Psychopathic Robots
Robots that Think Like People Must Also Be Prone to Psychopathic BehaviorA robotics expert has attempted to create a geometrical model of the human psyche, with the aim to develop "psycho-robots" - Artificial Intelligence (AI) systems that exhibit important elements of the human psyche.
Andrei Khrennikov, researcher at the University of Vaxjo in Sweden, says that psycho-robots will have elements of a human-like self-developing psyche, with cognitive systems that evolve through interactions with humans and other psycho-robots. Not only would these robots exhibit emotions and desires like a human, but the machines would also necessarily evolve human psychological problems, including psychopathic behavior.
Khrennikov, who has studied AI psychoanalysis for more than a decade, applied his mental AI model to Freud's psychoanalysis. The researcher's model is based on the geometric representation of mental processes, where ideas are represented by points of "mental space" (similar to metric space). The flow of information—through the unconscious, subconscious, and conscious—is modeled similar to the dynamic evolution of a physical system.
"Our model describes (and can be even used for a mathematical simulation) of such basic features of psychoanalysis as repression of forbidden wishes, desires and impulses (coming to the subconsciousness from the unconscious and going to the consciousness), complexes and corresponding symptoms," Khrennikov explained in a recent paper, "Toward Psycho-robots," currently at arxiv.org. "On the basis of the presented models, we can create AI societies of psycho-robots interacting with real people and observe evolution of the psyche of psycho-robots (and even people interacting with them). … Our aim is precisely to understand the human mind and psychology via AI-modeling."
In past models, Khrennikov and other researchers have used high levels of mathematics, making the models non-readable for people working in AI, computer science, and psychology, he explained. To overcome this obstacle, the current model doesn’t utilize the formal mathematics, and also does not attempt to specify the set-theoretic and topological structure of mental space.
Freudian Psycho-Robots
As Khrennikov describes, a cognitive system has three domains: the unconscious, subconscious, and conscious. An external stimulus is first sent to the unconscious domain, interpreted as a mental state at a control center, and is then produced as an "idea-attractor" (e.g. a wish, desire, impulse, etc.). Khrennikov explains that the idea-attractor can be thought of as the solution to some problem. Interestingly, the unconscious produces new mental states practically automatically, meaning that the unconscious domain - where ideas originate - is not based on the rule of reason.
From the unconscious domain, ideas are then sent to the subconscious domain, which forms a queue of all the ideas waiting to be realized (sent to the conscious domain for action). Ideas in the queue are ordered for successive realizations by a quantitative "measure of interest," where ideas are more interesting when they're new, and lose interest over time. The subconscious control center deletes ideas below a "realization threshold."
Khrennikov describes this geometrically, where the brain measures the distance between ideas. Specifically, the subconscious domain contains a continuously-updated database of interesting ideas formed on the basis of previous mental experiences. It then measures the distance between an idea and the set of interesting ideas, as well as the distance between two ideas. Interest, then, is an inverse relation where smaller distance corresponds to more interesting ideas.
In addition to the measure of interest, the unconscious control center also considers a "measure of interdiction." In other words, some ideas are forbidden by the environment, so that some very interesting ideas may not be realized if they have a high measure of interdiction—or are located at a short distance to the subconscious' set of forbidden ideas.
Taken together, interest and interdiction are defined by a "measure of consistency," and it is this measure that the queue of ideas is ordered by. From here, the most interesting ideas get realized in the conscious domain.
Khrennikov hopes that this model - where the brain is literally described with the metaphor of a computer - will provide the foundation of AI psychoanalytic research for psycho-robots.
Domain of Doubts
But the model doesn't stop there. Khrennikov goes on to explain what happens to the ideas that have a very high measure of interest but are permanently forbidden from being realized. Due to their high interest, these ideas can never be deleted. Where do they go?
Khrennikov introduces the "domain of doubts," where an idea waits for a long time because a cognitive system cannot make a decision about the realization of the idea (on the basis of the measure of consistency). In Freudian terminology, the idea is repressed. Thus, the domain of doubts is the origin of many psychical problems and mental diseases, because the brain can not realize these ideas nor simply delete them.
The human cognitive system never developed a neuronal structure that could 100% isolate such repressed ideas, Khrennikov explains. He goes on to describe how these ideas can still influence idea-attractors, and cause disguised substitutes to be realized in place of the forbidden ideas. The significant part of this property of the cognitive system is that it is inherent, and therefore, psycho-robots that exhibit elements of the human psyche must also necessarily be prone to psychotic behavior.
Like many other robotics researchers, Khrennikov thinks that robots with a more human psyche will be very useful as domestic robots, providing a more intuitive interface. Despite potential "psychopathic robots," Khrennikov explains that such forms of artificial intelligence could not only improve human's lives through physical labor, but also by providing a method for understanding our own human psyche.
Source: Robot World News