Why Aren’t We Measuring How AI Affects Humans?

Quick take

AI systems get an intense amount of technical scrutiny. Researchers measure how fast models process data, how well they answer reasoning tests, and various performance metrics. However, what rarely gets measured is the human side of AI’s impact: how these systems affect people’s psychology, behavior, and social well-being.

Imran Khan, who leads psychosocial evaluation of AI at the Center for Humane Technology, spotlights this blind spot. The essay argues that as AI gets smarter and more embedded into daily life, it is critical to track what it does to humans. Ignoring this metric leaves a huge piece of AI’s impact unmeasured, and unregulated.

Why it matters

Businesses building or deploying AI need to know that the effects on users extend beyond typical performance benchmarks. Products that improve efficiency but degrade trust or mental health will create long-term risks for brand, engagement, and regulatory blowback.

Investors and operators should pressure teams to develop metrics capturing human outcomes. That includes emotional well-being, cognitive load, and social dynamics influenced by AI. Without these data points, risk assessments remain incomplete and biased toward technical performance.

For regulators, the lack of human impact evaluation weakens oversight and encourages narrow compliance. Rigorous psychosocial monitoring could shift incentives toward safer, more humane AI design. This in turn influences UX decisions, liability, and competitive differentiation.

AI operators need to think beyond throughput and accuracy. Human-centered metrics force reconsideration of what success means and where trade-offs lie. This expands accountability beyond models into the experience of the end user, the workforce, and society at large.

AI Quick Briefs Editorial Desk