The Real Reason You Can’t Change Someone

Hard to Change One's Nature

Changing a person is hard because you think you are arguing with his views. You are negotiating with his history.

His history is tens of thousands of days long. Every day he experiences reward and punishment, safety and threat, being seen and being ignored, control and loss of control. Those experiences do not disappear. They become connection weights in neural circuits. They become tension, avoidance, pleasing, or aggression in the body before conscious thought. They become an automatic predictive system. What you see is his worldview, values, temperament, and insistence. What you cannot see is the thick sediment beneath them.

Changing a person often feels like changing a world. His whole system is maintaining stability according to past experience.

I

A person says, “I don’t need anyone.” This may be a value. It may be a survival strategy. A person says, “Effort is useless.” This may be a worldview. It may be a prediction verified again and again. A person says, “Everyone eventually leaves.” This sounds pessimistic. It may be the statistical conclusion of decades of relational experience.

We treat the “three views” as independent modules: worldview, values, outlook on life. From reinforcement learning and neuroscience, they look more like narrative layers on the surface of water. Beneath them sit reward functions, predictive models, action strategies, emotional conditioning, and identity structures. Values are often the verbal expression of reward functions and identity structures.

When two people’s values conflict, the disagreement is often less logical than historical. One benefited from cooperation; the other survived betrayal. One associates expressing needs with care; the other associates expressing needs with humiliation. Arguing about “whether people are trustworthy” is like arguing about the shape of an iceberg’s tip. The main body sits in deep water.

II

Reinforcement learning has a core concept: reward. Human rewards go far beyond money, food, and pleasure. Positive rewards include happiness, pride, aesthetic experience, insight, self-efficacy, identity, resonance, social recognition, safety, novelty, well-being, and a sense of existence. Negative rewards, or punishment, include anxiety, conflict, shame, danger, pain, fear, disappointment, social rejection, hatred, incompetence, despair, and insignificance.

A child has his needs interrupted every time he expresses them. He does not learn “this time no one paid attention to me.” He learns that the belief “expressing needs brings returns, attention, and love” loses one point of confidence. After this repeats hundreds or thousands of times, “not expressing” becomes the default strategy. Later, even if he meets a few people willing to listen, the past experience does not immediately lose its force. The old strategy was not built in a day. A new experience cannot cover it in one encounter.

The reverse also holds. Every time a person solves a problem others cannot solve by using an unconventional method, his belief “being unconventional and divergent can solve problems” gains a point. “I have agency, I am useful” gains a point. Over decades, every core belief and coping strategy accumulates a large amount of historical reward.

We can imagine, imprecisely but usefully, that everyone has a ledger in their mind.

“Other people can be trusted”: (-6812)
“Exercising agency can change objective things”: (+9087)
“The world is systematic, traceable, and predictable through science”: (+6273)
“Victim narrative can win sympathy and solve problems for gain”: (-4425)
“Solving the problem matters more than maintaining surface harmony”: (+5194)

These numbers are not precise calculations. They form a mental model. A person’s degree of conviction does not come from nowhere. It is the integral of historical rewards and punishments.

III

This reinforcement enters learning circuits involving the basal ganglia, striatum, and dopamine reward system. The dopamine system constantly compares “what was expected to happen” with “what actually happened,” then adjusts the tendency of the next behavior according to the difference. This is reward prediction error.

If reality is better than expected, the dopamine signal rises, and the behavior is reinforced. If reality is worse than expected, the dopamine signal falls, and the behavior is inhibited. It does not care about truth. It cares about prediction and survival. As long as a strategy repeatedly reduced pain, increased safety, or brought recognition in history, it will be written into the system. Even if it is harmful in the long run.

The prefrontal cortex participates in more abstract rules and plans. The hippocampus participates in contextual and event memory. The amygdala and related networks participate in threat and emotional salience learning. The basal ganglia constantly compress certain reactions into low-cost, automatic action patterns. Physiological imprinting means that past experience leaves a memory that can be narrated. It also changes the connection weights between neural circuits, the threshold for initiating behavior, and the body’s way of responding to certain situations.

Over time, a person does not abstractly “believe he is unlovable.” When a relationship gets close, he automatically becomes tense, tests, retreats, or controls. Even before consciousness has time to explain, the body and behavior have already made a choice. He may verbally know “this person has not hurt me.” His reward system and defense system still operate according to past experience. Reason enters language and declarative knowledge. What determines action is another set of strategies that have been repeatedly trained.

