MinimalCognitiveBehaviour

Prev: Body Evolution and Morphological Computing	Up: Home	Next: Learning by demonstration

Table of contents

• Minimal Cognitive Behaviours

Minimal Cognitive Behaviours

The term Minimal Cognitive Behaviour have been introduced by Randall Beer (1996) to describe experiments that enable to study the evolution of cognitive capabilities, such as categorization or attention, in experimental settings which are as simple as possible. The simplicity of the agent and of the environment, in fact, can enable the experimenter to carry on detailed analysis of the agent/environmental dynamics that can be important to elucidate the mechanisms that are at the basis of the studied cognitive capacity.

In the research carried out by Randall Beer and collaborators, this has been typically realized using agents which can move only left or right along a single dimension provided with a simple “eye’ realized by using seven sensors that can measure the relative distance of simple objects falling down from the top with varying direction and speed (see Figure). The agents’ neural controllers include seven corresponding sensory neurons, a variable number of internal neurons with recurrent connection, and two motor neurons that control the direction and the speed of the agents’ movement along the horizontal axis. This type of simple experimental setting can be used to study different cognitive capacities. For example, experiments involving circle-shaped and diamond-shaped falling objects, in which the agent should “catch” (i.e. collide with) the former but not the latter type of objects, can be used to study active categorization (Beer, 2003). Experiments involving two circular objects falling down at different time and locations, in which the agent should catch the first and the second landing object in sequence, can be used to study selective attention (Goldenberg et al., 2004).

The MinimalCognitiveBehaviourExperiment plugin can be used to replicate the experiments on selective attention described in (Goldenberg et al., 2004) and can be used as a basis for implementing other related experiments.

References

Beer, R.D. (1996). Toward the evolution of dynamical neural networks for minimally cognitive behavior. In P. Maes, M. Mataric, J. Meyer, J. Pollack and S. Wilson (Eds.), From animals to animats 4: Proceedings of the Fourth International Conference on Simulation of Adaptive Behavior (pp. 421-429). MIT Press.

Goldenberg, E., Garcowski, J. and Beer, R.D. (2004). May we have your attention: Analysis of a selective attention task. In S. Schaal, A. Ijspeert, A. Billard, S. Vijayakumar, J. Hallam and J.-A. Meyer (Eds.), From Animals to Animats 8: Proceedings of the Eighth International Conference on the Simulation of Adaptive Behavior (pp. 49-56). MIT Press.

Beer, R.D. (2003). The dynamics of active categorical perception in an evolved model agent. Adaptive Behavior 11(4): 209-243.