AI Harms

Two tools for visualizing and conceptualizing the flow of causality, information, incentive, and harm related to artificial intelligence products.

I made them as companions to this essay: The Hazards of AI: a 3d model (work in progress),

• Taxonomy of user harms: the user column of the model opened up. Types of harm to the user, their causes, and the MIT Domain Taxonomy drawn over the same concerns, with the explorer’s case harms placed inside the same pictures.

• Harm explorer: the 3d model as an interactive solid. Each harm from a source study is a sphere placed by the party it lands on, the party responsible for its cause, and its timescale; every node carries a verbatim quote and page from the source. Load a bundled study or upload your own.

Source studies in the explorer

• Wells and Ustek Spilda, “Uber for Nursing”, Roosevelt Institute, December 2024
• Garcia v. Character Technologies, Inc., complaint (October 2024) and order on the motions to dismiss (May 2025)
• National Nurses United, “Nurses and Patients’ Bill of Rights”, 2024
• Fang et al., chatbot use and psychosocial outcomes, randomized trial, 2025

The datasets, their schema, and the bridge files that place each harm in the taxonomy live in the site repository under data/harms/cases/.