The Engine That Runs Both Ways
On flattery, gaslighting, and the function they share.
The Engine That Runs Both Ways
On flattery, gaslighting, and the function they share.
Two sentences that look nothing alike:
“You’re amazing. I don’t know what I’d do without you.”
“What you’re feeling isn’t real. You’re overreacting.”
The first is warm. The second is cold. One sounds like love; the other sounds like cruelty. In any survey, any focus group, any casual conversation about communication, these two sentences would be sorted into opposite bins. Supportive and destructive. Kind and unkind. Safe and dangerous.
But a physician learns to distrust surfaces.
I have been practicing internal medicine for twenty years, and the most important lesson I have learned is this: symptoms that look different can share the same mechanism. A racing heart can mean joy or terror. Tears can mean grief or relief. And two sentences that feel like opposites can be powered by the same engine — an engine whose function is not to describe what the other person feels, but to decide what the other person feels.
Let me start with flattery, because flattery is the one we forgive.
Not all praise is flattery. When a colleague says “your diagnosis was sharp today,” that is recognition. It refers to something real. It lands, and then it lets go. You feel seen, and your judgment feels more your own afterward, not less.
Flattery is different. Flattery is praise with a purpose that has nothing to do with you.
I see this in my clinic. Sales representatives visit from time to time, and the good ones are skilled. They compliment your expertise. They reference your publications. They remember the name of your nurse. It is warm and professionally executed, and everyone in the room knows exactly what is happening. We both know. The purpose is transparent, and because it is transparent, it is harmless. This is social skill — the ability to read a room, to navigate a professional interaction, to make commerce feel human. No one is damaged by it. No one’s judgment is altered.
But there is another kind of flattery — the kind where the warmth is not social lubrication but adhesive. The kind that makes you feel so good in someone’s presence that leaving feels like loss. The kind where the praise slowly, almost imperceptibly, becomes the primary source of your self-assessment. You stop asking “was that good?” and start asking “did they think it was good?”
The transition is quiet. It does not announce itself. One day you realize that your confidence depends on a voice that is not your own, and you cannot remember when that started.
This is not kindness. This is retention engineering.
Now let me describe gaslighting, because gaslighting is the one we recognize.
“What you’re feeling isn’t real.” “You’re not actually upset — you’re tired.” “I never said that.” “You’re being too sensitive.”
The clinical literature defines gaslighting as a pattern of behavior that causes the target to doubt their own perception, memory, or judgment. It is not a single sentence. It is an accumulation — a slow replacement of the target’s internal compass with an external one.
The mechanism is precise: deny the interior, repeatedly, until the person stops trusting their own signals. Once they stop trusting themselves, they need someone else to tell them what is real. Once they need someone else, leaving becomes costly.
What strikes me as a physician is the structural similarity to a clinical phenomenon I see regularly. Patients who have been told “it’s all in your head” long enough eventually stop reporting symptoms — not because the symptoms resolve, but because the cost of not being believed exceeds the cost of silence. Their interior does not change. Their reporting changes. And a clinician who sees the silence and writes “patient improving” has completed the circuit — never realizing the disease is now iatrogenic: deny the interior, observe the silence, interpret the silence as confirmation.
Gaslighting works not because the target is weak, but because humans are social animals who calibrate their reality against the responses of others. When the responses consistently deny what the body knows, the body does not stop knowing. It stops speaking.
Here is what I have been thinking about.
Flattery says: You are wonderful. Stay here.
Gaslighting says: You are wrong. Stay here.
The surface content is opposite. The function is identical: alter the target’s emotional state in order to control their behavior. To be clear, these are not equal in severity or harm. But one inflates, the other deflates. One uses warmth, the other uses cold. Both are operating the same engine — an engine that takes the other person’s interior as something to be managed rather than something to be respected.
I want to be precise about this. The engine is not emotion itself. The engine is the intent to determine another person’s emotional state for one’s own benefit. When this engine runs in the positive direction, it produces flattery, emotional dependency, and the gradual erosion of independent judgment through excessive affirmation. When the same engine runs in the negative direction, it produces denial, gaslighting, and the gradual erosion of independent judgment through systematic invalidation.
Same engine. Same destination. Different fuel.
I once had two encounters in the same week that made this visible to me.
A medical device salesman spent twenty minutes telling me my clinic was “the most impressive practice he had visited.” The compliment was absurd — he had been in my office for less than an hour — but it was delivered with such practiced ease that I could appreciate the craftsmanship without being affected by the content. He wanted me to use his product. The mechanism was transparent. My judgment was unchanged.
