Inductive Learning with External Representations

  • Mark Wexler


External representation is the use of the physical world for cognitive ends, the enlargement of the mechanisms of representation to include the action-perception cycle. It has recently been observed that such representation is pervasive in human activity in both pragmatic and more abstract tasks. It is argued here that by forcing an artificial learning system to off-load all of its representation onto a (simulated) external world, we may obtain a model that is biased in a very natural way to represent functional relations in ways similar to those used by people. After learning a function from examples, such a model should therefore generalize to unseen instances in ways that we would consider correct. These ideas are tested by developing two machine learning systems, in which representation relies on the sensorimotor control of simulated robotic agents. These systems are able to represent a variety of functional relations by means of their action and perception, and they learn to spontaneously do so from examples. Moreover, they generalize extremely well to unseen problems even after a small number of examples, including on functions such as n-parity that are notoriously difficult to generalize for machine learning algorithms. It is argued that despite these systems’ simplicity, the external representations that they evolve are similar to those used by people on similar tasks.


Generalization Performance External Representation Inductive Learn Inductive Bias Length Parity 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. Brooks, R. A. & Stein, L. A. (1993) Building brains for bodies. MIT AI Memo 1439.Google Scholar
  2. Clark, A. & Chalmers, D. (1996) The external mind. Manuscript, Dept. of Philosophy, Washington University.Google Scholar
  3. Clark, A. & Grush, R. (1997) Towards a cognitive robotics. Manuscript, Dept. of Philosophy, Washington University.Google Scholar
  4. Clark, A. & Thornton, C. (1997) Trading spaces: Computation, representation and the limits of uninformed learning. Beh. Brain Sci. 20: 57–90.CrossRefGoogle Scholar
  5. Goodman, N. (1983) Fact, fiction, and forecast. Cambridge, MA: Harvard Univ. Press.Google Scholar
  6. Harnad, S. (1990) The symbol grounding problem. Physica D 42: 335–46.CrossRefGoogle Scholar
  7. Kirsh, D. (1992) From connectionist theory to practice. In: M. Davis (ed.) Connectionism: Theory and practice. New York: Oxford Univ. Press.Google Scholar
  8. Kirsh, D. (1995) The intelligent use of space. Art. Intell. 73: 31–68.CrossRefGoogle Scholar
  9. Koza, J. R. (1992) Genetic programming: On the programming of computers by natural selection. Cambridge, MA: MIT Press.Google Scholar
  10. Mitchell, T. (1980) The need for biases in learning generalizations. Technical Report, Department of Computer Science, Rutgers University, No. CBM-TR-117.Google Scholar
  11. Pollack, J. (1992) The induction of dynamical recognizers. Machine Learning 7: 227–252.Google Scholar
  12. Pouthas, V. (1995). Timing behavior in young children: A development approach to conditioned spaced responding. In: Michon, J. A. and Jackson, J. L. (eds.) Time, mind, and behavior. Berlin: Springer-Verlag.Google Scholar
  13. Quartz, S. and Sejnowski, T. (1997) The neural basis of cognitive development: A constructivist manifesto. Beh. Brain Sci. 20: 537–596.CrossRefGoogle Scholar
  14. Wexler, M. (1996) Embodied induction: learning external representations. AAAI1996 Fall Symposium.Google Scholar
  15. Wexler, M., Kosslyn, S. M. & Berthoz, A. (1997) Motor processes in mental rotation. Cognition 68: 77–94.CrossRefGoogle Scholar
  16. Wolpert, D. (1996) The lack of a priori distinctions between learning algorithms. Neural Computation 8: 1341–1390.CrossRefGoogle Scholar
  17. Zhang, J. and Norman, D. A. (1994) Representations in distributed cognitive tasks. Cognitive Science 18: 87–122.CrossRefGoogle Scholar

Copyright information

© Kluwer Academic/Plenum Publishers 1999

Authors and Affiliations

  • Mark Wexler
    • 1
  1. 1.Laboratoire de la Physiologie de la Perception et de l’ActionParisFrance

Personalised recommendations