Simulating problem-solving

UNIVERSITY PARK, Pa. — Christopher McComb, assistant professor of engineering design and mechanical engineering, and researchers at Carnegie Mellon University have received funding from the Defense Advanced Research Projects Agency (DARPA) to design a computer algorithm that is able to simulate and enhance human problem-solving capabilities.

McComb and Jonathan Cagan, the George Tallman and Florence Barrett Ladd Professor of Mechanical Engineering at Carnegie Mellon University, will create a computer platform that uses a two-tiered problem-solving language. Computational agents will primarily operate in the higher tier of the language, which is composed of general problem-solving operations that are agnostic to problem type. Higher-tier operations will be translated into the lower-tier of the language, enabling direct modification of problem-specific solution concepts.

These two language tiers will allow computers to understand and solve problems and interact with human team members. The platform will also provide dynamic data mining opportunities for collecting human behavior data.

“Traditionally, humans have had a master-tool relationship with computers. What this means is that our biases and tendencies were projected onto the way that we used these tools,” McComb said. “Designing computational algorithms that function as equal partners is a natural progression and it will help us be more balanced and less biased as we search for innovative solutions to difficult problems.”

Computers currently act as a problem-solving tool and not as an active problem-solving participant. Traditional research on individual and team problem-solving can take a significant amount of time to collect data. In contrast, the computational model the research team is designing will allow for quick changes to be made to problem-solving strategies in real time. This will help determine best practices for varying types of problems. The model will allow for various simulations, including changes in goals, environment, team structure, and team member interaction and expertise.

“Problems in which everything stays constant are the exception, not the rule. Real life is dynamic. In engineering design your competitor may release a new product or your supplier may change their pricing,” McComb said. “To keep up with these changes you usually need to change something about your approach.”

The team’s research and modeling will provide human teams with the ability to interact smoothly with a computer equipped with an intelligent problem-solving framework. The computer will be able to reason, decipher concepts and work with human team members to manage the problem-solving process. 

This research will help to optimize problem-solving processes for teams, especially in the case of problems with dynamic and ever-changing situations. The results will be applicable in creative design and also in military problem-solving situations that include strategic mission planning and deployment logistics.

“Problem-solving is a canonical activity that is at the center of many different domains. If we can create computer agents that partner with humans in general problem-solving, our approach can be used in fields as diverse as business strategy, pharmaceutical design, and military operation planning,” McComb said.

Cagan serves as the principal investor (PI) on the project, with McComb serving as a co-PI.

Media Contacts: 
Last Updated October 11, 2017