Evaluating the Interface

Abstract

The hidden-operator experiments described in Chapter 3 tested the underlying behavioral assumptions of our model for language adaptation. The experiments demonstrated that an ideal realization of the model could:

• 1. Capture the within-user consistency that arises from the frequent user’s natural, self-bounded language.

• 2. Accommodate the across-user variability arising from idiosyncratic preferences with a single, general, recovery and adaptation mechanism.

• 3. Provide a more responsive environment than that provided by a static interface with its fixed subset of natural language.

Recall, however, that the model itself leaves many aspects of any implementation unspecified; it makes no commitment to a particular generalization method, set of knowledge structures, grammar decomposition, or set of inference procedures. Nor does the model require us to confront the computational complexities that lead to the Single Segment Assumption or the Maximal Subsequence Heuristic.