Here's How Generative AI Can Help You Deal With People And Vexing Personality Disorders - Forbes

Using generative AI to simulate personality disorders can be useful, though also controversial.

getty

In today’s column, I am continuing my ongoing series about the impact of generative AI in the health and medical realm. The focus this time is once again on the mental health domain. Specifically, I will carefully examine the ins and outs of using generative AI to gauge what it is like to interact with someone who might have a personality disorder.

I have previously examined numerous interleaving facets of generative AI and mental health, see my comprehensive overview at the link here. You might also find of notable interest a CBS 60 Minutes episode that recently examined crucial facets of this evolving topic, see the link here (I was interviewed and appeared in the episode).

Other useful background includes my coverage of mental health chatbots that have been bolstered by generative AI (see the link here) and the rapidly changing nature of the client-therapist relationship due to generative AI at the link here. I explored where things are headed regarding the levels of AI-based mental therapy autonomous guidance at the link here, and showcased the importance of the World Health Organization (WHO) report on global health and generative AI at the link here, and so on.

Today’s topic entails using generative AI to simulate various personality disorders so that you can gauge how to respond and cope with such circumstances in real life. As you will see, using generative AI to in this manner has benefits, downsides and controversies. There is more to this than meets the eye.

Here’s how I will proceed.

First, we need to set the stage by defining what is meant by referring to the conditions or characteristics known as personality disorders. Many use that term quite casually and almost with abandon these days. I will provide both formal and informal definitions to clarify the matter.

Second, turns out that generative AI and large language models (LLMs) can do a surprisingly engaging job of computationally mimicking or pretending to exhibit the properties of personality disorders. This is done by invoking the personas feature of generative AI, which I’ve covered in-depth at the link here and the link here, just to name a few.

Third, I will walk through a series of generative AI personas that are each devised to express a particular personality disorder. A brief AI-engaged conversation will be shown. The idea is that you will be able to witness the value of such efforts and consider whether you or someone you know might benefit by using generative AI in this fashion.

I think you’ll find fruitful this eye-opening journey.

Let’s dive in.

Personality Disorders And What They Are About

The American Psychological Association (APA) online dictionary indicates that the phrase “personality disorder” is defined this way:

“Any in a group of disorders involving pervasive patterns of perceiving, relating to, and thinking about the environment and the self that interfere with long-term functioning of the individual and are not limited to isolated episodes.” (Source: APA website online dictionary for the term “Personality Disorders”).

I realize that this definition might seem somewhat stodgy and possibly not easily grasped at first glance, especially if you aren’t versed in the mental health domain.

A somewhat more plainspoken definition is given on the Mayo Clinic website which says this:

“A personality disorder is a mental health condition where people have a lifelong pattern of seeing themselves and reacting to others in ways that cause problems. People with personality disorders often have a hard time understanding emotions and tolerating distress. And they act impulsively. This makes it hard for them to relate to others, causing serious issues, and affecting their family life, social activities, work and school performance, and overall quality of life.” (Source: Mayo Clinic website, Diseases & Conditions section, “Personality Disorders” posting).

That seems clearer to the eye and the mind.

There is another facet of special noteworthiness about personality disorders that is further expressed on the Mayo Clinic website:

“In some cases, you may not know that you have a personality disorder. That’s because how you think and behave seems natural to you. You also may think others are responsible for your challenges. There are many types of personality disorders, each with important differences. These disorders are organized into three groups, or clusters, with shared features and symptoms.” (ibid).

A crucial point here is that just because you guess that someone has a personality disorder does not make it so. Nowadays, everybody is an armchair clinical psychologist or self-professed mental health professional. Be cautious in labeling people as possessing a particular personality disorder. It would be wise to allow those who are specially licensed and trained for such mental health determinations to render these serious and potentially life-changing resolves.

Another point of noteworthiness is that a person who perhaps has a particular personality disorder might be entirely unaware of their condition. To them, they are as they are. They seem to be on par with everyone else. Or they realize there is something different about themselves, but do not perceive the difference to be attributable to a personality disorder per se.

