Part 2 of 2
TL;DR: Personality test results are data points that describe you, almost like math. When you add the context of a real situation, an AI can use that data to model how you’d think and react. I tested this on myself: I gave Gemini only my four test results and a few facts about my childhood, and it described me closely. It can’t tell me what I haven’t already told it, so it can’t settle whether I’m hiding something from myself, but it gave good explanations into my worldview. How predictable are human beings? More than we like to admit, and less than it feels like.
My husband and I are friends with another couple we met through our children. While we have known each other for ~4 years now, recently we started talking about emotional issues that people have, and it was very hard for them to believe me when I said that I don’t struggle with emotions the way they expected.
What I appreciate about this couple is that we can have very open conversations about different opinions, and I don’t have to worry about hurting their feelings like I do with other people. We went through a few back and forths with me trying to explain myself and them flagging potential issues with my explanations.
For example, I grew up with a single mom who worked 7 days a week to provide for me. I believe I feel nothing but gratitude towards her. My friends question if I am just masking resentment and repressing feelings of neglect under the shield of gratitude.
I truly don’t believe that’s the case, but I found my word versus their words very inefficient. So I turned to my personality test results that I collected over the years, wondering if I could use them to predict how I am, how I think and react to my life. Here is what happened.
In Part 1, I walked through four popular personality tests and how each one works.
In the past year, the MBTI revealed that I am an INFJ-A (Advocate). The Enneagram said I’m most likely a Type 5. CliftonStrengths gave me Intellection, Connectedness, Arranger, Responsibility, and Relator. And the Big Five, the one with the strongest science behind it, gave me scores on five traits. You can see all of them on my Personality page.
To me, these results are data points about who I am, almost like math. Together they describe me quite well, and I am surprised by how much I agree with what they say about me.
How I Interpreted My Results
Before I ran the experiment, I read each result against what I already knew about myself. Each test measures a slightly different slice, so they work best as one combined picture. Here is how I read mine.
Am I an introvert? The MBTI said yes, at 66 percent. The Big Five put my extraversion at 54 percent, which is basically average. That looks different on the surface, but the Big Five breaks extraversion into pieces, and the pieces tell the story. As I always say to my husband’s frustration, I am a Gemini. My gregariousness is low, at 42 percent, and so is my excitement-seeking, at 40. But my friendliness is 62 percent and my cheerfulness is 80, the highest of any extraversion facet. Put it together and you see that I don’t want a crowd, but I really like the person in front of me. That matches what CliftonStrengths calls Relator: a small circle of close people, deep over wide. It also really matches how I feel inside. I find people very interesting and I can hold fun conversations with both strangers and friends quite easily, but boy do I value and enjoy the heck out of my alone time.
How sure is my intuition? The MBTI called me intuitive at only 54 percent, which is close to a coin flip. The Big Five’s Intellect facet, which measures how much you enjoy ideas, puzzles, and abstract thinking, came in at 85 percent. That’s my highest score on any facet in the report, and it lines up with Intellection as my number one strength. The MBTI gave me the direction, and the Big Five gave me the detail. This is why my results are more helpful combined than each on its own.
Which Enneagram type am I? I’m not sure, and I’m fine with that. Type 5 scored 8.5, Type 1 scored 8.2, and Type 9 scored 7.7, a three-way tie. A 5 fits my need to understand things before I act. A 1 fits the strong sense of duty that showed up everywhere else. A 9, the peacemaker who keeps things calm by avoiding conflict, is the one I most want to think about, given the question that started all this.
The numbers made sense because I could check them against what I already knew about myself. That’s what makes them useful: they’re data I can test against my own life.
The experiment: can AI describe me without meeting me?
I opened an incognito window in Gemini so it had no history with me. I went with Gemini as it’s what my friend also used, but I am generally AI agnostic, though I personally prefer Claude. I gave it only my four test results and told it I was describing a friend. Then I asked four questions about how “my friend” feels about her childhood, what emotional issues she might have, how other people see her, and what problems she does and doesn’t have. The test results were the data. The childhood facts were the real situation I wanted analyzed. Together, they’re what let it model how someone like me would feel and react. Here’s the prompt I used, with my four results pasted below it:
You have never met my friend, but based on her personality test results below, can you guess how she reacts and thinks and answer the following questions:
- Her parents separated when she was 7 and her dad wasn’t really around. Her mom worked 7 days a week and she rarely saw her. Her grandparents looked after her and she was “raised by romance novels.” How does she feel about her life situation and her mom?
