Envisioning Stakeholder-Action Pairs to Mitigate Negative Impacts of AI: A Participatory Approach to Inform Policy Making

TL;DR

This work tackles the lack of lay stakeholder input in AI governance by introducing a participatory foresight method to map stakeholder-action pairs (SAPs) for mitigating negative AI impacts in the news/media domain. It uses narrative scenarios to elicit SAPs from lay participants, a second survey to rank them by agreement and priority, and a large language model to convert results into concise policy fact sheets for policymakers. The study reveals a rich set of SAPs across 10 impact types, with fact-checking, transparency, and digital literacy frequently prioritized, and actors such as news publishers, technology companies, and governments identified as responsible for action. The approach offers a translational tool to enrich policy deliberation with diverse public input, while acknowledging limitations around representativeness and the need for broader validation.

Abstract

The potential for negative impacts of AI has rapidly become more pervasive around the world, and this has intensified a need for responsible AI governance. While many regulatory bodies endorse risk-based approaches and a multitude of risk mitigation practices are proposed by companies and academic scholars, these approaches are commonly expert-centered and thus lack the inclusion of a significant group of stakeholders. Ensuring that AI policies align with democratic expectations requires methods that prioritize the voices and needs of those impacted. In this work we develop a participative and forward-looking approach to inform policy-makers and academics that grounds the needs of lay stakeholders at the forefront and enriches the development of risk mitigation strategies. Our approach (1) maps potential mitigation and prevention strategies of negative AI impacts that assign responsibility to various stakeholders, (2) explores the importance and prioritization thereof in the eyes of laypeople, and (3) presents these insights in policy fact sheets, i.e., a digestible format for informing policy processes. We emphasize that this approach is not targeted towards replacing policy-makers; rather our aim is to present an informative method that enriches mitigation strategies and enables a more participatory approach to policy development.

Figures (5)

• Figure 1: Flow diagram detailing the approach proposed in this work. We first ask lay stakeholders to brainstorm possible stakeholder-action pairs to mitigate negative impacts from AI (Survey 1). We then ask lay stakeholders to rank these in terms of agreement and priority (Survey 2). Finally, we utilize LLMs to synthesize this information into a one-page fact sheet to inform policy makers, and have an expert on our team validate them for usefulness.
• Figure 2: Stacked bar chart displaying how often lay stakeholders allocate responsibility to various stakeholders to take an action, sorted by frequency, split by impact types.
• Figure 3: Heatmap displaying the frequency of stakeholder-action pairs across the various impact types brainstormed by lay stakeholders, sorted by vertically by action frequency and horizontally by stakeholder frequency.
• Figure 4: Stacked bar chart displaying how often lay stakeholders brainstormed various actions they wanted stakeholders to take, sorted by frequency, split by impact type.
• Figure 5: Example "Policy Fact Sheet" for unemployment impacts generated using GPT-4o, prompt engineering, and the data from this study.