Knowledge acquisition and fault diagnosis: experiments with PLAULT

A series of experiments using a paradigm called PLAULT, where subjects have to discover the structure of a logical network, is reported. The links in the network are invisible, and subjects have to infer where they are by examining the status of the logic gates under a variety of conditions. Three experiments explore the interaction of features of the task, such as its spatial and graphical complexity, with the cognitive limitations of the learner. A combination of these features predicts which parts of a network subjects can and cannot learn, when knowledge is tested by requiring subjects to draw diagrams of the supposed network structure. These task features also predict efficiency of transfer to a fault diagnosis task. The experiments demonstrate the predictive power that can be gained by identifying task features that are likely to prevent thorough learning of a system. These results are also important for understanding the dissociation of the ability to perform routine tasks and the ability to respond to abnormal events