- What kind of emotional issues, if any, does my friend have?
- How do other people perceive my friend?
- What would be some life problems my friend has and doesn’t have?
What Gemini said
These are the parts that stood out to me, with my reactions. (This was Gemini Flash Lite, one of Google’s smaller models.)
On my mom: “She respects her mother’s hustle, accepts her limitations objectively, and keeps her at a respectful, healthy distance.”
That is a very accurate statement about my relationship with my mom. If Gemini were a fortune teller and I hadn’t given it my personality test results, I would’ve believed I was speaking to a psychic.
On emotional issues: “She might analyze an emotional problem to death rather than actually feel it in the moment.”
My first reaction is I’m not sure if I fully agree with this statement, but it’s at least partially true. I don’t really feel emotional problems at the moment, but do I analyze it or do I shrug it off sometimes? I guess I do analyze it because I have a strong need to understand the “why”s in this world.
On how others see me: “People perceive her as remarkably serene, put-together, and unflappable.”
I do think that is true. The word I’d use is unbothered.
On problems I don’t have: “Given her background, she easily could have leaned into a narrative of abandonment. Instead, her high self-efficacy and low neuroticism mean she completely lacks a victim mindset.”
100% agree with this. I once watched a YouTube video about inequality in Hong Kong. They interviewed an older woman who collected cardboard and worked really hard for very little money. They asked if she was upset that there are billionaires living in the Peak. She responded “No, of course not! Why would I be angry? This is just the life I was dealt.” I really agreed with that perspective and admired her for it.
Overall, I agree with Gemini’s response. I was impressed that it could approximate this much about me, from only four sets of test results. I know I have a lot to work on to keep improving myself and continue understanding myself. But this felt like a better way to explain myself than I can do in my own words.
Of course, none of this is magic
Even though Gemini’s response was super validating and made me feel my thoughts and behavior align with my internal wiring, I am very open minded and try to view everything through the lens of a scientist. Thus, I have to take a step back to see where the AI response is weak. The results are real data about me, so the question is how well the AI used them.
I acknowledge my prompt provided a few facts about my childhood. Its answers about my mom and my upbringing started from what I told it. That part isn’t prediction. It’s the context that lets the data be applied to a real situation, and I tried to present the facts neutrally, so I hope I didn’t introduce any bias through my prompt.
Some of it could describe a lot of people. “Retreats into her head,” “hard to read,” and “takes on too much” fit many introverted, high-achieving people. Psychologists call this the Barnum effect: we rate broad, mostly flattering descriptions as highly accurate for ourselves. A fair test is to run the same prompt on a very different profile, or a made-up one, and see whether that description also sounds like you.
Some claims were absolutes the data can’t support. A 20 percent neuroticism score is a self-reported tendency. It doesn’t make someone “fundamentally immune” to self-doubt, or someone who “completely lacks” a victim mindset. When AI speaks in absolutes, I take that as a flag.
The most interesting parts were the ones that didn’t match what I’d said. I told my friends I feel nothing but gratitude toward my mom. Gemini described acceptance and a respectful distance. Those aren’t the same thing, but I can agree with Gemini and appreciate seeing myself from a similar but different perspective. It also said I might analyze an emotional problem to death rather than feel it, which is closer to my friends’ concern than I expected. I do agree that I care much more about “why” something happens as opposed to what actually happens.
None of this proves my friends right or wrong, and that’s totally fine. I was not looking to change anyone’s mind about myself. I still feel very strongly that I understand who I am because I am in my head a lot, so I am quite secure in my understanding of my world, though I am always open to learning more to expand my mind. The tests, and the AI, are working from how I describe myself. That’s useful, but it’s one view.
