Foundations Of Cognitive Science

Credit Assignment Problem

The credit assignment problem concerns determining how the success of a system’s overall performance is due to the various contributions of the system’s components (Minsky, 1963).  “In playing a complex game such as chess or checkers, or in writing a computer program, one has a definite success criterion – the game is won or lost. But in the course of play, each ultimate success (or failure) is associated with a vast number of internal decisions. If the run is successful, how can we assign credit for the success among the multitude of decisions?” (p. 432). This kind of problem was important to the decline of Old Connectionism, and the birth of New Connectionism.  The credit assignment problem that faced Old Connectionism was its inability to assign the appropriate credit – or more to the point, the blame -- to each hidden unit for its contribution to output unit error.  Failure to solve this problem prevented Old Connectionism from discovering methods to make their most powerful networks belong to the domain of empiricism, and led to its demise (Papert, 1988).  New Connectionism arose when this problem was solved, permitting multi-layered networks to be trained (Rumelhart, Hinton, & Williams, 1986).  For instance, the backpropagation of error algorithm defines hidden unit error as the total weighted error signal coming from output units through connections between output units and a hidden unit.


  1. Minsky, M. L. (1963). Steps toward artificial intelligence. In E. A. Feigenbaum & J. Feldman (Eds.), Computers And Thought (pp. 406-450). New York, NY: McGraw-Hill.
  2. Papert, S. (1988). One AI or many? Daedalus, 117(1), 1-14.
  3. Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by back-propagating errors. Nature, 323, 533-536

(Added November 2010)