Problem solving

  • Egri, L., & Shultz, T. R. (2015). Constraint-satisfaction models. In J. D. Wright (Ed.), International Encyclopedia of the Social & Behavioral Sciences (2 ed., Vol. 4, pp. 716–723). Oxford: Elsevier. - A survey of constraint-satisfaction models covering both symbolic approaches from artificial intelligence (AI) and neural network approaches from psychology and cognitive science. Newer constraint satisfaction research has made considerable progress with systems that optimize solutions to hard problems even when it is impossible to satisfy all constraints perfectly.
  • Dandurand, F., Cousineau, D., & Shultz, T. R. (2012). Solving nonogram puzzles by reinforcement learning. In N. Miyake, D. Peebles, & R. P. Cooper (Eds.), Proceedings of the 34th Annual Conference of the Cognitive Science Society (pp. 1452-1457). Austin, TX: Cognitive Science Society. - Reinforcement-learning (RL) solvers learn near-optimal solutions that outperform a heuristic solver with the explicit rules often given to Hanjie players. Only RL solvers that used a neural-network to remember the quality of actions generalized their knowledge to generate good solutions. pdf
  • Dandurand, F., Shultz, T. R., & Rey, A. (2012). Including cognitive biases and distance-based rewards in a connectionist model of complex problem solving. Neural Networks, 25, 41-56. - A connectionist-based model of complex problem solving that integrates cognitive biases and distance-based and environmental rewards under a temporal-difference learning mechanism.
  • Prime, H., & Shultz, T. R. (2011). Explicit Bayesian reasoning with frequencies, probabilities, and surprisals. In C. Hoelscher, T. F. Shipley, & L. Carlson (Eds.), Proceedings of the 33rd Annual Conference of the Cognitive Science Society (pp. 1918-1923). Boston, MA: Cognitive Science Society. - People use both priors and likelihoods in Bayesian directions, but the effect of likelihood information is stronger than that of prior information. Use of frequency information and surprisal measures increase deviations from Bayesian predictions. pdf
  • Dandurand, F., & Shultz, T. R. (2009). Connectionist models of reinforcement, imitation and instruction in learning to solve complex problems. IEEE Transactions on Autonomous Mental Development, 1, 110-121. - Computational models of three experimental conditions in learning to solve a complex problem: reinforcement, imitation, and instruction.
  • Dandurand, F., Shultz, T. R., & Onishi, K. H. (2008). Comparing online and lab methods in a problem-solving experiment. Behavior Research Methods, 40, 428-434. - Online and laboratory participants performed similarly in learning to solve a complex problem.
  • Dandurand, F., Shultz, T. R., & Rivest, F. (2007). Complex problem solving with reinforcement learning. Proceedings of the Sixth IEEE International Conference on Development and Learning (pp. 157-162): IEEE. - A Softmax SARSA-based model of reinforcement learning to solve a complex problem. pdf
  • Dandurand, F., Shultz, T. R., & Onishi, K. H. (2007). Strategies, heuristics and biases in complex problem solving. In D. McNamara & G. Trafton (Eds.), Proceedings of the Twenty-ninth Annual Conference of the Cognitive Science Society (pp. 917-922). Mahwah, NJ: Lawrence Erlbaum. - People used means-ends analysis to solve a complex problem. Instruction and imitation each enabled them to overcome symmetry and simplicity biases.
  • Dandurand, F., Bowen, M., & Shultz, T. R. (2004). Learning by imitation, reinforcement and verbal rules in problem-solving tasks. Proceedings of the Third International Conference on Development and Learning: Developing Social Brains (pp. 88-95). LaJolla, CA: The Salk Institute for Biological Studies. - In learning to solve a complex problem, imitation and verbal instruction are each more effective than reinforcement.
  • Shultz, T. R. (2001). Constraint satisfaction models. In N. J. Smelser & P. B. Baltes (Eds.), International Encyclopedia of the Social and Behavioral Sciences (Vol. 4, pp. 2648-2651). Oxford: Pergamon. - A survey of constraint-satisfaction reveals that most modelers favor connectionist over symbolic techniques.