No Plan, Yet Human: A Reactive Robotics Model Predicts Human Planning Failures on a Clinical Task

In three linesAICON, a reactive robotics model using gradient descent, better predicts human planning failures on the Tower of London cognitive test than planning baselines. Without lookahead, it reproduces the difficulty ordering of 24 problems and fails similarly to Parkinson's patients, suggesting reduced planning capacity shifts behavior toward reactive modes.

Summary generated by Claude — human-verified