A recent MIT study investigated how humans solve complex problems by examining how people predict a ball’s path through a maze when the ball is hidden from view. Researchers found that instead of trying to track all possible trajectories simultaneously, which is impossible for humans, people break down the problem using two key strategies. The first is hierarchical reasoning: breaking the problem into manageable steps. The second is counterfactual reasoning: imagining alternative scenarios when their initial prediction doesn’t match the evidence.
The study revealed that people flexibly switch between these strategies based on their confidence in their memory. Notably, those with better memory recall are more likely to use counterfactual reasoning to revise their predictions. Interestingly, when researchers programmed similar cognitive limitations into machine learning models, the artificial systems adopted the same human-like problem-solving strategies. Overall, these findings demonstrate that humans act rationally within their computational constraints, using clever shortcuts and heuristics to solve problems that would otherwise be too complex to handle optimally.
Image credit: Photo_Concepts / iStock