Soar Technology

Soar Technology
Name	Soar Technology
Type	Private
Industry	Artificial intelligence
Founded	1983
Founders	John Laird; Paul Rosenbloom; Allen Newell
Headquarters	Ann Arbor, Michigan
Products	Cognitive architecture implementations; simulation software

Soar Technology is a company and research lab that developed and commercialized a cognitive architecture originally created in the 1980s. The organization has linked academic research in cognitive modeling with applied systems for simulation, human–machine interaction, and decision support. Its work connects to a lineage of cognitive science, artificial intelligence, and human factors research derived from prominent laboratories and projects.

History

Soar Technology traces intellectual roots to research groups at Carnegie Mellon University, University of Michigan, and Rensselaer Polytechnic Institute where researchers such as Allen Newell, John Laird, Paul Rosenbloom, and collaborators advanced production-system models and cognitive modeling. Early milestones include the formulation of a unified theory of cognition in the 1980s and the creation of an architecture intended to model general intelligent behavior, intersecting with work at RAND Corporation, Stanford Research Institute, and laboratories influenced by Herbert A. Simon. Institutional activities spurred spin-offs and collaborations with centers like the Defense Advanced Research Projects Agency and the Naval Research Laboratory. Over time the organization engaged with government programs such as DARPA challenge initiatives and partnered with contractors including Lockheed Martin and SAIC on simulation projects and training systems. Leadership changes and funding cycles mirrored broader shifts in Artificial Intelligence research funding in the 1990s and 2000s, aligning the company with programs at National Science Foundation and Office of Naval Research.

Soar Architecture

The core product is an implementation of the Soar cognitive architecture, a symbolic, rule-based system designed to support problem solving, learning, and decision making. The architecture integrates concepts from production systems developed at institutions like Carnegie Mellon University and theoretical frameworks promoted by Allen Newell and Herbert A. Simon. Key components include a working memory, a long-term procedural memory of productions, and mechanisms for chunking and reinforcement-style learning; these mechanisms relate to models used at MIT, Yale University, and University of California, Berkeley in cognitive modeling. Soar implementations interface with external perceptual and motor subsystems in simulation environments such as those from MAK Technologies and Bohemia Interactive Simulations, and with middleware like HLA and DIS used by defense and aerospace institutions including Northrop Grumman and Boeing.

Applications and Implementations

Soar-based systems have been applied in simulated training, command-and-control prototyping, human-systems integration, and autonomous agents. Military simulation applications connected the architecture to platforms from U.S. Army training centers and contractors like Raytheon for scenario generation, after-action review, and virtual role players. Commercial applications included integration into serious games and synthetic teammates alongside engines by Epic Games and Unity Technologies. Research implementations have interfaced with robotics efforts at NASA facilities and with unmanned systems developed by General Dynamics and BAE Systems. In academic settings, Soar models have been used in cognitive tutors and intelligent tutoring system research at institutions such as Carnegie Mellon University and University of Pittsburgh.

Research Contributions

Soar Technology contributed to formalizing mechanisms for incremental learning and problem decomposition in cognitive architectures, building on work from MIT and Brown University. The architecture’s chunking mechanism links to psychological theories tested in labs like Harvard University and Princeton University. Publications and demonstrations with partners such as DARPA and National Institutes of Health explored decision support under uncertainty, human–agent teaming, and explainable behavior, resonating with efforts at Columbia University and University of California, San Diego. Comparative studies positioned Soar alongside other architectures developed at Oregon State University and University of Edinburgh, influencing benchmarks for cognitive fidelity used by researchers at Stanford University and University of Texas at Austin.

Commercial Products and Partnerships

The company commercialized tools and toolkits enabling integration of the architecture into simulation and training products. Partnerships involved defense contractors and systems integrators including Lockheed Martin, Northrop Grumman, BAE Systems, and General Dynamics, and collaborations with simulation vendors such as MAK Technologies and Bohemia Interactive Simulations. Projects for governmental customers interacted with procurement programs at U.S. Department of Defense agencies and research solicitations from DARPA and NSF. Commercial offerings supported interfaces to standards embraced by organizations like IEEE and enabled deployment in programs run by U.S. Army Training and Doctrine Command and other service branches.

Community and Education

An active research community around the architecture includes researchers and students at universities such as Carnegie Mellon University, University of Michigan, University of Southern California, and University of California, Berkeley. Workshops and conferences where work has been presented include venues like International Joint Conference on Artificial Intelligence, Conference on Cognitive Modeling, and meetings sponsored by Association for the Advancement of Artificial Intelligence. Educational uses include graduate- and undergraduate-level courses at institutions such as University of Michigan and Rensselaer Polytechnic Institute where models based on the architecture are used to teach cognitive modeling, human–computer interaction, and agent-oriented design.

Category:Artificial intelligence companies Category:Cognitive architectures