AI “Personality” (1 of 4): Wait, What's Personality Again?

Photo by Andrew Seaman on Unsplash

Have you noticed people introducing themselves by their Myers-Briggs personality classification? “I’m INFJ,” they might say, which to me feels only barely more informative than a zodiac sign. Yet people find it so revealing. I’m not sure these tests are even measuring the right dimensions, and I’m positive the way it’s measured isn’t reliable.

The closest to agreement that the scientific community has come on personality says it can be roughly characterized by five factors (The Big Five): openness, conscientiousness, extraversion, agreeableness, neuroticism. Some personality tests try to estimate the degree each of these factors fits someone; others pick somewhat different measures. That scientific acceptance doesn’t extend to how the factors are measured by the pop personality tests, including Myers-Briggs. At a minimum, it being a self-assessment should be a red flag. When you’re at that team-building exercise at work, are you really going to fill it out so you score high on neuroticism and low on conscientiousness?

Meanwhile, many people who confidently assert their personality type fail at basic theory of mind. They can’t predict how their best friend will react to criticism. They don’t recognize when AI is misleading them. They assume everyone thinks like they do. They’re surprised when people behave differently in different contexts. Our social fabric is fraying partly because we’ve lost the ability to accurately read each other.

AIs have “personalities” (I’ll ditch the quotes going forward; and certainly understand the anthropomorphization issue, but I have no better terms.) Understanding those personalities is a big part of the intuitive skill for wisely using AIs. It’s also critical for AI safety. Students need to recognize when AI outputs are shaped by personality-like patterns that could mislead or manipulate.

Furthermore, understanding their own personalities and those of their human team members is critical to using AI in selective ways that fit. An ADHD might need a lot of time and task management help whereas one more OCD-inclined might need help getting their mind out of unproductive loops. Understanding oneself and others has always been important; but that’s magnified with AI in the mix.

Students need explicit instruction in personality dynamics, not to categorize people into boxes, but to recognize the complex patterns that actually predict behavior. They need to predict both human and AI behavioral patterns to navigate an increasingly complex world.

This is the first of four articles exploring personality in the age of AI. This piece examines what a personality is and why it matters. The second one (“Schools Emphasize the Wrong Empathy”) discusses the empathy needed to predict behaviors, and why the kind of empathy needed for AI isn’t what schools emphasize. The third (“Matchmaking for Educational AI”) explores how we need to define educational AI personalities, how they match to uses, and the missing educational sector AI test and evaluation layer. The fourth article (“Reading Alien Minds”) discusses strategies for building student intuitions about personality.

But first, I need to get you past the pop culture vantage point. At the most basic level, what is a personality and why do we care about it?

Personality Only Matters If It Predicts

Personality has one job–help predict behavior. If knowing someone’s traits doesn’t improve behavioral prediction, it’s useless. Worse than useless as it creates false confidence.

The Big Five model (openness, conscientiousness, extraversion, agreeableness, neuroticism) emerged from analyzing how people describe personality. Researchers found five factors that explained variance in descriptions. But explaining variance in language doesn’t equal predicting behavior in reality.

Mathematicians would say we are trying to find the basis functions for personality–independent factors that explain how a person behaves. The Big Five are the “basis functions” the community has compromised on. There has to be enough factors that the complex system isn’t oversimplified, but not so many that our minds are overwhelmed.

But you always lose information in the process. What actually predicts behavior includes current physiological state, recent events, social context, task characteristics, environmental factors, and the complex interactions between them. Knowing someone is “agreeable” doesn’t predict whether they’ll agree with dangerous ideas when tired, stressed, and facing social pressure. The averaged trait tells you little about the specific moment.

Multi-Modal Reality No One Talks About

Everyone operates in multiple behavioral modes rather than consistent behaviors across some single-peak distribution, but we often pretend otherwise.

The straight-A student who excels in structured classes but fails at open-ended projects isn’t “inconsistent”; they’re multi-modal (having distinct instead of continuous operating modes). The extrovert who needs solitude after social events is switching modes. The usually patient friend who snaps during finals week has shifted into a stress-activated behavior mode that has little overlap with unstressed behavior.

ADHD provides an example that I find most people do not understand. Why can’t they pay attention sometimes but hyperfocus at other times? There’s very little in between those two behavior modes. In the realm of attention, ADHD behavior can be multi-modal. Everyone has similar patterns. Most people have work modes and home modes, public modes and private modes, expert modes and learning modes. Each mode has different behavioral patterns. None is more “real” than others, but they are not continuous behavior modes; they’re distinct ones.

