Artificial Intelligence between Cognition and Motivation
Summary: Here I re-stated the argument that current Artificial Intelligence (AI) is cognition-oriented, and suggested that cognition is not enough to mimic human behavior because it cannot define internal context. For that AI will need to incorporate models of motivation.
Cognition and motivation are different mechanisms, each one contributing to the behavior of living organisms in different ways. Common sense tells us that humans show more cognitive behavior than animals. Dictionary definitions of cognition describe it as awareness and knowledge of the outside world. The categories of the outside world are generally fixed in time, or static: you can’t compare (the cognitive categories of) apples and oranges irrespective of time. There are numerous views linking cognition to Artificial Intelligence (AI) such as
Artificial Intelligence theory is the engineering counterpart of ‘cognitive science’, I suggest that AI at present is cognition-oriented; AI activities largely involve building intelligent tools based on static categories in order to save time and labor. However, AI involves more than building software tools; it is
concerned with making computers behave like humans, and in order to fulfill this role I believe that AI will need more than cognition.
On the other hand, animals show more motivational behavior than humans. One definition of motivation is:
an internal state or condition that activates behavior and gives it
direction. The root of the word “motivation” comes from moving, the implied time dependency of the internal state is obvious. Motivation is dynamic because its state changes with time. Progression from static to dynamic can be seen as natural as going from simple to complex; it makes sense to expect the progression from cognition to motivation. But what does motivation bring to the table that’s not found in cognition? In a word-context: subjective internal context to be precise and I suggest that motivation defines it. Software tools don’t need a context of their own; they are an extension of the user’s context. A human-like artificial personality does. AI workers and laypersons alike take it for granted that many topics are meaningful standing alone, while in reality they have implicit context. For example, aggression is not really meaningful without a purpose, or a context. Is it for self-preservation or to protect someone else, or to feed oneself? And hunger for food, is eating a response to homeostatic need or a defensive reaction? This is context! Pets, are they companions, workers or toys? And humor, why do some of us see a joke in context while others don't? Even rationality is not rational without context! You can be objective and answer "all of the above" to such questions, but this is not a good answer if you want to mimic human behavior; humans are not only objective, they are also subjective. The subjectivity should be modeled otherwise computers cannot “behave like humans.”
Conclusions: Cognition alone is not enough to mimic human behavior for AI, perhaps a model of motivation is more suited as a starting point than a cognitive one, because motivation will create the internal context within which cognition can process information like a human.
About the author: Faisal L. Kadri is an independent researcher not affiliated with any educational institution. His research interest since 1986 is in applying the mathematical tools of nonlinear systems engineering to modeling motivational mechanisms in animals and humans. For more information please visit:
www.artificialpsychology.com
Copyright © 2006 Faisal L. Kadri, all rights reserved. Reproduction of this article in Internet media is permitted on condition that all links are maintained. Use in college essays welcome.