The math-solving horse who fooled scientists still haunts AI research

In early twentieth-century Germany, Wilhelm von Osten exhibited a horse named Clever Hans who appeared to solve math problems, tell time, and answer complex questions by tapping his hoof. Scientists were impressed by the horse's apparent abilities — until testing revealed Hans wasn't actually doing arithmetic. The horse had learned to read tiny behavioral cues from observers themselves: when to stop tapping based on shifts in their stance, their breathing, or their posture. Boing Boing reports that the Clever Hans case has become a cautionary tale for modern AI researchers, who face a similar risk: sophisticated-seeming capabilities that turn out to be pattern-matching exploiting blind spots in how we design and test systems. As AI grows more complex, the century-old story of a horse who fooled scientists offers a sharp reminder to audit carefully.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton