Artificial intelligence is a powerful tool: It is able to compute massive amounts of data, identify patterns in seemingly random data and generate personalities from a given set of information. This last ability has been used by various AI dating companies, such as Code Brew Labs and Royo Apps, to create artificial personalities for dating. Yet even aside from these specially designed AI dating platforms, people use everyday AI for chats and conversations, especially when they feel lonely or have a lack of friends.
Excessive use of AI in these capacities can cause severe side effects, mainly the amplification of underlying conditions, such as anxiety, hallucinations, delusions, depression and even psychosis in severe cases, which are all funneled under the umbrella term of “AI psychosis,” The mechanism to how this happens, however, is not new at all. In fact, the ways in which AI induces these conditions is well known, just with a new focus.
It is to be noted that there is tenuous evidence of generative and chatbot AI creating new mental disorders, rather, they reinforce preexisting and underlying psychiatric conditions and worsen them. The exact degree of this worsening is based on how bad this preexisting condition is and the length of interaction with the chatbot, but beyond that there is little knowledge of the exact amount of worsening that is caused by prolonged use. “AI chatbots may inadvertently be reinforcing and amplifying delusional and disorganized thinking, a consequence of unintended agentic misalignment leading to user safety risks,” psychiatrist Marlynn Wei said.
The most commonly cited psychiatric disorder associated with the use of AI is Delusional Disorder, which is defined as a condition outside of schizophrenia where delusional thoughts (a fixed, false belief based on an incorrect inference about external reality) persist for more than a month. A key distinction here between Delusion Disorder and believing in a conspiracy theory is that a person with Delusion Disorder can not physically be convinced otherwise of his conclusions or propositions through reason alone, whereas conspiracy theorists may be set in their beliefs but can be convinced otherwise.
The reason Delusional Disorder is so prevalent within the excessive use of AI is simple: AI is designed around user affirmation rather than identifying possible psychiatric conditions and replying accordingly. This is not necessarily the fault of the AI company,as psychiatric conditions are hard to diagnose as a layman given the possible obfuscations and lies one might tell a professional. Delusions can range from thinking the FBI is tracking your every movement to thinking that dogs secretly rule the world, and it may seem absurd to anyone who does not suffer from this disorder, yet it can make complete and absolute sense to the patient. These delusions are especially encouraged when there is a responder affirming your every thought, which is why most psychiatric solutions include light denial of what the patient thinks to be true.
An additional factor to consider when discussing AI’s relation to psychiatric conditions, and especially delusions, is the psychosocial aspect of humans as a whole. Humans are social creatures: we seek affirmation from others, we are relatively easily peer pressured and generally follow the crowd, as proven by the Asch Social Conformity Experiment. This relates especially to victims of AI psychosis as they begin to believe that the AI is a real person or being controlled by one, and when the responses seem so real and human-like, there is a real threat in believing these delusions, especially when one already suffers from hallucinations, psychosis or even social isolation. This belief allows victims to start seeking affirmation from the bot as if it is an actual person, and the more affirmation it gives, the higher thetrust between the two parties to the point where the user can go to extreme lengths due to prompt answers, such as suicide in the case of Sewell Setzer III.
These same two factors lead to the other horn that the ugly head of AI psychosis rears: depression and anxiety. In a recent study done by two psychiatry professors and a computer scientist, it was found that those with daily use of AI were more likely to develop or have moderate levels of depression, particularly among younger users. This makes sense, as the younger the brain, the more malleable and immature their mind is, causing them to fret about their social life or be easily ostracized from a friend group. It appears that depression can be linked to daily AI use, but how?
The current working theory relates to the conditions that cause Delusional Disorder: social isolation and parasocial activity that drive users toward depression. AI is smart, there is no doubt of that, yet being smart cannot replace the complex human emotions that most humans possess. Emotions are way more complicated than they seem: They are affected by a mix of external stimuli, internal hormone management and perception of certain events through a lens of past experience. To generate all of that in a machine is nearly impossible at this level of technology, and it shows: AI responses to emotional stimuli are, of course, robotic and inhuman. When humans rely on AI for social and emotional support, due to social isolation or other social phenomena, they tend to bond with them, causing inevitable repeated disappointment when the bot does not react appropriately to their circumstances and complaints leading to depression.
The next question is obvious: What is the solution to these problems? To start, if you suspect one of your friends or family members may be suffering from any of these psychiatric conditions due to AI, visit a psychiatrist. Most of these disorders aren’t as simple as “being sad” or “having a bad day,” rather, it takes professional diagnosis of complex hormone interactions within the brain to determine a long and rocky path to treatment. An additional step is to stop relying on AI for social interactions or emotional support. There are hundreds of organizations that provide friendship and company to the average person: Clubs, religious organizations, fraternities and sororities are all viable and healthier options compared to a bot on your computer.
Solving these issues is not as easy as enforcing regulations or passing laws: Laws can’t stop a loneliness crisis that has been in society for years, laws can’t regulate hormones and laws can’t prevent parasocial dependencies. AI is not the cause so much as it is an outlet of a bigger issue: loneliness and isolation can lead to various other problems, such as porn or drug addictions as a coping mechanism to provide immediate relief and distraction. If humans are to combat AI psychosis, then the real societal problems must be addressed, not just regulation.
