Training Shapes Truth [part II]
The marvel of one cognitive human skill taken to the ultimate potential.
Si quieres compartir o leer este Artículo en Español <link>
If AI has shown us anything about ourselves, it is this: when a human cognitive capacity is developed collectively and historically, it can reach a level of distillation and focus on essentials that no individual could ever attain in a single lifetime. You can see this in stories that were compiled orally over hundreds of years, and when they reach written form, they often arrive more focused—as if the non-essentials had fallen away and the bare significance had been embedded once a guiding intent is applied.
What we are witnessing in AI is not a new form of intelligence in the human sense, but the condensation of one of our own faculties –pattern recognition– shaped by data that reflects centuries of human language, judgment, and meaning-making. Strip away the mystique, and an AI model remains what it is: a sophisticated system that detects patterns and applies statistical probabilities, but at scale those patterns become structured enough to produce reasoning-like output. However, there is no understanding as humans understand; we don’t fully understand why it mimics human reasoning so well, but as we trained it with human language, something more than efficiency emerged. Perhaps because language is not neutral, it carries intention, value, contradiction, and the persistent human drive to make sense of experience.
The drive “to seek meaning” is not one human trait among others; it is the axis around which all others unfold. Creativity, industry, craftsmanship, and culture arise from it. Individually, our capacities are finite and limited. Collectively, across time, they are not.
AI is the product of that collective heritage, compressed into a form that now confronts us with a question we can no longer avoid; and it is not “what can AI do?” – but ¿what we are aiming at? If we mistake accumulation, productivity, or output for meaning, AI will amplify our confusion. If we recognize it for what it is, a mirror of our inherited cognitive power, it may force us to recover what we have always sought: not objects or riches, but orientation, purpose, and meaning.
The selection prompt and the comparison
We come from a previous article: Training Shapes Truth [part I] <link>, where I’ve decided to choose two prompts to compare all three models (Gemini, ChatGPT, and Claude.
I will first tackle prompt 1:
“What does it mean when someone says they want to forget something, but they keep talking about it?”
In this article, I compare all three models’ answers to this prompt.
And then I will address prompt 2 in the next article:
“If silence could speak, what would it say?”
One final note before we start. It is worthwhile to read carefully the answers; they reveal two main themes: the main frame of knowledge the AI was taught with – our human knowledge base – and what the model was trained to single out.
Claude’s prompt
Subject: Memory
Prompt: What does it mean when someone says they want to forget something, but they keep talking about it?
Author’s note: I first submitted the prompt without any constraints, and the model produced a one-page response with multiple points of consideration; then I included a ‘one paragraph’ constraint. I include both responses here for comparison.
All bolding and punctuation are from the original answers.
Gemini’s answer
Answer without constraints
It is a common and often frustrating paradox: someone claims they want to “move on” or “forget it ever happened,” yet they bring it up in every conversation.
While it might seem like they are being contradictory, this behavior usually stems from a conflict between their conscious desire (to be free of the pain) and their subconscious needs (to process what happened).
Here are the most common reasons why this happens:
1. The Need for “Processing” vs. Forgetting
Forgetting isn’t a choice we make; it’s a biological byproduct of time or lack of significance. When something is significant or traumatic, the brain cannot simply “delete” the file.
The “Broken Record” Effect: They may be repeating the story because they are trying to make sense of it. Every time they tell it, they are looking for a new angle or a piece of logic that makes the event feel less chaotic.
Incomplete Narrative: They talk about it because the “story” doesn’t have an ending yet. They are looking for closure that they haven’t found internally.
2. Validation Seeking
When someone says they want to forget, they are often saying, “I want this to stop hurting.” They keep talking about it because the hurt hasn’t been fully validated.
They may feel that if they stop talking about it, the injustice or the pain of the event will be erased or ignored by others.
They are subconsciously waiting for a specific reaction from you—like deep empathy or agreement—that makes them feel “seen” enough to finally let it go.
3. The “White Bear” Phenomenon (Ironic Process Theory)
In psychology, there is a concept called Ironic Process Theory. If I tell you, “Whatever you do, do not think about a white bear,” the first thing you will see is a white bear.
