Regulating AI: Where U.S. State Policy and HCI (Mis)align

Paper Title

Regulating AI: Where U.S. State Policy and HCI (Mis)align

Publication Info

• Topic area: Analysis of U.S. state-level AI policy development and its alignment with HCI research.
• Keywords: AI policy, state-level governance, HCI, socio-technical systems, AI risks, AI benefits, responsible AI, participatory design, AI literacy, AI regulation.

Background and Problem

• Problem / challenge: U.S. state-level AI committees emphasize AI benefits over risks and lack alignment with socio-technical concerns emphasized in HCI research. Discussions of risks are cursory and lack specificity, with limited inclusion of diverse perspectives.
• Significance: As AI adoption accelerates, state-level policies will shape societal impacts, making it critical to address gaps in risk assessment, inclusivity, and socio-technical framing.
• Motivation and related work: While HCI research highlights socio-technical risks and emphasizes participatory design, U.S. state AI committees focus narrowly on economic growth and operational efficiency. There is a lack of systematic studies examining U.S. state-level AI policy development.

Solution

• Proposed approach: Mixed-methods analysis of 18 U.S. state-level AI committee reports to compare policymakers’ priorities with HCI research on AI risks and benefits.
• Novelty:
  1. Development of a typology of motivations for forming AI committees.
  2. Comparison of AI-related risks and benefits across sectors in state reports.
  3. Alignment analysis between state reports and HCI literature using the AI Risk Repository.
  4. Recommendations for bridging gaps between policy and HCI through participatory design and socio-technical perspectives.
• Procedure and key techniques:
  • Quantitative coding of AI benefits and risks across nine sectors.
  • Comparison of committee-reported risks with the AI Risk Repository taxonomy.
  • Thematic analysis of mitigation strategies and values in committee reports.
  • Statistical analysis (Wilcoxon signed-rank test, mixed-effects logistic regression) to assess emphasis on benefits versus risks.

Results

• Concrete findings:
  • AI benefits were emphasized more than risks across all sectors (e.g., government services: 83% benefits vs. 58% risks).
  • Key risks discussed in HCI literature, such as environmental harm and labor devaluation, were underrepresented in committee reports.
  • Committees frequently lacked definitional clarity for AI and proposed ambiguous mitigation strategies.
• Advantage over baselines:
  • Identified significant misalignment between state-level AI policy priorities and socio-technical concerns in HCI research.
  • Highlighted gaps in inclusive governance and operational definitions of AI risks.
• Experiments / evaluation:
  • Analysis of 18 state-level AI committee reports using a mixed-methods approach.
  • Comparison with the AI Risk Repository, which synthesizes over 700 risks from HCI and related fields.
  • Statistical tests confirmed systematic prioritization of benefits over risks.
• Limitations and future work:
  • Focused only on U.S. state-level committees, excluding federal and international contexts.
  • Relied on committee reports rather than direct interviews with policymakers.
  • Future work should explore broader stakeholder inclusion and global AI policy initiatives.

Summary

This study analyzed 18 U.S. state-level AI committee reports to understand how policymakers prioritize AI-related benefits and risks and how these align with HCI research. Findings reveal that state committees emphasize benefits over risks, lack socio-technical framing, and propose ambiguous mitigation strategies. Key risks highlighted in HCI literature, such as environmental harm and labor devaluation, are underrepresented. The paper recommends leveraging HCI methods, such as participatory design and AI literacy initiatives, to bridge gaps between policy and socio-technical research, ensuring more inclusive and effective AI governance.