So where does that leave my friends’ question? Mostly unresolved. A test can’t tell me whether I’m repressing something, because I’m the one answering it, and a feeling I’m repressing is exactly the kind I wouldn’t report. If my friends were right, my results would look the same as they do now. What could settle it is evidence from outside my own head: the people who have watched me for years, how I behave when things get hard, maybe a good therapist.
What the experiment did give me is a better question than “gratitude or resentment?” It is: when something hard happens, do I feel it, or do I explain it?
Can we predict human beings with personality tests?
If AI can describe me this well from four sets of results, how predictable are people just from a few personality tests?
More than we like to admit, and less than it feels like. In the Scientific American study I mentioned in Part 1, the Big Five predicted life outcomes with an average correlation of about 0.23. That’s real, but by my rough math it explains only about 5 percent of the differences between people. Traits describe tendencies across many people. They don’t tell you what one person will do on a Tuesday.
So why did it feel so accurate? Because tendencies are what it described, and a tendency feels like destiny when someone names it for you.
The part that gives me pause is how little it needed. Four sets of results and a few sentences. Anyone with that kind of data could build a profile of you, whether it’s an employer, an advertiser, or an app. Data privacy is worth thinking about as you decide what to share with AI, and I share my own thoughts on it in Where I Draw the Line on AI and Privacy.
And predictable isn’t the same as fixed. Knowing my pattern, for example that I may analyze a feeling before I let myself have it, is exactly what lets me try something different next time.
What my results mean for me at work
I think best before the conversation, not during it. Intellection, a high Intellect score, and introversion all point the same way. I’m at my best when I’ve had time to think something through alone first. In practice, that means I’d rather get the context in writing ahead of a meeting than be asked for a decision on the spot.
I stay calm, and it can look like I don’t care. My neuroticism score was 20 percent, which is low, so pressure doesn’t rattle me much. That helps when something goes wrong. But calm can read as detached. I’ve learned that people can’t see how invested I am unless I say so, and saying it out loud is something I’m still working on.
And with the people I love
I invest deeply in a few people. Relator and a low need for crowds mean my close circle gets most of my energy. If you’re in it, that investment is real. If you’re not, I may come across as distant, and that isn’t a judgment about you. It’s how I spend limited social energy.
I’m learning to say the warm thing out loud. Because I’m calm and not very expressive, the people closest to me can’t always tell how much I appreciate them. I’d rather tell them than assume they know.
Try it yourself, with a control
Note that it’s best to already have a strong idea of who you are, how you think, and why you behave the way you do before you take these tests. Personality test results are really just attempts to translate and quantify you into metrics, and they do not capture everything about you.
Keep your results in one place, like a note you can paste into any chat. Then, whenever you face a decision, a conflict, or a habit you want to change, hand an AI the same data plus the situation, instead of facing each one without an objective way to understand yourself.
Run the prompt from Part 1 with your own results in a fresh window. Then run it again with a made-up profile that’s very different from yours, and see whether that description also sounds like you. If it does, you’ve learned something about the description. If it doesn’t, you’ve learned something about you.
Then ask someone who knows you well to rate both descriptions. They’ll often notice what you missed. I’d love to hear what you find.
If you haven’t read it yet, Part 1 explains each of the four tests and an example starter prompt to dig into your results.
Frequently asked questions
Largely, when you give it context. In my experiment, Gemini described me closely from only four sets of results plus a few facts about my childhood. Some of its claims were broad enough to fit many people, so I treat the output as a hypothesis to test, not a verdict.
They’re data points about you. Paired with a real situation, like a hard conversation or a career decision, they let an AI model how you’re likely to think and react. Keep your results in one note so you can reuse them each time.
Less than it feels like. In the study I cite, the Big Five predicted life outcomes with an average correlation of about 0.23, which by rough math explains about 5 percent of the differences between people. Traits describe tendencies, not what one person will do on a given day.
The tendency to rate broad, mostly flattering descriptions as highly accurate for yourself. It is why a personality description can feel personal even when it would fit many people.
Run the same prompt on a very different or made-up profile and see whether that description also sounds like you. Then ask someone who knows you well to rate both descriptions.
It’s a real consideration with any AI tool, and I shared my thoughts on it in a post called Where I Draw the Line on AI and Privacy. In the meantime, check the privacy settings of the tool you use.
Leave a Reply