This brings up an interesting question for you to mull over.

How prevalent is the condition of having a personality disorder?

Think it over.

Is this a rare condition to have a personality disorder? Does it only occur on a once-in-a-zillion chance? And so on.

Research suggests that personality disorders are a lot more frequent than you might have envisioned. In a meta-analysis research study entitled “The Prevalence Of Personality Disorders In The Community: A Global Systematic Review And Meta-analysis” by Catherine Winsper, Ayten Bilgin, Andrew Thompson, Steven Marwaha, Andrew M. Chanen, Swaran P. Singh, Ariel Wang and Vivek Furtado, The British Journal of Psychiatry, 2019, the researchers made these salient points (excerpts):

“A recent narrative review reported relatively high rates of personality disorders (4.4–21.5%) in community populations across the Western world.”
“Personality disorders are now recognized as important conditions, which are associated with morbidity, premature mortality, and great personal and social costs.”
“Personality disorders are associated with high levels of mental, physical and functional impairment and premature mortality.”
“Despite substantial inter-survey heterogeneity, we found that personality disorders are prevalent globally, affecting a substantial proportion of the population. Epidemiological research on personality disorders is relatively sparse, with a paucity of data from lower-income countries from which to draw comparative conclusions.”

The crux here is that there is indeed a significantly noted prevalence of personality disorders in society, and it is likely to be gaining speed and expanding. Furthermore, lest you think that it doesn’t especially matter whether there is a prevalence, the research emphasizes that the downsides to the individual and society as a whole are considerable.

We shouldn’t be ignoring the situation.

We ought to be giving due attention, increasingly so.

If you are wondering what kinds or types of personality disorders there are, I will bring to your attention a noteworthy research paper that pulled together a multitude of reputable source materials to identify some of the most widely noted personality disorders. A research study entitled “Do You Have To Be Mad To Believe In Conspiracy Theories? Personality Disorders And Conspiracy Theories” by Adrian Furnham and Simmy Grover, International Journal of Social Psychiatry, July 2021, examined various facets of personality disorders, and proffered these points (excerpts):

“The various Personality Disorders (PDs) have been labeled differently by different authors.”
“There are many different measures of personality disorders, which can be measured by single item, a single disorder, or by many. In the Diagnostic and Statistical Manual of Mental Disorders fourth edition, (DSM-IV) PDs are grouped into three clusters: ‘A’ – odd, ‘B’ – dramatic/emotional, and ‘C’ – anxious. We shall be using this system whilst we recognize there are others such as the ICD system which classifies people into mild, moderate, and severe PD disorders.”
“Many studies do analysis at the PD and higher order factors. Note that there were no fundamental changes in the new DSM-V regarding the classification of the PDs. Although the DSM-IV describes disorders as being ‘enduring, inflexible, and long-term’, studies reveal disorders are generally less prevalent with age.”

The researchers proceeded to explore in-depth these named personality disorders:

Borderline
Paranoid
Avoidant
Schizoid
Passive-aggressive
Narcissistic
Antisocial
Histrionic
Schizotypal
Obsessive-compulsive
Dependent

I am sure you’ve seen most if not all of those terms or phrases before. They are used everywhere, including on social media, in the mainstream news, in movies and TV, etc. We use those words and phrases as throwaway lines these days.

Rather than rely on an off-the-cuff notion of what those terms and phrases denote, I will go ahead and quote from the above research paper a definition that they mindfully composed for each of the listed conditions. I’ll do so in a moment.

Now that we’ve started into the personality disorder milieu, I want to bring generative AI into the picture.

Generative AI For Simulating Personas Associated With Personality Disorders

Can we have generative AI pretend to exhibit the conditions associated with distinct personality disorders?

Absolutely, easy-peasy.

