TU Berlin

Cognitive Modeling in dynamic Human-Machine SystemsPredicting User Performance and Errors


Page Content

to Navigation

Predicting User Performance and Errors on the Basis of the Computational Introspection of Model-Based User Interfaces

Dissertation of Marc Halbrügge

Maintaining the usability of applications that target several device types at once (e.g., desktop PC, smart phone, smart TV) is a challenging problem. The current work proposes to combine cognitive modeling with model-based user interface development (Meixner, Paternò, & Vanderdonckt, 2011) to tackle this issue. Model-based applications provide interesting meta-information about the elements of the user interface (UI) that are accessible through computational introspection. Cognitive user models can capitalize on this meta-information to provide improved predictions of the interaction behavior of future human users of applications under development.

In order to achieve this, cognitive processes that link UI properties to usability aspects like effectiveness (user error) and efficiency (task completion time) are established empirically, are explained through cognitive modeling with ACT-R (Anderson, 2007), and are validated in the course of the work. In the case of user error, an extended model of sequential action control based on the Memory for Goals theory (Altmann & Trafton, 2002) is developed and confirmed in different behavioral domains and experimental paradigms.

The newly developed model of user cognition and behavior is implemented using the MeMo workbench (Engelbrecht, Kruppa, Möller, & Quade, 2008) and integrated with the model-based application framework MASP (Blumendorf, Lehmann, & Albayrak, 2010) in order to provide automated usability predictions from early software development stages on. Finally, the validity of the resulting integrated system is confirmed by empirical data from a new application, eliciting unexpected behavioral patterns.


Quick Access

Schnellnavigation zur Seite über Nummerneingabe