Causality as a key to the frame problem

Abstract

In a formal description of actions and changes, there is a famous problem known as the “frame problem”. In its generalized form, the frame problem is that it is practically impossible to describe, or infer, all the necessary preconditions (qualification problem) and all the possible consequences of a given action (ramification problem). It is not only difficult to determine what changes and what does not change. It is also difficult to determine what is relevant. Our observations show that even humans cannot solve the frame problem. Humans simply behave as if there were no such problem. The solution is to use some kind of heuristics. We claim that the notion of causality is the heuristics necessary to solve the frame problem. The most important working hypothesis is that causality is not a logical relation. As one of the heuristics it is subject to change as more information becomes available. There are two kinds of nonmonotonicity in the reasoning of changes: one due to a lack of information, the other due to change itself. We need separate mechanisms for them because they are used in combination to solve the frame problem.