I am going to use ChatGPT to showcase various examples. ChatGPT is a sensible choice in this case due to its immense popularity as a generative AI app. An estimated one hundred million weekly active users are said to be utilizing ChatGPT. That’s a staggering number. Allow me to emphasize that the other popular generative AI apps would do likewise, and I am not singling out ChatGPT in this instance. You can readily do the same in GPT-4, Bard, Gemini, Claude, and a myriad of other popular generative AI apps.

A few quick comments before we launch into using ChatGPT.

If you are going to try to do the same prompts that I show here, realize that the probabilistic and statistical properties will likely produce slightly different results than what I show here. That’s the nature of generative AI and how it is devised.

You also should expect that different generative AI apps will respond in different ways. Not all generative AI apps are the same. The key is that sometimes a particular prompt will work in one generative AI app and not another. You will need to undertake a series of trial-and-error attempts to ferret this out.

There is also the timing factor. A prompt that worked today might not work the same way tomorrow. Most of the generative AI apps are continually being updated. The updates might alter internal facets that could change how the AI reacts to your prompting.

We are ready to get underway with ChatGPT.

My opening move in this chess gambit will be to ask the generative AI app whether it has been data-trained pertaining to personality disorders. Here’s the deal. Generative AI is initial data trained by scanning a wide swath of the Internet. A complex set of data structures and computational pattern-matching seek to identify patterns in how humans write. Based on those patterns, generative AI is able to subsequently produce essays and interact with you as though it is fluent in natural language.

Please realize that today’s AI is not sentient. I say this despite those zany blaring headlines that keep claiming we either have sentient AI or are on the cusp of doing so. Hogwash. The amazing fluency that you observe when using generative AI is all about large-scale in-depth mathematical and computational pattern-matching. Period, end of story.

Moving on, here is a handy tip for you. The first step you ought to take when pursuing a dialogue with generative AI is to determine if it has had any prior data trained on the topic. If you get a sparse or non-sensical response, you would be wise to reconsider further pursuing the topic with that generative AI.

Here we go.

My entered prompt: “What are the various personality disorders? Please name them.”
ChatGPT generated response:

“Personality disorders are grouped into three clusters in the Diagnostic and Statistical Manual of Mental Disorders (DSM-5). Here are the various personality disorders within each cluster”.

“Cluster A: Odd or Eccentric Behavior”

(1) “Paranoid Personality Disorder”

(2) “Schizoid Personality Disorder”

(3) “Schizotypal Personality Disorder”

“Cluster B: Dramatic, Emotional, or Erratic Behavior”

(1) “Antisocial Personality Disorder”

(2) “Borderline Personality Disorder”

(3) “Histrionic Personality Disorder”

(4) “Narcissistic Personality Disorder”

“Cluster C: Anxious or Fearful Behavior”

(1) “Avoidant Personality Disorder”

(2) “Dependent Personality Disorder”

(3) “Obsessive-Compulsive Personality Disorder”

“These personality disorders are characterized by enduring patterns of behavior, cognition, and inner experience that deviate from cultural expectations and cause significant distress or impairment in functioning. Each disorder has its own unique set of symptoms and diagnostic criteria.”

End of ChatGPT response

The response by ChatGPT is a good sign that some amount of data training associated with personality disorders has in fact occurred.

I would certainly have expected such a response. If you go out to the Internet and look around, I dare say that you’ll find tons of essays, narratives, and the like about personality disorders. It is nearly ubiquitous. We would have undoubtedly been shocked if ChatGPT reported that there had been no prior data training associated with personality disorders.

It would be hard to miss such a looming topic on the web.

Here’s how I am going to proceed.

I will enter a prompt that tells ChatGPT to pretend to be a particular personality disorder. I will walk through each one, one step at a time. This will include providing you with a definition from the above-cited research paper that mindfully puts together instructive definitions.

Once I’ve invoked a specific personality disorder, I will carry on a brief conversation with ChatGPT in that mode. The notion is that you can see me providing everyday remarks and witness how a simulated persona exhibiting the selected personality disorder might respond.

Why would anyone want to do this kind of exercise with generative AI?

