Phrenology — the nineteenth-century pseudoscience that claimed to read character and intelligence from the bumps and contours of a human skull — was not fringe lunacy. It was respectable science, backed by leading researchers, endorsed by institutions, and used to make consequential decisions about individuals' lives. It was also completely wrong, and its wrongness caused real harm before it was finally discredited.
Stuart McClure asks whether AI-based assessments of human personality, cognitive ability, and workplace fit are at risk of repeating the same pattern. The concern is not that AI itself is pseudoscientific — the computational techniques are real and powerful. The concern is about what those techniques are actually measuring and whether the claims made about them are warranted by the underlying science.
Stuart holds a degree in Psychology and Philosophy, which gives him an unusual foundation for this inquiry. He knows what behavioral science actually says about the stability of personality traits, the predictive validity of various assessment instruments, and the gap between what a model can learn from behavioral data and what that data actually tells you about a person. That gap, he argues, is often wider than AI vendors acknowledge.
This essay is not an argument against using AI in talent management — Stuart's own company Wethos AI does exactly that. It is an argument for doing it right: grounding AI assessments in validated behavioral science, being honest about what the models can and cannot reliably infer, and designing systems that augment human judgment rather than substitute for it. The alternative is a new phrenology — impressive-sounding, authoritatively presented, and fundamentally misleading.