Identity Signals for Enabling Participatory Governance
Abstract
Michael Bernstein, Stanford UniversityAmy X. Zhang, University of WashingtonApproved for Public Release. Research problem: Online co mmunities are digital fiefdoms, each run by an administrator withtotal control. The moral and practical limitations of this model ha ve become clear, with users engagingin collective action against administrators and platform owners, and platform owners strugglingt o develop policy that represents the population. Underlying these challenges are issues withtrust and reputation -- users are often not given voice because the ease of creating new accountsenables Sybil attacks. Enabling participatory policymaking online requires trustable reputationsystems as well as mechanisms for combining those voices.Technical approach: Drawing on the licensing and identi ty approaches of offline governmentorganizations and civil society groups, we will develop techniques that allow communities tocreat e and publish reputational signals, algorithms to help communities identifytrustable signals,and democratic policymaking approaches that utilize these signals. Our first objective is infrastructure for communities to author and publish verifiable reputationsig nals. For example, Stanford University might publish a signal that verifies a particular personis an alumni of the university; a civ il society organization might publish a signal that verifies thebearer has volunteered 100 hours with the community; the U.S. govern ment might publish signalsto verify that the person has voting rights, or holds a particular role or office. Underlying thiscontribu tion will be techniques for ensuring verifiability of these signals, and for portability ofthese signals across domains and platform s. Our second objective will be to develop algorithms that empower communities and organizationsto better calibrate levels of tr ust in signals that they are unfamiliar with. For example, acommunity might see a sudden influx of new participants, and be unfamili ar with the reputationalsignals that they carry, wondering, "Should I trust these signals?" Our goal is to enable visualizationof tr ust levels for unfamiliar reputation signals -- are they trusted by people and communitiesI trust? -- as well as automatic filtering and moderation of untrusted content. So, we will adaptnetwork centrality algorithms such as personalized PageRank to provide visual izations of localtrust -- trust as calibrated to a given community or, to those closest to you in the network. We will then brin g these signals together to enable participatory policymaking within andjointly across online communities. This final stage will con tribute designs and systems that enablereputation-powered policy nomination and voting. For example, several communities might wantt o jointly develop and vote on policies against misinformation or harassment, and need to ensurethat the voting body are trusted, act ive members of each community. We will enable communitiesto author criteria for suggesting policy, voting on policy, and for tabulat ing results.Anticipated outcome: This research will create infrastructure, algorithms, and systems to enablepolicymaking online. The se systems will be deployed and evaluated with online communities.Online governance complicates the "one person, one vote" criteria, and establishing trustedreputation signals will reduce raiding and increase trust in participatory decision-making online.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 20, 2021
- Source ID
- N000142112839
Entities
People
- Michael Bernstein
Organizations
- Office of Naval Research
- Stanford University
- United States Navy