Robot Mental Healthcare

I am not deeply considering the question of whether real AI agents will go insane, especially since the sorts of AI agents depicted in these movies are many years away. But, it is interesting to speculate about what characteristics of AI might lead to insanity on the part of the agent.

In humans, insanity seems to arise because of deficient or defective perceptual, memory or reasoning mechanisms. As well as these physical disorders, mental impairment could arise from a variety of social causes such as isolation, inadequate or inappropriate socialization, sensory deprivation, torture, mental cruelty, antisocial companionship, lack of attention, contradictory obligations, hazardous circumstances, etc. AI agents are unlikely to be different. Also, because AI agents have different reasoning mechanisms, they are likely to manifest unexpected aberrant behavior as well, but the forms of this are harder to predict.

In the same way as human society has a variety of ways to cope with aberrant human behavior, we should anticipate movies exploring how aberrant AI agents might be controlled. Thus, as AI sophistication develops, we might see movies that investigate how human society might:

• Detect, prevent, control, or repair `insane' AI agents (such as isolation, group therapy, `psychotherapy', rehabilitation and destruction).
• Create an acceptance of and a way to act in irreconcilable conflict situations, perhaps by using a structured behavior priority system, such as Asimov's Three Laws of Robotics [3], to help resolve goal conflicts.
• Create an acceptance of agent `death', or limiting the `will' to survive.
• Engage AI agents in a `society' that provides social feedback, behavior modification and constraints.
• Use `drug' control (e.g. by producing modifiers to limit reasoning and perceptual abilities).