Why is it so hard for a robot to understand what changes when it takes an action - and what stays the same?

The answer

The frame problem, first identified by John McCarthy and Patrick Hayes in their 1969 paper 'Some Philosophical Problems from the Standpoint of Artificial Intelligence,' addresses the challenge of representing what does NOT change when an action is performed. If a robot moves a block from the table, it must infer that the table's color, the room's temperature, and countless other facts remain unchanged - without explicitly stating each one. In logical AI systems, this creates a combinatorial explosion: specifying all 'frame axioms' (non-effects) for every action becomes computationally intractable. The problem extends beyond logic to a deeper philosophical question about how minds efficiently update their model of the world, and it has been called by philosopher Daniel Dennett 'one of the most profound challenges in AI and cognitive science.'