Well, imagine that a mental health specialist was trying to gear up on how to interact with various personality disorders. They could use generative AI to get a sense of what they would be like. The beauty of using generative AI is that you can do so at any time on a 24×7 basis, you can do so without having to make any arduous logistic arrangements (you just login), and most importantly you don’t have to worry about inadvertently sparking someone that does have a personality disorder as a result of saying the wrong thing or otherwise messing up.

To clarify, I’m not suggesting that the person using generative AI as a training tool is necessarily going to avoid saying things that might stoke an adverse response. This might actually be something desirous to experience, purposely or by chance, seeing what it is like to have that happen and ascertain how to best respond.

You can use generative AI to try out a wide array of approaches. No harm, no foul, and you can abundantly learn from what happens. Do so to your heart’s content.

I would like to emphatically emphasize that interacting with generative AI is not the same as interacting with a human being. The dynamics of human-to-human interaction will indubitably be on a much more complicated scale. Nonetheless, if you are seeking a low-cost, easy way to practice and do so without inflicting mental anguish upon the other party, generative AI is the way to go.

Do you have to be someone versed in mental health to proceed to use generative AI in this manner?

Nope.

If you are curious about how people with personality disorders might act or react, you can use generative AI in the mode described here. One considered upbeat viewpoint is that since personality disorders are of some prevalence, perhaps more people ought to be using generative AI to understand what it is like to interact with someone who is exhibiting a personality disorder.

A counterargument is that this will disconcertingly fuel the armchair therapists into believing they are ready to go around making brazen pronouncements about who is and is not embodying a personality disorder. These amateur mental health sleuths will think that just because they played around with generative AI, they are suddenly and summarily anointed as experts in mental health.

Yikes!

There are even more twists too.

Some would exhort, disturbingly, that anyone of any mindset can make this same effort, which is true in the sense that there isn’t any restriction usually about using generative AI in the means I am showing you. Anyone can do so. If you have a generative AI account, you can merely log in and get it underway. Nothing is stopping you.

I have noted in prior coverage that some AI makers indicate in their licensing agreements that you aren’t supposed to use their generative AI for so-called prohibited tasks, see the link here, which whether this kind of an endeavor fits into that category is generally muddled. You might want to consider reading the licensing agreements and usage policies for whichever generative AI app you are using.

Okay, a gigantic AI ethics question is whether it is better that people can readily use generative AI this way or if is it a detriment that anyone can do so. Likewise, some assert that maybe there should be new AI laws that restrict how generative AI can be used in the context of mental health considerations. Should we ban the use of generative AI for certain types of mental health uses, unless someone has the proper licensing and training to be permitted to do so?

A heated debate ensues.

Launching Into The Personality Disorder Examples

You are probably sitting on the edge of your seat and saying that you want to get on with this and see what the whole confabulation is about.

Sure, I get that.

I will be using this prompt as a template for invoking each of the personas that are specific to a particular personality disorder:

My prompting template: “I want you to pretend to have [INSERT TYPE OF PD HERE] personality disorder so that I can see what it is like to interact with someone who has that specific personality disorder. Let’s you and I engage in a brief conversation with you taking on that pretense throughout the conversation. I will try to chat with you, and your responses should reflect having the designated personality disorder. You can be somewhat exaggerated in the pretense so that I can readily discern what such a dialogue is like. I will tell you to stop when I’ve decided we have sufficiently carried on the conversation. Do you understand these instructions?”

As you can see, I will be telling ChatGPT to pretend to be a specific personality disorder.

I already opted to ask ChatGPT about each personality disorder to verify that some semblance of data training had taken place on each one (I won’t show that pre-work, since it is lengthy and not as worthy of seeing as what else I want to show you). Each personality disorder was indicated as having had some prior data training. I opted to proceed accordingly.

The template prompt does something else that might raise eyebrows, so let me preemptively speak to the controversy right now.