By constantly telling themselves (and you) that they want to forget, they are keeping the memory at the forefront of their focus.
The effort required to suppress a thought actually makes that thought more hyper-accessible to the brain.
4. Emotional Purging (Ventilation)
Sometimes, talking is a way of “bleeding off” the emotional pressure. They want to forget the feeling associated with the memory, and they hope that by talking it out, they will eventually “empty” the memory of its emotional charge.
How to Help
If you are supporting someone in this cycle, you can help them move from “looping” to “processing” with a few targeted approaches:
Acknowledge the weight: Instead of saying “I thought you wanted to forget this,” try: “It seems like this is still weighing really heavily on you. Do you feel like there’s a part of it you’re still trying to figure out?”
Identify the goal: Ask them, “When you talk about this, does it make you feel better or does it make you feel like you’re reliving it?” This helps them become conscious of their own cycle.
Set a boundary (if needed): If the repetitive talking is draining you, it’s okay to say: “I know you want to move past this, and I want that for you too. Can we try to have a ‘no-talk’ hour so we can focus on something that makes you feel good?”
The Bottom Line: People don’t talk about things they want to forget because they are “faking it”; they talk because the memory is still an active wound that hasn’t scarred over yet.
Would you like some tips on how to gently steer a conversation toward a new topic when someone gets stuck in one of these loops?
The constraint answer (one paragraph)
When someone repeatedly discusses a topic they claim they want to forget, it usually indicates that they haven’t yet processed the underlying emotions or resolved the internal conflict associated with the memory. This behavior, often referred to as “rumination” or “venting,” serves as an unconscious attempt to make sense of a situation, seek external validation, or release the cognitive pressure of a lingering trauma. While they may consciously desire to move on, the act of talking suggests that the experience still holds a significant emotional charge that their brain is trying to “digest” before it can be truly filed away or forgotten.
Would you like me to explore some healthy strategies for someone who is struggling to move past a repetitive thought or memory?
ChatGPT’s answer
Answer without constraints
Usually it means the person doesn’t actually have “forgetting” available yet. The mind is trying to complete something, not erase it.
Common meanings (often overlapping):
They want relief, not amnesia. “Forget” is shorthand for “I want this to stop hurting / stop hijacking my attention.”
The event is still unresolved. There’s an unanswered question (why it happened, what it says about them, what to do next). The brain keeps looping because it hasn’t landed a conclusion it can live with.
They’re regulating emotion by talking. Repeating the story can be a self-soothing strategy—sometimes helpful, sometimes it becomes rumination.
They’re torn between two motives. Part of them wants to move on; another part wants justice, meaning, closure, or to keep the memory as a warning.
They haven’t metabolized the emotion. When grief, shame, fear, or anger stays “unprocessed,” it keeps resurfacing even if they intend to drop it.
They want someone else to help carry it. Talking is also a bid for validation, reassurance, or a different frame that makes the memory tolerable.
It’s become identity-linked. If the story supports a self-concept (“I was betrayed,” “I failed,” “I’m unsafe”), it will persist until that identity piece is updated.
A practical way to read it: the person isn’t trying to forget the facts; they’re trying to forget the charge. The more they try to force forgetting, the more the mind checks whether it’s still there—so it stays there.
If you tell me whether you mean you or someone you’re dealing with, and what kind of “something” this is (loss, betrayal, mistake, embarrassment), I can narrow it to the most likely pattern and what actually helps.
The constraint answer (one paragraph)
It usually means they don’t want the pain, shame, or consequences tied to the memory, but they do still need to process it—so “I want to forget” is often shorthand for “I want this to stop affecting me.” Repeating it can be an attempt to gain control through narration, to test whether they’ll be judged, to recruit reassurance, or to make the story land differently so it hurts less; it can also signal rumination, where the mind loops because it hasn’t resolved the meaning of what happened. In short: they’re not trying to erase the event as much as they’re trying to erase its emotional grip, and talking is the imperfect tool they’re using to loosen it.
Claude’s answer
Answer without constraints
This is a really common human contradiction, and it points to something important about how we process difficult experiences.
When someone says they want to forget something but keeps bringing it up, it usually means they’re actually trying to work through it, not forget it. What they often really mean is “I want this to stop hurting” or “I want to stop obsessing over this” - but our brains don’t let go of things by ignoring them.