This is the deep ditch between “knowing” and “doing.” The goal-directed system led by the prefrontal cortex is slow, energy-consuming, and verbalizable. The habit system led by the basal ganglia and emotional systems is fast, automatic, and energy-saving. When stress rises, the prefrontal cortex goes offline, and old habits take over. A problem can be calmly discussed in a consulting room and then completely relapse after returning to the family of origin. This happens because the new learning has not yet crossed enough contexts or accumulated enough strength to compete with the old system.

IV

Cognitive behavioral therapy divides this structure into automatic thoughts, intermediate assumptions, and core beliefs.

Automatic thoughts are the surface layer. For example, “He didn’t reply to my message, so he doesn’t care about me.” Intermediate assumptions are more stable rules. For example, “If I don’t constantly confirm the other person’s attitude, I will be abandoned.” Core beliefs are deeper. For example, “I am unlovable,” “Others cannot be trusted,” “The world is unsafe.” Core beliefs determine how a person interprets events. Intermediate assumptions determine what strategies he uses. Automatic thoughts pop up quickly in specific situations and guide the immediate response.

A person’s beliefs form a net. A conclusion is usually not supported by a single piece of evidence. It is jointly supported by family experience, group culture, identity, morality, ideology, and material interests. A person believes “no one can be relied on.” What supports it is unstable childhood care, a career that rewards independence and toughness, and a friend circle that regards vulnerability as incompetence. This belief also protects him from facing the fear in intimate relationships.

You refute with one logical sentence: “But some people are worth relying on.” You have not touched the network that supports this belief. It is like A, B, and C jointly produce D. You temporarily press D down. A, B, and C are all still there. After some time, the system restores D.

In more extreme cases, belief dictates what counts as evidence. Information supporting it is regarded as true. Information opposing it is interpreted as coincidence, disguise, manipulation, or naivety. At this stage, belief no longer just explains the world. It manages the rules for interpreting the world.

V

What locks the system is often negative reinforcement.

Negative reinforcement means removing an aversive stimulus, thereby increasing a behavior. It is often confused with punishment. It is more hidden and more solid than positive reinforcement.

A person fears rejection, so he never actively expresses himself. Every time he does not express himself, his anxiety temporarily drops. As a result, “not expressing” is reinforced. A person fears failure, so he procrastinates for a long time. Every time he procrastinates, he temporarily escapes the pressure of having his ability tested. As a result, procrastination is reinforced. A person fears abandonment, so he repeatedly checks, interrogates, and controls his partner. These behaviors temporarily reduce uncertainty. Even if they are destroying the relationship in the long run, the system still treats them as working.

This forms a stable feedback loop. Core beliefs produce automatic thoughts. Automatic thoughts trigger emotions and defensive behavior. Defensive behavior changes the reactions of people around. The reactions of people around in turn become evidence for the core belief. A person who believes “everyone eventually leaves me” may, through constant testing and attacking, push people away. When they leave, he only says: “See, I knew all along that everyone would leave.”

The old belief actively shapes a reality that fits itself.

VI

Denial, rationalization, projection, intellectualization, and dissociation are not just “wrong thinking.” They usually carry specific regulatory functions: blocking shame, reducing fear, maintaining self-esteem, and keeping the system from overloading in the present.

A person insists, “All problems are other people’s fault.” This damages reality judgment. This attribution may also temporarily spare him from the collapse brought by “I may have failed.” You directly tear down the defense without giving him new emotional carrying capacity. The result is often stronger attack, deeper withdrawal, or the rebuilding of an even more solid defense.

Changing a person cannot rely only on tearing down walls. You also have to give him a new foundation, new rooms, new load-bearing structures. Otherwise, when the old wall falls, the person will not stand in an open field. He will panic.

VII

Trauma further increases the rigidity of this structure. Trauma does not need to happen ten thousand times. A single event of intense loss of control, inability to escape, life threat, or attachment betrayal can produce extremely strong threat learning. After that, the system raises its alert level and brings similar sounds, expressions, scenes, and relationship structures into the danger zone.

At the rational level, he may already know “this is not the past.” Heartbeat, freezing, avoidance, and attack are still triggered first. Threat learning and linguistic reasoning are not completely on the same level. Old fear memories are usually not deleted. A new path forms beside them: “This situation may be safe.” This new path needs a great deal of repetition and generalization across contexts to compete with the old path under pressure.