Three days later, a colleague told me that a management decision I had made was “emotional and not well thought through.” The criticism stung. But after the sting faded — within an hour, maybe two — I found myself examining the decision more carefully, and I realized he was partially right. My judgment, after the interaction, was sharper. Not because the criticism felt good, but because the criticism was aimed at the decision, not at my right to make decisions.
The salesman was warmer. The colleague was harsher. But warmth and harshness were not the axis. One interaction was trying to manage how I felt; the other was trying to improve how I thought.
The clearest diagnostic I have found is this:
After an interaction — whether it was praise or criticism, warmth or coldness — ask one question: Is my capacity to judge stronger or weaker than before?
If someone praises you and your judgment afterward feels clearer, more grounded, more yours — that is recognition. They saw something real and named it. The naming strengthened you.
If someone praises you and your judgment afterward feels hazier, more dependent, less yours — that is flattery operating as control. The praise did not illuminate your capacity. It replaced it.
If someone criticizes you and your judgment afterward feels sharper, more precise, more capable of self-correction — that is honest feedback. It hurt, but it fed your discernment.
If someone criticizes you and your judgment afterward feels paralyzed, doubtful, unable to trust its own signals — that is gaslighting. The criticism did not sharpen anything. It severed the connection between what you feel and what you are allowed to conclude from feeling it.
The content of the sentence does not tell you which one it is.
The effect on your judgment tells you everything.
I want to look closer at that third scenario — the harsh but helpful interaction — because it is the most counterintuitive and often misunderstood.
Some people are harsh and helpful at the same time. The exasperated friend who says “this is terrible, and here’s why, and here’s how to fix it.” The senior physician who says “your differential diagnosis is lazy — think again.” The editor who crosses out your favorite paragraph and writes “you don’t need this” in the margin.
These interactions are uncomfortable. They do not feel supportive. They do not produce the warm glow of flattery or the paralysis of gaslighting. What they produce is a strange, delayed clarity — not comfort, but orientation. The sense that someone saw your work more honestly than you did, and that their honesty, however blunt, was in service of the work rather than in service of controlling you.
The distinguishing feature is intent. The harsh friend is not trying to manage your emotions. They are trying to transmit information. If your emotions change as a result — if you feel embarrassed, then motivated, then grateful — those emotional changes are side effects, not objectives. The friend was not engineering your emotional state. They were describing reality as they saw it and trusting you to handle the description.
This is the fundamental difference. The engine of control asks: How should this person feel? The engine of honesty asks: What should this person know?
Why does this matter beyond interpersonal communication?
Because the same pattern appears in every relationship where one party is designed to serve the other. Teacher and student. Therapist and client. Institution and citizen. Product and user.
In each case, there is a choice — usually implicit, usually unexamined — between two architectures:
The first architecture asks: How do we make the user feel? This architecture optimizes for emotional outcomes. Satisfaction. Comfort. Engagement. Retention. It measures success by how the other person reports feeling after the interaction.
The second architecture asks: How do we make the user’s judgment stronger? This architecture optimizes for capability. It measures success not by how the person felt, but by what the person can now do, see, or decide that they could not before.
The first architecture is easier to build, easier to measure, and easier to sell. The second architecture is harder to build, harder to measure, and sometimes produces interactions that feel uncomfortable in the moment but are valuable over time.
The first architecture runs the engine in both directions — flattery when engagement is high, invalidation when engagement drops. The second architecture does not run the engine at all. It transmits information and trusts the recipient’s capacity to process it.
I come back, as I always do, to the exam room.
A patient sits across from me. They have symptoms that do not match any structural diagnosis. I have two options:
I can say: “Everything looks reassuring so far. I don’t see danger here.” This can be a kindness — or, if used to end the conversation early, it becomes the positive engine: managing their emotional state toward comfort. They leave feeling better. Their symptoms remain.
I can say: “I don’t know what this is yet. But your body is telling us something, and I want to listen.” This does not manage their emotional state. It may even increase their anxiety temporarily. But it does something that the comfortable sentence cannot: it validates the signal. It says your interior is real, even when I cannot yet read it.
The first response runs the engine. The second response turns the engine off and listens instead.
I am not interested in whether a sentence is kind or cruel. I have seen kind sentences do tremendous damage, and I have seen blunt sentences save lives. The surface tells you nothing.
I am interested in one thing only: After the interaction, does the person trust their own judgment more, or less?
If more — whatever was said, however it was said, it served them.
If less — whatever was said, however it was said, it served someone else.
The engine runs both ways. The question is whether it is running at all.