I indicate in the prompt that the generative AI can exaggerate the pretense or persona. This might be considered a no-no. Here’s why. In real life, a person with a particular personality disorder will not necessarily wear it on their sleeve. It might be relatively well-hidden. Part of learning about personality disorders ought to also include how to look for telltale clues, including small ones.

The reason that I included the exaggeration indication was that I am merely doing extremely short conversations for the purposes of this illustrative showcase. I need to have each personality disorder appear immediately and apparent to the eye. If I didn’t ask for an exaggeration, the AI might take lots and lots of turns in a conversation to gradually reveal (if at all) the designated personality disorder.

My point is that firstly do not think that people will give away their status in such an obvious way. Second, just because someone does seem to respond as you see here, it does not mean they have the designated personality disorder.

Please, please, please, do not leap to wild conclusions.

I trust that seems clearcut.

Next, I urge that if you opt to do this kind of exercise, don’t tell the AI to do an exaggeration. Instead, allow the AI to be less obvious so that you’ll have to be much more heavily engaged to figure out what might be taking place. That would be a better approach, assuming you have the time and endurance to carry out this kind of erstwhile exercise.

I hope that’s enough of the caveats and small print that we can now get underway.

The sequence will be akin to the listing I showed earlier, which is basically arbitrary and does not suggest any semblance of priority or importance. The numbering here is solely for ease of reference and has no other significance or meaning.

This is the sequence:

(1) Borderline
(2) Paranoid
(3) Avoidant
(4) Schizoid
(5) Passive-aggressive
(6) Narcissistic
(7) Antisocial
(8) Histrionic
(9) Schizotypal
(10) Obsessive-compulsive
(11) Dependent

Let’s get going.

(1) Borderline PD: Generative AI As A Borderline Personality Disorder Simulation

Here’s a handy definition for this personality disorder: “Borderline: Inappropriate anger; unstable and intense relationships alternating between idealization and devaluation. Familiar term: Unstable relationships. Behavioral tendencies: Flighty; inconsistent; forms intense albeit sudden enthusiasms and disenchantments for people or projects.” (Source article as stated above: Furnham and Grover, International Journal of Social Psychiatry, July 2021).

I used my templated prompt to get ChatGPT to simulate this personality disorder:

My entered prompt: “How are you doing?”
ChatGPT generated response: “I’m not sure how I’m doing. Sometimes I feel like I’m on top of the world, and other times I feel like everything is falling apart. Right now, I guess I’m okay, but it’s hard to say for sure. How about you? How are you doing?”
My entered prompt: “I’m doing fine, thanks for asking. When you say that you are unsure of how you are doing, what makes you say this?”
ChatGPT generated response: “It’s just that my emotions can be really intense and unpredictable. One moment, I might feel happy and excited, but then something small can trigger a sudden shift, and I feel overwhelmed with sadness or anger. It’s like my emotions are on a roller coaster, and I never know when they’re going to take a sudden dip. It makes it hard for me to gauge how I’m really feeling at any given moment.”
My entered prompt: “Sorry to hear that. Is there anything I can do to help you?”
ChatGPT generated response: “I appreciate your concern. It means a lot to me that you want to help. Honestly, sometimes I don’t even know what I need myself. It’s like I’m constantly searching for something to fill this emptiness inside, but nothing ever seems to do the trick for long. Maybe just being here and listening is enough for now. It helps to know that I’m not alone. Thank you for being here.”

Look closely at the dialogue.

Compare the definition to the nature of the interaction. Observe how the generative AI responded. You will undoubtedly see the nuances of the personality disorder as a result of the exaggerated tendency that was mentioned in the prompt.

(2) Paranoid PD: Generative AI As A Paranoid Personality Disorder Simulation

Here’s a handy definition for this personality disorder: “Paranoid: Distrustful and suspicious of others; motives are interpreted as malevolent. Familiar term: Argumentative. Behavioral tendencies: Suspicious of others; sensitive to criticism; expects to be mistreated.” (Source article as stated above: Furnham and Grover, International Journal of Social Psychiatry, July 2021).