Using cognitive modeling for adaptive automation triggering

Abstract

The Multi-Level Cognitive Cybernetics (MLCC) [1] approach provides a methodological approach to studying adaptive automation and advancing its development across multiple levels of analysis. We follow up on the previous paper [1] by focusing on the cognitive modeling level of MLCC. Adaptive aids must only be triggered when the inclusion of the aid will boost performance relative to what it would be without the aid. Computational cognitive modeling provides a means to represent the cognitive sequence that completes a task. Using cognitive modeling and MLCC, we discuss two ways to provide optimal triggering for adaptive automation. First, models will provide a mapping of which cognitive stages caused the most difficulty for individuals and therefore aids may be designed to support those cognitive functions. Second, models may provide information about optimal thresholds for determining when a user is having difficulty, allowing more timely aid interventions than without the model.