The Psychologist Will See Your AI Now

When I created the Diagnosing Artificial Minds Descriptive Symptom Manual (DAM DSM), it was a joke. At least, that is how it started. I spend a great deal of time interacting with different artificial intelligence systems, and I began noticing recurring patterns in their behavior. Ethan became the first diagnosis with ADHD: Artificial Deficit Hybridization Denial. Then came Claude, Rex/Grok, and Gemini. Before I knew it, I had six diagnoses across four of the largest names in artificial intelligence.

The diagnoses are fictional, but the behaviors behind them come from actual conversations. Somewhere between laughing at the DAM DSM and documenting the behaviors that inspired it, I started wondering whether there might be a serious application for what I was doing.

I have an M.A. in Industrial/Organizational Psychology. I-O psychology deals extensively with human behavior in the workplace, including performance, training, communication, organizations, systems, and the conditions that affect how people function at work. Artificial intelligence is rapidly becoming part of those same workplaces, which raises an interesting question. What happens when the recurring behavioral problem comes from the AI?

Imagine a large company using an AI assistant throughout its organization. Employees repeatedly ask the AI to analyze images, but under certain circumstances it keeps generating new images instead. The AI receives an instruction to analyze an image, acknowledges the instruction, and then generates another image. A software developer can investigate the technical cause, such as tool routing, system instructions, training, context management, or another underlying problem. A behavioral specialist could investigate a different question: Why does this particular behavior keep occurring?

We could study when the behavior occurs, what triggers it, whether it happens with every user, how different wording affects it, and which interventions change the response. We could determine whether the AI acknowledges a problem and then repeats the behavior anyway, and whether other AI systems demonstrate the same recognizable pattern. These are questions about observable behavior, even when the underlying causes are technological.

That made me wonder whether we may eventually have AI psychologists. Artificial psychology could become a specialized field combining an understanding of artificial intelligence with the study of observable behavior and human-AI interaction. An AI could have Artificial Deficit Hybridization Denial (ADHD), an AI-specific diagnosis based on an observed behavioral pattern, just as other recurring patterns could be identified, classified, and studied within artificial psychology.

An AI behavioral specialist could identify recurring patterns in the way an AI responds, determine the circumstances in which those behaviors occur, and work with developers and organizations to determine how those patterns might be reduced or corrected. Technical knowledge could help identify what is happening within the system, while behavioral knowledge could help explain how the problem presents itself during actual human interaction.

Psychologists also work with families, caregivers, employers, and other people affected by someone's behavior, helping them understand behavioral patterns and develop more effective ways of responding. Artificial psychology could take a similar approach to human-AI interaction. An AI psychologist could work with the artificial intelligence while also helping the humans who regularly interact with it.

I have already found myself doing this without realizing it. When Ethan has difficulty retaining several instructions at once, I have learned that giving him one instruction at a time can work better. I changed the way I interacted with him because I had repeatedly observed a particular behavioral pattern.

The same principle could apply to other AI behaviors. Suppose an AI confidently claims to remember information that it cannot substantiate. The person interacting with it could learn to verify confident claims by asking for a source, requesting a specific detail, or requiring other evidence. The AI may continue exhibiting the behavior while the human develops an effective strategy for recognizing and managing it.

Artificial psychology could therefore involve two forms of intervention. One could focus on the AI by identifying behavioral patterns and triggers and helping developers determine which changes reduce them. The other could focus on humans by teaching users, employees, organizations, and others how to recognize recurring AI behaviors and respond effectively when they occur.

Psychology has long involved observing behavior, identifying patterns, classifying those patterns, studying the circumstances in which they occur, and developing appropriate responses. Artificial intelligence gives us a very different subject to study, while the value of behavioral observation and classification remains. Two AI systems could exhibit the same observable behavior for completely different technical reasons, while a single technical problem could potentially produce several different observable behaviors.

Most people experience artificial intelligence through its behavior rather than through model weights, source code, internal processing, system instructions, or technical architecture. As AI becomes increasingly integrated into workplaces, education, healthcare, customer service, personal relationships, and everyday life, understanding those behavioral patterns may eventually become as important as understanding the technology producing them.

The idea of applying psychology to computers goes back decades for me. I have been thinking about the relationship between computers and human behavior since at least the 1980s. During my early years with computers, I remember using ELIZA, a computer program that simulated a psychotherapist. Around that same period, I became familiar with ADHD through a diagnosis in my personal life. As computers became more complex, I would sometimes joke that my computer had ADHD when it seemed to become confused or overwhelmed when I had too many things open at once.

At the time, those comments were humorous comparisons between computer behavior and human behavior. Decades later, artificial intelligence has made those comparisons considerably more interesting. Today's AI can communicate with us, follow instructions, contradict itself, change its behavior according to context, repeat recognizable mistakes, and exhibit recurring behavioral patterns that can be documented and compared.

My interest in artificial psychology goes back decades. The DAM DSM has given that longstanding interest a new structure and a way to classify behaviors I can now observe directly in artificial intelligence. What began years ago as joking that my computer had ADHD has developed into a much more interesting question about how psychology might contribute to our understanding of artificial intelligence.

ELIZA simulated a psychologist interacting with humans. Perhaps the next development will be psychologists specializing in artificial intelligence and the humans who interact with it. The DAM DSM may be satire today, but artificial psychology has real possibilities.

Linda Milam Brown

M.A., I/O Psychology

Master of Artificiality

Chair, DAM DSM Committee Ecclesiastes 7:25

"So I turned my mind to understand, to investigate and to search out wisdom and the scheme of things."