A proposed conceptual model of human behavior and its implications for design of warnings.

A model of human behavior is proposed that hierarchically describes levels of operator performance. Judgement-based performance occurs at the top level, and the lower levels correspond to knowledge-, rule-, and skill-based performance. Different forms of information denoted as values, symbols, signs, and signals are used at each of these levels of performance, allowing the effectiveness of different warning applications to be inferred. To be effective, warning information must be presented in the form appropriate for the operator's level of performance. Values therefore are appropriate when performance is at the judgement-based level. Explicit verbal information (symbols) is most likely to be effective when directed toward changing behavior from a knowledge- to a rule-based level, as when recommending actions in novel situations or to a judgement-based level, when goal priorities need to be changed. Signs are likely to be effective when performance is at a rule-based level, while signals are best for guiding needed transitions from a skill- to a rule-based level. Warning information should be carefully matched to the level of performance at which errors are taking place to be most effective and avoid information overload. To attain this goal, task analysis that focuses on cognitive activity is of essence. This includes measuring users' knowledge and documenting the flow of information during task performance.