Biological Models for Emotion

To gain a better understanding of emotional reactions in mammals, it's necessary to model their cognitive abilities at least in part. Indeed, animals must perceive and interpret their surroundings before they can feel emotions.

Researchers in various fields related to AI attempt to reproduce different aspects of biological creatures to simulate adaptive behaviors. Generally, systems are based on two different components: cognitive and emotional. This is often done in the spirit of nouvelle AI, but with additional biological insights:

• The emotional component is often hard-coded (phylogenetic). With this approach, the designer can impose biologically accurate emotions (for instance, fear, pain, and pleasure) and test their role in the learning. Complementary emotions can be considered as extreme values of a single variable, used to drive the learning toward positive emotions.

• The adaptive component provides the ability to learn behaviors dynamically during the simulation (ontogenetic). Techniques such as neural networks and reinforcement learning allow the creature to adapt to its emotional status and behave in a better fashion (as explained in Part VII).

The primary objective of such biologically inspired approaches is to build understanding in mammals. However, it's possible to investigate the benefits of emotions as a tool for creating intelligent creatures—with less focus on biological accuracy.