Abstract Human sensorimotor decision-making has a tendency to get ‘stuck in a rut’, being biased towards selecting a previously implemented action structure (‘hysteresis’). Existing explanations cannot provide a principled account of when hysteresis will occur. We propose that hysteresis is an emergent property of a dynamical system learning from the consequences of its actions. To examine this, 152 participants moved a cursor to a target on a tablet device whilst avoiding an obstacle. Hysteresis was observed when the obstacle moved sequentially across the screen between trials, but not with random obstacle placement. Two further experiments (n = 20) showed an attenuation when time and resource constraints were eased. We created a simple computational model capturing dynamic probabilistic estimate updating that showed the same patterns of results. This provides the first computational demonstration of how sensorimotor decision-making can get ‘stuck in a rut’ through the dynamic updating of its probability estimates. Significance Statement Humans show a bias to select the organisational structure of a recently carried out action, even when an alternative option is available with lower costs. This ‘hysteresis’ is said to be more efficient than creating a new plan and it has been interpreted as a ‘design feature’ within decision-making systems. We suggest such teleological arguments are redundant, with hysteresis being a naturally emergent property of a dynamic control system that evolved to operate effectively in an uncertain and partially observable world. Empirical experimentation and simulations from a ‘first principle’ computational model of decision-making were consistent with our hypothesis. The identification of such a mechanism can inform robotics research, suggesting how robotic agents can show human-like flexibility in complex dynamic environments.