Speakers: Daniel Shank, PhD
Daniel Shank, PhD
Associate Professor of Psychological Science
Missouri University of Science & Technology
Dr. Daniel B. Shank is an associate professor of psychological science at Missouri University of Science & Technology specializing in the areas of social psychology and technology. His research primarily focuses on social psychological interactions with and perceptions of artificial intelligence, including morality, emotions, relationships, impressions, and behavior toward AIs. He has received over 3 million dollars in research funding and published over 50 peer reviewed articles. In 2025, he published a book The Machine Penalty: The Consequences of Seeing Artificial Intelligence as Less Than Human and presented a TEDx talk on The Harm of AI Romance.
AI Warnings: How People Understand and Misunderstand Simple AI Model Information
AI systems in critical domains like housing and health may provide simple AI model information as a warning for users to evaluate the AI’s capabilities. Ideally, users would use this information to determine the capability, and thus appropriateness, of the AI in that domain. In a series of 9 experiments on matching home buyers to potential homes, we provided a simple AI training limitation (e.g., “The AI is not trained on buyers with ages 45 or below.”) and asked participants to rate the AI’s capacity to make an accurate recommendation. In general, people understand whether cases are in or out of scope of the AI’s training and rate the AI as more capable when it is in scope. Yet, they also often believe the AI is somewhat capable on cases it’s not trained on, especially people who understand and trust AI more. Additionally, people ignore whether the limitation actually applies to the housing domain: they make similar distinctions for an AI training limitation of income (highly related to house buying) as they do for a limitation of handedness (not related to house buying). I conclude by drawing parallels from house buying to kidney transplantation.