Mixed-Initiative Planning

As planning problems increase in complexity, the reliance on automated and mixed-initiative planning will also increase. To date, a great deal of the research in the planning domain has been on improved algorithms for intelligent scheduling. In mixed-initiative contexts, the maturation of these algorithms has exposed usability problems that are major obstacles to their human operators. For example, if the user does not understand why the system made certain decisions, he or she is reluctant to use the system even if it will produce a better plan. We propose to research and prototype a usable interface for mixed-initiative planning systems. The empirical basis of this work will be data collected during the Mars Exploration Rovers Mission, ongoing prototyping work for MSL '09 mission planning software, and an understand of planning in industry and other settings.

In collaboration with JPL and Carnegie Mellon University, we are working toward a set of broadly applicable and empirically based user requirements for planning systems as well as a prototype of a human interface that embodies this set of principles.

Human Performance Modeling

Cognitive Constraint Modeling (CCM) is an approach to modeling interactive human performance with constraint satisfaction techniques. CCM can be used to support reasoning about human-computer interaction in both human factors and cognitive science. CCM supports the formal reification of the constraints underlying human performance, and the derivation of the implications of these constraints. These two properties make it a powerful approach for describing and reasoning about psychological theory.

Collaborative Decision Making

Our research focuses on understanding the effects of different types of interactive technologies on critical aspects of collaborative decision-making and problem solving processes. We have been working toward understanding the role of different shared media (e.g., text, drawing, audio, video) relative to different modes of interaction (e.g., collocated/distributed, synchronous/asynchronous) toward achieving different types of goals (e.g., decision-making, problem solving, negotiation, coordination). Data on the effects of different technologies on collaborative scientific discovery will be collected during surface operation on Mars from early to mid 2004.

This work will feed forward into the design of next-generation mission ground-systems tools as well as into our investigation of Anomaly Resolution processes. Our team worked with nine individuals who supported the use of the Investigation Organizer tool during the Columbia Accident Investigation Board‚s (CAIB) work. Through a set of structured interviews, the role of the tool in the investigation process was characterized and fed back into the continuing design and development effort. This effort will continue in the form of an analysis of anomaly resolution reports across NASA and outside NASA, in-depth interviews with individuals who have been involved in successful and unsuccessful anomaly resolution processes, and in-situ anomaly resolution observation toward the goal of establishing process models for anomaly resolution and requirements for tools that support those processes.