In fact, everytime that we have a limitation in term of perception, memory, analysis, we are exposed to the risk of wrong interpretations of the present situation in which we have to take a decision.
In Science, the same thing happens.
The first examples of such effect in term of reasoning were given in the field of AI and more specifically by the programmers of chess engines.
In 1973, Hans Berliner(1) described the Horizon Effect during a conference at Stanford University.
He described the phenomenon as constituted of 2 parts:
- a negative part creating diversions which:
- ineffectively delay an unavoidable consequence
- or make an unachievable consequence to appear achievable.
- a positive part pushing much too soon a consequence that can be imposed on an opponent at leisure, frequently in a more effective form.
Clearly, the Horizon Effect is a consequence of the depth limitation in the search algorithm and the challenge is to limit its effects.
Recently, it was proposed the following definition: "In computer game playing or other search processes, a large search tree has to be explored. It is usual to set a maximum depth limit (D) beyond which it is considered uneconomic to search further. The horizon effect refers to the fact that interesting results will always exist beyond any depth D and therefore in any given search will not be discovered. Variable evaluation functions and dynamic search-depth controls have been used in attempts to deal with this problem." (A Dictionary of Computing, Oxford University Press)
to be continued...
1- Berliner, Hans J. (August 20–23, 1973). Some Necessary Conditions for a Master Chess Program. Proceedings of the 3rd International Joint Conference on Artificial Intelligence. Stanford, CA, USA, August 20–23, 1973: 77–85.