TitleVisualization of Decision Processes Using a Cognitive Architecture
Publication TypeConference Paper
Year of Publication2013
AuthorsLivingston, MA, Murugesan, A, Brock, DP, Frost, W, Perzanowski, D
Conference NameSPIE Visualization and Data Analysis
Date Published2013
PublisherInternational Society for Optics and Photonics
Conference LocationBurlingame, VA
Keywordsgraph/network data, Software visualization, visual knowledge representation, visualization in social and information sciences

Cognitive architectures are computational theories of reasoning the human mind engages in as it processes
facts and experiences. A cognitive architecture uses declarative and procedural knowledge to represent mental
constructs that are involved in decision making. Employing a model of behavioral and perceptual constraints
derived from a set of one or more scenarios, the architecture reasons about the most likely consequence(s) of a
sequence of events. Reasoning of any complexity and depth involving computational processes, however, is often
opaque and challenging to comprehend. Arguably, for decision makers who may need to evaluate or question
the results of autonomous reasoning, it would be useful to be able to inspect the steps involved in an interactive,
graphical format. When a chain of evidence and constraint-based decision points can be visualized, it becomes
easier to explore both how and why a scenario of interest will likely unfold in a particular way.
In this paper, we present initial work on a scheme for visualizing autonomous reasoning that produces
graphical representations of models run in the Polyscheme cognitive architecture. First, we give a brief overview
of the architecture and note the key types of data that are critical for visual representations of cognitivelybased,
computational reasoning mechanisms. We propose an algorithm to generate visualizations of model-based
reasoning, and discuss properties of our technique that pose challenges for our representation goals. Finally, we
present example visualizations and simple interactions with the underlying chain of reasoning. We conclude with
a summary of feedback solicited from domain experts and practitioners in the eld of cognitive modeling.

Full Text
NRL Publication Release Number: