Research map / Large language models / AI and society

Position: Generative AI Regulation Can Learn from Social Media Regulation

ICML 2025 · Large language models › AI and society · approach: Human-centered trust and agency

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Position paper arguing that generative AI regulation can learn from social media regulation, comparing affordances of both media and deriving four policy recommendations.

Problem

Debate over how to regulate generative AI: whether existing laws should be extended or entirely new laws are needed, with unclear lessons from prior technology regulation.

Approach

Adapts Clark's media affordance features (medium, control, immediacy) to compare generative AI and social media in a table, finds shared features that make social media a useful analogy, and derives policy recommendations (transparency and researcher access, democratic input, user choice, trust and safety, computational social science, global perspective) from social media regulatory history.

Key results

• Comparison table of 24 media affordance features finds generative AI and social media share several key features (e.g., personalization, recording, use of AI, black-box, content moderation), while differing on invisible and pre-generation moderation and simultaneity.
• Proposes four policy recommendations: counter bias and perceptions (transparency, researcher access, oversight boards, democratic input), address specific concerns and invest in trust and safety, promote computational social science, and take a global perspective.

Surprising findings

• Generative AI content moderation is less visible to users than social media moderation, since refusals usually occur before content is shown and often lack explanations.

Limitations

• Analogy between generative AI and social media is admittedly imperfect
• Recommendations are argued qualitatively and not empirically tested
• Affordance analysis is based on the author's informal analysis of commonly used applications

Experimental setup

Builds on: Clark's media affordance features (face-to-face communication) Digital Services Act

Closest papers by meaning

• Braess’s Paradox of Generative AI (AAAI 2025 · similarity 0.84)
• Position: Near to Mid-term Risks and Opportunities of Open-Source Generative AI (ICML 2024 · similarity 0.84)
• Position: A Safe Harbor for AI Evaluation and Red Teaming (ICML 2024 · similarity 0.83)
• AI-Mediated Communication Can Steer Collective Opinion (NeurIPS 2026 · similarity 0.83)
• Position: If Innovation in AI systematically Violates Fundamental Rights, Is It Innovation at All? (NeurIPS 2025 · similarity 0.83)
• Watermarking Without Standards Is Not AI Governance (NeurIPS 2026 · similarity 0.83)