The Nature of Psychosis
Our traditional understanding and classification of a psychotic state comes from psychiatry, which aims to identify and treat different forms of mental illness. In the psychiatric model, psychosis may include holding delusional beliefs, experiencing hallucinations (stereotypically the experience of 'hearing voices' although such hallucinations may occur in any of the senses) and having muddled or incoherent thoughts (known as thought disorder). The experience of psychosis is commonly associated with schizophrenia or manic depression although it is not necessarily specific to these conditions. Psychosis has been reported to occur in the context of everything from flu to HIV to Parkinson's Disease.
The content of the psychotic experience can be particular to each individual. John Nash, the Nobel Prize winning mathematician, held the delusional beliefs that he was the 'left foot of God' and in charge of a military conspiracy about to take over the world. Kay Redfield Jamison, noted psychologist and author, hallucinated blood dripping from every surface and believed that all the green plants in the world were slowly dying. Not everyone experiences such sinister episodes and some people may come to believe themselves film stars or endowed with unlikely special abilities such as space travel or universal communication.
Simulating the Psychotic Mind
The metaphor of the computational mind has become increasingly popular in psychology, replacing the stimulus-response theories that were popular in the first half of the 20th century. With a theory promoting thinking and brain function as computation and the widespread availability of computers, psychologists were quick to create artificial models of how the mind (or at least a part of it) might work.
Although usually more concerned with patient assessment and therapy, the development of computer based models of mind and brain caught the attention of clinicians working with people who were diagnosed with mental illness or had suffered brain injury. They wanted to take these simulations of normal thinking a stage further, to see what happened after they were exposed to simulated damage. Particularly with a view to seeing if the models made any useful clinical predictions or helped explain how the damage took effect.
An early simulation of paranoia by psychiatrist Kenneth Colby, based his theory on a flow chart understanding of the mind, charting mental function as a process of manipulating symbols, segments and sequences of natural language thinking. Procedures such as the 'self-scanner' would check self generated 'speech' for topics related to currently held delusions and would increase the 'FEAR' variable if found, similar procedures would affect the values of 'MISTRUST' and 'ANGER', supposedly simulating the levels of these emotions during social interaction.
Compared to more recent and sophisticated simulations Colby's flowchart approach (published as a 1975 book) seems a little crude. Perhaps even a little naive, taking the popular approach to programming at the time and applying it wholesale to the mind. However, Colby is in good company. As psychologist Douwe Draaisma has charted in his history of ideas about the mind, psychology has a long tradition of co-opting current technological developments as a metaphor for mental function.
More recent simulations have tended to use artificial neural networks, usually 'trained' to simulate certain mental functions and then 'damaged' to simulate illness or injury. Neural networks have a number of advantages for scientists. Principally, they aim to simulate (however loosely) the workings of neurons and do not assume any particular method for representing knowledge beforehand. Although it is controversial as to how accurate a computer model of neurons can and should be when simulating behaviour, artificial neural networks at least have a level of plausibility in combining approaches which aim to understand the link between the biology of the brain and the phenomenon of the mind.
Ralph Hoffman and Thomas McGlashan, two psychiatrists from Yale, created ambitious neural network models of memory and speech perception. They were particularly interested in 'damaging' their computer models in a way that simulated the brain changes in schizophrenia. Crucially they hoped that their simulation would become psychotic, perhaps exhibiting virtual delusions or hallucinations, suggesting a way in which these experiences might arise from a disordered brain.
Their simulations produced strange and unrequested output when damaged, output which Hoffman and McGlashan consider to be the equivalent of psychosis in humans. Furthermore, they concluded that levels of activation in brain areas involved in speech perception may be important in producing hallucinated voices. Applying the conclusions from their simulation Hoffman and McGlashan used a technique called Transcranial Magnetic Stimulation (where magnetic pulses are used to temporarily stimulate small areas of brain) to successfully treat a series of people who were being troubled by hallucinated voices.
There is no doubt that this was a successful and useful application of simulated psychosis, leading to important scientific findings and a hopeful practical treatment. Nevertheless, psychosis is notoriously difficult to pin down and it is becoming clear that the boundaries between computer science, philosophy and medicine will need to be broken down before we can be sure that similar simulations are not useful by accident rather than by design.
Better Living Through Madness
Our idea of what psychosis consists of has been inherited from psychiatrists. Because of their role as medics for mental distress, they have traditionally suffered from a sample bias. A person is only likely to show up in front of a psychiatrist if they are either distressed or causing distress to others. A person who has wild and extensive hallucinations is unlikely to ever be a psychiatric patient if they are never troubled and can continue their lives successfully. Many prophets and visionaries throughout history and in modern times have had a radically altered perception of reality but have contributed much of benefit to the world.
Furthermore, there is increasing evidence that a significant minority of the population hold strange and unusual beliefs and may have sensory experiences that would otherwise be considered as part of psychosis if it were not for the fact that they are rarely troubled by them. It seems that psychosis may not always be a sign of mental illness, but simply another way of constructing reality.
If the 'illness' part of mental illness is not to do with the strange experiences, but with the distress caused by them, computer models may eventually be inadequate. Whilst computer simulations may be able to produce virtual hallucinations, they are unable to produce virtual distress. This cuts to the heart of the debate about artificial intelligence, as it could be argued (and frequently is) that computers are incapable of adequately producing emotional states and always will be. Scientists could simply add a 'DISTRESS' variable, but the problem of when this should be increased would remain unsolved, putting the project back to square one.
Philosophically, the problem remains unsolved but the fact that useful treatments are starting to emerge from computer simulations is reason enough to continue. And with our history of using technological metaphors to understand the mind, future technology may yet hold the key to cracking the core of psychosis.