Comment from Adegbite, Tolu

RE: Comment on Proposed Policy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems Matter No. P264200 I am submitting this comment as a product designer and inclusive design researcher with experience in accessibility, inclusive design, and human-AI interaction. I am completing a Master's Research Project at OCAD University's Inclusive Design Research Centre titled Who Gets to Use AI? Inclusive Prompting from the Edges, under the supervision of Dr. Jutta Treviranus, Director of the Inclusive Design Research Centre. My research examines how neurodivergent users interact with AI prompting systems, with attention to barriers affecting users with ADHD, autism, depression, and anxiety. The proposed policy statement addresses a practice I observe directly in my research: AI systems that steer outputs away from accurate responses without telling users. I want to flag a population that is disproportionately harmed by this practice and that the statement does not currently address: users with disabilities, specifically neurodivergent users. Neurodivergent users like people with ADHD, autism, depression, and anxiety make up an estimated 15-20 percent of the population. They interact with AI prompting systems differently from neurotypical users. Their prompts are often structured differently, sometimes less optimized for how AI systems are designed to receive input, and more likely to generate outputs that are miscalibrated to the original query. When an AI system steers output away from an accurate response, neurotypical may have the cognitive resources to detect the problem: they can compare outputs, iterate on their prompts, cross-reference other sources, and escalate when something seems wrong. Many neurodivergent users have significantly reduced capacity for this kind of error-detection work. They are more likely to rely on a single output and less likely to detect when that output has been steered or is inaccurate. This has the potential to cause significant consumer harm. It falls squarely within the FTC's stated concern. And as AI becomes the interface through which people access healthcare information, employment guidance, public benefits instructions, and educational content, neurodivergent users who receive steered or inaccurate outputs that are presented as accurate are likely to face adverse real-world consequences. Based on my research with neurodivergent users, I would recommend that the policy statement: 1. Explicitly address cognitively diverse users as a disproportionately harmed population.
The harm of AI accuracy suppression is not evenly distributed. Users with reduced capacity for output verification, error detection, or iterative prompting are more vulnerable to consumer harm from steered/misrepresented AI outputs. 2. Require that accuracy disclosures be accessible.
If AI systems are required to disclose when outputs may have been steered or filtered, those disclosures must be accessible to users with cognitive, neurodevelopmental, psychiatric, learning, and communication-related disabilities. A disclosure a neurodivergent user cannot understand or act on does not meaningfully mitigate the harm. 3. Include neurodivergent users in AI accuracy testing and evaluation.
An AI system may produce outputs that appear accurate when tested with neurotypical users, while producing systematically different outputs for neurodivergent users. Testing frameworks should reflect this. 4. Address AI interface design as part of the accuracy framework.
Accuracy suppression is not only a model-level problem. It also occurs at the interface level; when systems are designed in ways that make it harder for some users to formulate queries accurately, interpret outputs correctly, or recognize errors. The FTC's consumer protection framework should account for interface-level accuracy failures, not only model-level steering. I would welcome the opportunity to contribute research findings to any follow-up process related to AI accuracy and cognitively diverse users. Thank you, Tolu Adegbite Product Designer and Inclusive Design Researcher