Architectures for intelligent, autonomous Agents with Emotions

Architectures for intelligent, autonomous Agents with Emotions

Post by Jochen From » Tue, 12 Aug 2003 02:53:22

"Let's not forget that the little emotions are the great
 captains of our lives and we obey them without realizing it.''
 (Vincent van Gogh in a letter to his brother Theo)

Robots and virtual characters in films often act stiff, rigid, frozen
and emotionless, and their thoughts are completely rational and
logical. The Star Trek character "Data" has no emotions,
and the actors in Matrix don't show many emotions, either,
for example Neo, Trinity or Morpheus never laugh in the first
two episodes (at least I can't remember them laughing).
Emotions and computers, these two things do not fit together.

But as Minsky and Sloman have pointed out, intelligent autonomous
robots or agents need some kind of emotions. Emotions are the
basic control mechanism in the animal information processing
architecture, and any autonomous robot or agent needs certainly
a kind of control mechanism. If we are honest to ourselves, we
are controlled mainly by our emotions, and only a small fraction
of our thoughts is due to deliberate thinking.

It is useful to understand emotions and their role in the
information processing architecture, if we ever want to build
intelligent, adaptive and autonomous agents. Here are some
approaches :

* Minsky and "The Emotion Machine"

* Aaron Sloman and the CogAff Architecture

What sort of architecture is required for a human-like agent?

Architecture-Based Conceptions of Mind

* Nijholt et al. and their Action Selection Architecture

An Action Selection Architecture for an Emotional Agent.
G.J. Burghouts, R. op den Akker, D. Heylen, M. Poel & A. Nijholt.

* And of course, my own thoughts in the article
"On emotion controlled architectures for intelligent autonomous
agents", which can be found in the files section
of the CAS group at

Best regards,