Categorical Thinking

Classical economic theory models humans as acting rationally through the optimization of their expected utilities. The paradigm of bounded rationality takes a step toward greater realism by placing computational limitations on agents' abilities to determine optimal decisions. Behavioral and cognitive studies reveal that humans are also categorically bounded, meaning that they use a finite categorization of the set of decision problems that may be posed, with a small number of categories. This project focuses on the use of quantization theory and information theory to establish foundations for the interplay between categorization and decision making. We aim to understand the impact of categorization on individual decision making, team decision making through voting, and sequential decision making. The theory developed in the project will include both analysis of situations in which the categorization is fixed and optimal design of categorizations. In addition to the behavioral justification for the study of categorization, informational limitations on learning suggest that categorization into classes of decision problems has ramifications for engineering design.

This project has the potential to influence economic theory and the understanding of certain social and organizational phenomena. Specifically, the project offers a way to understand the decision making performance of teams, with the incorporation of certain human limitations and the potential for differing preferences (e.g., between Type I and Type II errors) among teammates. Differences in preferences lead to a quantifiable penalty of team discord, even when the team shares the common goal of making correct decisions. There is also a quantifiable advantage from team diversity in the sense of obtaining better performance when teammates apply different categorizations. These new concepts could contribute to principles for team formation and for how data gathering policies can be optimized with the goal of fair decision making.

Publications

• Quantization of Prior Probabilities for Hypothesis Testing
  K. R. Varshney and L. R. Varshney, IEEE Trans. Signal Processing, vol. 56, no. 10, pp. 4553-4562, October 2008.
• Mean Bayes Risk Error is a Bregman Divergence
  K. R. Varshney, IEEE Trans. Signal Processing, vol. 59, no. 9, pp. 4470-4472, September 2011.
• Quantization of Prior Probabilities for Collaborative Distributed Hypothesis Testing
  J. B. Rhim, L. R. Varshney, and V. K. Goyal, IEEE Trans. Signal Processing, vol. 60, no. 9, pp. 4537-4550, September 2012.

• Categorical Decision Making by People, Committees, and Crowds
  L. R. Varshney, J. B. Rhim, K. R. Varshney, and V. K. Goyal, Proc. Information Theory and Applications Workshop (La Jolla, CA), February 2011.
• Collaboration in Distributed Hypothesis Testing with Quantized Prior Probabilities
  J. B. Rhim, L. R. Varshney, and V. K. Goyal, Proc. IEEE Data Compression Conf. (Snowbird, UT), March 2011, pp. 303-312.
• Conflict in Distributed Hypothesis Testing with Quantized Prior Probabilities
  J. B. Rhim, L. R. Varshney, and V. K. Goyal, Proc. IEEE Data Compression Conf. (Snowbird, UT), March 2011, pp. 313-322.
• Multilevel Minimax Hypothesis Testing
  K. R. Varshney and L. R. Varshney, Proc. IEEE Int. Workshop on Statistical Signal Process. (Nice, France), June 2011, pp. 109-112.
• Distributed Decision Making by Categorically-Thinking Agents
  J. B. Rhim, L. R. Varshney, and V. K. Goyal, Proc. NIPS Workshop on Decision Making with Multiple Imperfect Decision Makers (Sierra Nevada, Spain), December 2011.
• Benefits of Collaboration and Diversity in Teams of Categorically-Thinking Decision Makers
  J. B. Rhim, L. R. Varshney, and V. K. Goyal, Proc. 7th IEEE Sensor Array and Multichannel Signal Processing Workshop (Hoboken, NJ), June 2012, pp. 181-184.
  • 1st Place in Student Paper Contest

Team

Acknowledgments

This material is based upon work supported in part by the National Science Foundation under Grant No. 1101147 of the Interface between Computer Science and Economics & Social Sciences (ICES) program. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.