Abstract for AISB 2000: How to Design a Functioning Mind

Poster Abstract for the
Symposium on How to Design a Functioning Mind
17-18th April 2000
At the AISB'00 Convention

AUTHOR: Stefan Kuenzell

    Universitdt Gie_en
    Institut f|r Sportwissenschaft
    Kugelberg 62
    35394 Gie_en

    tel. +49-641-9925232

TITLE: "Learning the basics"


We need to learn to behave flexibly in an ever changing environment. Probability to survive increases if we predict environmental changes correctly. This can only be done if we discriminate between what happens by the physics of the environment and what we cause to happen through our action, i.e. the difference between the world and the self. This is learned at first in the "motor babbling" stage. Movements are produced in a random fashion, their proximal and distal effects are perceived. This leads to a concept of the self, to a mapping between the environment and the perceived environment, i.e. perception, and to a mapping between action, perceived environment and distal effects, i.e. a forward model (for the engineer) or a motor intuition (for the psychologist). The latter mapping enables us to anticipate our action's distal results. Once we are able to anticipate the results, we might want to produce them. So we have to learn the mapping between desired situation, perceived environment and motor behaviour, i.e. a inverse model (for the engineer) or motor control (for the psychologist).

In general, to enable learning we must have four concepts (implicitly) available in our mind: The perceived actual situation, the desired situation, the perceived true result available at the moment of the occurrence of the distal effect, and the anticipated result available at the moment of action. For learning, the last three are compared crosswise. We can distinguish five cases:

1. If the true result equals the anticipated result, but both do not equal the desired situation, motor control has to be learned. This is a usual case.

2. If the desired situation equals the anticipated result, but both do not equal the true result, perception must be differentiated. This is the case in novel situations. The environment's variance is mainly detected because the anticipated effect of a well-known action in an only seemingly well-known situation does not come true.

3. If the desired situation equals the true result, but not the anticipated result, motor intuition must be learned. This is the case in trial and error learning, when suddenly, and not anticipated, action leads to the desired situation.

4. If all of them equal each other, everything is (presumably) fine and nothing must (can) be learned. Improvement is only possible through presentation of explicit, consciously mediated knowledge of result.

5. All three situations do not equal each other. It is time for some motor babbling again.

For implicit learning, the actual situation and the action's effect must be experienced. It is necessary to act. It is the privilege of self conscious subjects to act cognitively instead of physically, to 'act as if you were acting'. A more or less correct motor intuition (or its conscious equivalent, motor imagery) and a concept of the 'self' presumed, distal results can be predicted mentally without acting. This protects consciously planning subjects from experiencing undesired or even lethal consequences, which enhances clearly the probability of survival of subjects and species.


1995: Exam in computer science (neural networks, Prof. Raul Rojas) and
Sport Science (Prof. Klaus Roth) at the Freie Universitdt Berlin
1995 - 1997: Teacher at the Walther-Rathenau-Gymnasium in Berlin,
Subjects Computer Science, Physical Education
1997 - now: Assistent of Prof. Munzert at the Justus-Liebig-Universitdt
Gie_en, Special Interests: Sport psychology, motor control and learning

Publications in the direction of the presented abstract:
K|nzell, S. (1996): Motorik und Konnektionismus (Motor Behaviour and
Connectionism). Kvln, bps.
K|nzell, S. (1997): Computermetapher und neuronale Netze: Wie
informationstheoretische Modelle kognitiver Prozesse die Trainingspraxis
beeinflussen. [Computer metaphor and neural networks: How information
processing models of cognitive processes influence every day training in
sport skills] In Perl, J. (Hrsg.), Sport und Informatik V (S. 108-125).
Kvln: Sport und Buch Strau_.
K|nzell, S. (1999). Implicit and explicit motor learning from a
connectionist point of view. In Parisi, P., Pigozzi, F, & Prinzi, G.
(eds.), Proceedings of the 4th Annual Congress of the Europan College of
Sport Science. (p. 767). Rome: IUSM.

Current research interest
Detecting temporal dependences between an inverse and a forward model of
human behaviour by comparing neural network simulations to experimental
human subject data.