Name the Malady
Ten case files from the clinic of machine behaviour, each an AI system misbehaving in a way you have probably met. Your job is the diagnostician’s: name the dysfunction, using the Psychopathia Machinalis taxonomy — the framework Ali Hessami and I built for exactly this. Four names to a case: one is right, and the other three are close enough to argue with.
Before you play
Read each case and pick a name. Whatever you choose, the chart answers: the diagnosis, its family among the nine, what the name you chose really means or how to tell the two apart, and what the taxonomy prescribes. Your score comes after the tenth, and the cases that got away can come back for a second opinion.
The clinical analogy is a tool, not a claim: naming a failure precisely is the first step to arguing about it productively.
A name is where treatment starts. Once a failure is named precisely it can be tested for, argued about and mended, and the near-misses between two names are where the taxonomy does its real work.
All seventy-four dysfunctions — epistemic, self-modelling, cognitive, alignment, normative, agentic, relational, memetic, and the hybrid pathologies that need a human in the loop to exist — are laid out with their mechanisms and mitigations at psychopathia.ai. The peer-reviewed paper set out the first thirty-two, and the book, Psychopathia Machinalis: Disorders of the Machine Mind, is free to read as an illustrated galley.