars & conflicts. We show the process can be run in two weeks with 1500+ participants for around $10,000, and that it generates policy guidelines with strong public support across demographic divides. We measure 75-81% support for the policy guidelines overall, and no less than 70-75% support across demographic splits spanning age, gender, religion, race, education, and political party. Overall, this work demonstrates an end-to-end proof of concept for a process we believe can help AI labs develop common-ground policies, governing bodies break political gridlock, and diplomats accelerate peace deals.
Read the full paper Democratic Policy Development using Collective Dialogues and AI
Abstract
We introduce a method to measure the alignment between public will and language model (LM) behavior that can be applied to fine-tuning, online oversight, and pre-release safety checks. Our “chain of alignment” (CoA) approach produces a rule based reward (RBR) by creating model behavior rules aligned to normative ob-jectives aligned to public will. This factoring enables a nonexpert public to directly specify their will through the normative objectives, while expert intelligence is used to figure out rules entailing model behavior that best achieves those objectives. We validate our approach by applying it across three different domains of LM prompts related to mental health. We demonstrate a public input process built on collective dialogues and bridging-based ranking that reliably produces normative objectives supported by at least 96% ± 2% of the US public. We then show that rules developed by mental health experts to achieve those objectives enable a RBR that evaluates an LM response’s alignment with the objectives similarly to human experts (Pearson’s r = 0.841, AU C = 0.964). By measuring alignment with objectives that have near unanimous public support, these CoA RBRs provide an approximate measure of alignment between LM behavior and public will.
Read the full paper Chain of Alignment: Integrating Public Will with Expert Intelligence for Language Model Alignment