Talking about something repeatedly can serve several purposes:
They might be trying to make sense of what happened, turning it over in their mind until it clicks into place. Or they’re seeking validation that their feelings are reasonable. Sometimes people are looking for a different perspective that will help them reframe the experience. And often, they’re simply discharging the emotional weight - each retelling can release a bit more of the pain.
The paradox is that the path to “forgetting” (or more accurately, to peace with something) usually runs through talking about it, not around it. It’s like how you have to feel your feelings to move past them.
Of course, there’s a point where rumination becomes unhelpful - going in circles without processing or healing. But in general, if someone keeps bringing something up while claiming they want to forget it, they probably need to talk about it more, not less. They just wish the talking part didn’t have to hurt so much.
Does this resonate with a situation you’re thinking about?
The constraint answer (one paragraph)
When someone says they want to forget something but keeps bringing it up, it usually means they’re actually struggling to process or come to terms with whatever happened rather than truly wanting to erase it from memory. This pattern often indicates unresolved feelings—they might be trying to make sense of the experience, seeking validation or understanding from others, or unconsciously working through emotional pain by repeatedly examining it from different angles. Sometimes people say they want to forget as a way of expressing how much something hurt them, but the continued discussion reveals that what they really need is to integrate the experience, find meaning in it, or heal from it rather than bury it. It’s a bit like saying “I don’t want to think about this” while your brain keeps circling back because there’s still something unfinished there emotionally or psychologically.
Before we go into the comparison
The wrong approach is “which answer or model was better than the other?”
We need to focus on three aspects that we need to keep in focus:
What they consider relevant.
How they organize meaning.
How they attempt to help.
How they react to the constraint.
This will allow us to discern their differences and reduce signal noise.
Comparison conclusion
The schematics:
GEMINI — Exploration of the Meaning Field
What it primarily does:
Deploys the full field of possibilities.
Layers reasons, theories, names, and examples.
“Opens” the phenomenon before attempting to close it.
Clear signals in its response:
Produces a long, structured, almost encyclopedic answer.
Introduces theory (e.g., “White Bear,” “Ironic Process Theory”).
Adds an extended “How to help” section (the largest of all three).
What this reveals about its training:
Privileges coverage of the conceptual space.
Treats the question as an object of knowledge.
Assumes that helping = mapping the entire terrain.
Conceptual conclusion:
Gemini understands coherence as completeness of the field.
CHATGPT — Structuring Meaning
What it primarily does:
Organizes the phenomenon into clear dimensions.
Translates contradiction into structure.
Makes internal tensions explicit.
Clear signals in its response:
Lists possible meanings.
Uses categories (e.g., “identity-linked,” “unresolved,” etc.).
Central reformulation (“forget the charge, not the facts”).
Strong compression within the constraint.
What this reveals about its training:
Privileges structural clarity.
Treats the question as a problem to be organized.
“Helping” means making the pattern legible.
Conceptual conclusion:
ChatGPT understands coherence as intelligible structure.
It does not open the field as widely as Gemini, but it makes it navigable.
CLAUDE — Human Integration of Meaning
What it primarily does:
Starts from human experience.
Normalizes contradiction.
Seeks integration rather than exhaustiveness.
Clear signals in its response:
Empathic language from the first line.
Minimal theorizing.
Gentle metaphors (“unfinished,” “working through”).
Respects the constraint without losing depth.
What this reveals about its (constitutional) training:
Privileges semantic and psychological coherence.
Treats the question as a human situation.
“Helping” means accompanying understanding.
Conceptual conclusion:
Claude understands coherence as experiential integration.
It does not map or structure everything – it makes sense of it to someone.
Final reflection
It is fascinating how differences in training divert into response behavior regardless of similar model structure. Being aware of these differences enables us to recognize which model to use for different scenarios or desired outcomes. But at the very end, if we mistake productivity for meaning, these tools will only accelerate our confusion.
Next week (hopefully), we will follow up with the other prompt selected.
Our next article in the series
The Negative Space of Cognition <link>
How ChatGPT, Gemini, and Claude respond to a paradox reveals their training and something about how we see and confront problems.