Theo Todman's Web Page - Paper Abstracts

Author’s Introduction

1. The Frame Problem is a well-known but often misunderstood problem in the foundations of artificial intelligence¹. In the real world, things change. A rational agent residing in a changing world must be able to reason about change and persistence. This requires four kinds of reasoning.
   • First, it must be able to dip into the world perceptually, acquiring information about the current state of the world.
   • Second, it must be able to combine information obtained from different perceptual excursions and, through inference, construct a coherent picture of the world.
   • Third, it must be able to detect changes perceptually and update its picture of the world accordingly.
   • Fourth, it must be able to acquire general causal information about "how the world works" and use that to predict the results of changes either observed by the agent or wrought by the agent's own actions.
2. Officially, the Frame Problem arises only in connection with the fourth constituent of reasoning about persistence and change, but the solution to the Frame Problem requires principles that bear upon the other constituents as well.