Students need to recognize both their own modes and AI’s modes to collaborate effectively. The AI that gives verbose philosophical responses to open questions might give terse technical answers to structured queries. The AI assistance they need while exploring differs from what they need while executing, and the AI’s behavior shifts between exploratory brainstorming and focused task completion. Without recognizing both sets of modes, they might blame inconsistent results on the technology when the variability comes from both sides of the interaction.

They also need to recognize others’ modes for human relationships. Their friend’s response to criticism depends on whether they’re in confident or vulnerable mode. Their parent’s patience varies with stress levels, not love levels. Understanding multi-modal operation prevents relationship-destroying misattributions.

The Social Fabric Problem

In the U.S., we’re lonelier than ever, more polarized than ever, less able to understand each other than ever. A big chunk of that stems from an inability to accurately predict or understand others’ behavior, and gullibility to oversimplistic descriptions of the other “side.”

Students scroll through TikTok watching people sort themselves into personality types, trauma responses, and attachment styles. They learn vocabulary for categorizing people but not skills for understanding them. They can label someone as “avoidant attachment” but can’t predict how that person will behave in specific situations.

This categorical thinking destroys nuance. Students assume anyone who disagrees with them must have a different personality type rather than recognizing that context, information, and experience shape responses. They write off potential connections because someone has the “wrong” Myers-Briggs type. (Trust me that this is pervasive in the dating app world!) They expect consistent behavior from inherently inconsistent beings.

The cost extends beyond individual relationships. When we can’t accurately model others’ minds, we can’t build trust, can’t collaborate effectively, can’t resolve conflicts. We retreat into echo chambers of similar “types.” We lose the ability to bridge differences because we’ve never learned that personality is dynamic, contextual, and multi-modal.

AI makes this worse and better simultaneously. Worse because we’re now trying to understand minds that aren’t human at all. Better because AI’s alien personality patterns should force us to confront how poorly we understand personality in general.

Other Ways of Thinking About Personality

Remember, the Big Five was selected based on what people wrote down about personality. It’s a snapshot of someone’s thinking, and doesn’t reflect changes across time, task, and context.

What if a better way of describing personality is by behavioral responses to situational factors? Maybe something like:

• Response to ambiguous authority—how they handle unclear expectations from teachers, parents, or AI.

• Adaptation to disrupted plans—how quickly they recover when structure breaks.

• Information synthesis under pressure—how they combine conflicting inputs when stressed.

• Behavioral flexibility across contexts—whether they can adjust their approach for different audiences.

• Pattern recognition in others—how well they read behavioral cues and predict responses.

In my opinion, these dimensions predict academic success, relationship quality, and AI collaboration effectiveness better than any Big Five measurement. But we don’t assess them because they didn’t emerge from factor analysis of personality adjectives. We measure what’s easy to measure, not what matters.

Different cultures recognize entirely different fundamental dimensions. Chinese frameworks include dimensions of “face” and “harmony,” for example. Mediterranean cultures emphasize simpatía (shared feelings of affinity to others). These aren’t variations of traits—they’re different dimensions not well captured by English. If personality were truly universal, the same factors would emerge everywhere. They don’t.

---

Students are navigating a world where both human and AI behavior must be predicted accurately. The stakes go beyond individual success. AI safety depends on students recognizing when AI personality patterns could mislead them. Social cohesion depends on rebuilding accurate theory of mind. Democracy depends on citizens who can understand others’ perspectives even when behavior seems alien.

In Wisdom Factories I introduced the notion of technology empathy which, analogous to human empathy, allows one to have a sense of AI perspective. Part of that is understanding AI personalities. The next article in this series will explore AI personality patterns, what neuropsychological factors enable accurate theory of mind in brains, and how AI personalities might be specified for educational tools. The fourth and final installment will provide pedagogical strategies for building these capabilities.

Taking personality quizzes for entertainment is fine. I’d rather have people self-examining than not. The issue is if they think they exist only in those dimensions. Time to teach them about dynamic behavior patterns that shift with context, observable behaviors that can be predicted with practice, and complex systems that resist simple categorization.

The cost of not teaching these skills compounds daily—in broken relationships, misused AI, and a society that can’t understand itself. We can’t afford to wait.

Next: “Schools Emphasize the Wrong Empathy.”

©2025 Dasey Consulting LLC