Surprisingness and Expectation Failure: What's the Difference?

Intelligent systems operate in the midst of a superabundance of information lacking the tags that indicate which few aspects are significant to the particular problems at hand at any given time and place. Given this wealth of information coupled with real-time processing constraints, selective attention is fundamental to any chance of success. In much of cognitive science, attentional focus is linked with the concept of expectation generation and failure. Unfortunately, research in AI has to finesse the problem of selective attention, although sometimes recognizing it as an important component of the frame problem (e.g., Schank, 1979; 1982).