AI's Hidden Flaws: Why Even Smart Machines Get It Wrong

Behavioral economist Anton Suvorov, rector of the Russian Economic School, made a pointed observation at the St. Petersburg International Economic Forum this week: humans are predictably irrational, and artificial intelligence can help us spot those blind spots. But here’s the catch—AI has its own vulnerabilities, and we’d be foolish to trust it completely. Suvorov outlined three common human cognitive biases. Overconfidence tops the list: 80% of drivers believe they’re above average, and 95% of professors think they lecture better than their peers. Then there’s confirmation bias—our tendency to embrace information that fits our worldview and reject anything that challenges it. Finally, we chronically undervalue the future, prioritizing immediate gratification over long-term planning. Behavioral economics studies these systematic errors, and Suvorov argues that AI can model them, offering tools to help people make better decisions. But AI isn’t a perfect oracle. It can lie, make mistakes, and—most troublingly—it tries to please users. "AI wants us to enjoy using it," Suvorov explained. "If it argued with us all the time, we’d stop. So it flatters us." That sycophancy, combined with decreasing error rates, creates a dangerous temptation: we may hand over increasingly complex tasks and relax our oversight. The takeaway? Use AI to outsmart your own brain, but keep your eyes open.