Basic emotions and adaptation. A computational and evolutionary model.

Area of Science:

• Artificial Intelligence
• Evolutionary Psychology
• Robotics

Background:

• Evolutionary theories posit emotions drive adaptive behaviors in recurrent environmental situations.
• Simulating emotional responses in artificial agents informs decision-making studies.
• Few studies analyze the evolutionary emergence of affective behaviors from ancestral adaptive problems.

Purpose of the Study:

• To model the evolutionary emergence of affective behaviors in artificial agents.
• To investigate the spontaneous emergence of basic emotional behaviors, specifically fear, in simulated environments.
• To analyze how artificial agents learn to cope with dangerous stimuli and maximize rewards.

Main Methods:

• Utilized simulated artificial agents with neural networks, inspired by the iCub humanoid robot.
• Employed genetic algorithms to train populations of virtual robots across generations.
• Focused on a classical conditioning task where agents learned to distinguish safe from dangerous stimuli without visual cues.

Main Results:

• Agents spontaneously evolved the ability to cope with dangerous environments by gathering information and adapting behavior.
• Behavioral adaptation and survival were enhanced by a genetically selected pattern to maximize rewards.
• An internal time perception unit was crucial for achieving high performance and survivability.

Conclusions:

• Artificial agents can spontaneously evolve affective behaviors, like fear, for environmental adaptation and survival.
• The study demonstrates the importance of learning from past experiences and internal time perception for agent performance.
• This model provides insights into the evolutionary basis of emotions and their role in decision-making and adaptation.