"It just requires so much more creativity": Barriers and Workarounds to Gathering Information for AI Contestation

Paper Title

"It just requires so much more creativity": Barriers and Workarounds to Gathering Information for AI Contestation

Publication Info

• Topic area: Information gathering and advocacy in AI contestation
• Keywords: AI contestation, advocacy, transparency, information work, social justice, legal barriers, policy, political economy, collaboration, risk

Background and Problem

• Problem / challenge: Advocates contesting harmful AI systems face significant barriers in accessing and using information due to opacity, adversarial tactics, and resource constraints. Existing transparency policies are insufficient to address these challenges.
• Significance: Effective contestation of AI systems is critical to mitigating harms in high-stakes domains like healthcare, housing, and employment, where AI often exacerbates social injustices.
• Motivation and related work: Prior research has focused on professional auditors and technical solutions, neglecting the broader advocacy community. There is limited understanding of how diverse advocates—lawyers, journalists, and policy advocates—navigate the information work required for contestation, especially in resource-constrained environments.

Solution

• Proposed approach: Empirical study of the information work of advocates contesting AI systems in the US, highlighting their strategies, barriers, and creative workarounds.
• Novelty:
  1. Characterization of the iterative and non-linear life cycle of information work in AI contestation.
  2. Analysis of creative workarounds, including interdisciplinary collaboration, direct engagement with AI systems, and innovative use of open records laws.
  3. Examination of the political economy of AI contestation, emphasizing structural disadvantages and risks faced by advocates.
  4. Critique of transparency policies and recommendations for strengthening community and collaboration in advocacy.
• Procedure and key techniques:
  • Conducted interviews with 18 US-based advocates (lawyers, journalists, auditors, policy advocates, and researchers) across domains like healthcare, housing, and employment.
  • Used inductive interpretive analysis to identify themes related to information work, barriers, and strategies.
  • Explored the implications of findings for transparency policies and advocacy practices.

Results

• Concrete findings:
  • Advocates face systemic opacity throughout the life cycle of information work, including harm discovery, strategizing, and sensemaking.
  • Creative workarounds include leveraging relationships, experimenting with AI systems, and using public records laws, though these are resource-intensive and risky.
  • Structural barriers include adversarial tactics by AI deployers, resource constraints, and governance structures favoring developers.
  • Risks include legal consequences, political backlash, and personal harm to advocates and affected individuals.
• Advantage over baselines:
  • The study highlights the insufficiency of transparency policies and the need for relational and coalitional approaches, which are often overlooked in technical and policy-centric research.
• Experiments / evaluation:
  • Interviews with 18 participants across diverse roles and domains provided qualitative insights into the challenges and strategies of AI contestation.
  • Data analysis revealed patterns of information work and systemic barriers, informing critiques of existing policy approaches.
• Limitations and future work:
  • Focused on the US context, limiting generalizability to other regions.
  • Did not include perspectives of individuals directly impacted by AI harms.
  • Future research should explore cross-cultural contexts and broader participant populations.

Summary

This study examines the information work of advocates contesting harmful AI systems in the US, revealing systemic barriers like opacity, adversarial tactics, and resource constraints. Advocates employ creative workarounds, including interdisciplinary collaboration and innovative use of public records laws, but face significant risks and structural disadvantages. The findings critique the limitations of transparency policies and emphasize the importance of relational and coalitional approaches to support advocacy. The study calls for infrastructure and procedural reforms to reduce the burdens of contestation and amplify solidarity among diverse advocates.