Closing the Loop: An Agenda- and Justification-Based Framework for Selecting the Next Discovery Task to Perform

We propose and evaluate an agendaand justificationbased architecture for discovery systems that contains a mechanism for selecting the next task to perform. This framework has many desirable properties: (1) its use of heuristics to perform and propose tasks facilitates the use of general discovery strategies that are able to use a variety of background knowledge, (2) through the use of justifications its mechanism for selecting the next task to perform is able to reason about the appropriateness of the tasks being considered, and (3) its mechanism for selecting the next task to perform also considers the users interests, allowing a discovery program to tailor its behavior toward them. We evaluate the extent to which both reasons and estimates of interestingness contribute to performance in the domain of protein crystallization. With both aspects contributing to task selection, a high fraction of discoveries by the HAMB prototype were judged interesting by an expert (21% interesting and novel; 45% interesting but rediscoveries).