When stress rises, the environment changes, or the person returns to the original relationship structure, adrenaline and cortisol levels rise, and the amygdala is activated. The old path, stress responses, and defense mechanisms may still regain dominance. New safe experiences rarely work once. They are not ineffective. The dose is not enough.

VIII

Changing beliefs is also an expensive system migration.

Even if the old belief causes pain, it may continue to provide predictability, group belonging, identity stability, and short-term risk control. It once worked. The new belief may be better in the long run. Migration requires immediate payment: relearning behavior, tolerating uncertainty, admitting past judgments may have been wrong, losing the original group, even rebuilding career and relationship networks.

Many people know the old pattern is problematic. They cannot afford the switching cost from the old equilibrium to the new equilibrium. You tell him to “set up a dream.” For him, this adds a new debt that cannot be repaid. You tell him to “open his heart.” For him, this exposes him to the place his old wound fears most.

A person’s conceptual system, self-identity, values, morality, defense mechanisms, and core beliefs usually do not have clear boundaries. Over decades they have tightly coupled and grown into a self-contained, hard-to-separate system structure. If you want to know why he insists on a certain judgment, keep asking: What does this judgment protect, what does it avoid, what does it prove, and what relationships, interests, shame, and fears is it connected to?

This model-identification process is expensive. Everyone’s developmental history, reinforcement path, and internal dependencies are different. If you are not closely related to a person, it is basically impossible to change them. One correct input is a drop in the bucket. Long-term, credible, repeated new experience is necessary. Even with recency bias, a single +5 cannot cover a historical experience of -9999.

You need to spend a great deal of time understanding his reward function, sources of fear, identity structure, and defensive logic. Then, in specific situations, help him form a new “action-result” loop. This cost is often hundreds to thousands of hours. Between strangers there is neither enough trust, nor enough time, nor anyone willing to bear the cost of another person’s failed change over the long term.

IX

Some people can be awakened by a single sentence. The magic usually does not lie in the sentence. The person has already accumulated a large amount of experience that the old model cannot explain. He is already close to the critical point of phase transition. That sentence provides a new way of organizing things. It suddenly connects his previously scattered doubts, failures, and observations.

This is like supercooled water. The structure is already prepared. The outside world only needs a tiny disturbance, and it quickly condenses into ice. An instant insight is still not stable change. The new understanding must enter behavior, undergo stress testing, form a new feedback loop, and only then can it become a new disposition.

X

Change cannot rely on persuasion. It requires the entire system, over a sufficiently long period of time, to continuously obtain new predictions, new actions, and new results. The old disposition must stop being the only option that can sustain survival.

This requires several conditions.

Safe relationships. Reduce threat, bring the prefrontal cortex back online, and loosen defense.

Small behavioral experiments. Let the new strategy gain returns, rather than remaining only in reason.

Repeated practice across multiple contexts. It should not work only in one consulting room. It must generalize under stress, intimacy, work, and conflict.

Identity and group support. The new narrative needs to be caught by new relationships.

The old strategy must be costly enough. The old equilibrium must be painful enough to be worth the risk.

Time and dosage. Often hundreds to thousands of hours. One insight cannot cover decades of sediment.

Critical point. When the old model has accumulated enough anomalies, a new way of organizing may instantly condense.

“Rivers and mountains may change, but a person’s character is hard to change” makes sense. It is not fate. Human beings remain plastic throughout life. Plasticity has a cost, conditions, and dosage. Change strengthens the new path so it becomes more easily retrieved and able to come online under pressure. The old path remains, but it no longer dominates.

You can replace a person’s city, job, partner, and living environment in a single day. If his reward structure, core beliefs, automatic thoughts, defense mechanisms, and action strategies have not changed, he will likely rebuild a world in the new environment that is highly similar to the old one.

People are prediction machines shaped by historical rewards. To change a person is to let this machine, over a sufficiently long period of time, repeatedly experience a world that the old model cannot explain and in which new strategies can obtain safety and reward. The new path must become more accessible than the old one.

If you want to change someone, do not give him only reason. Give him time, safety, and repeated new results.
If you want to change yourself, do not just swear oaths. Design environments, relationships, and behavioral experiments. Let the new path be walked through again and again.
Until one day, the old disposition is no longer the only viable option.