Is Changing Your Mind After Persuasion a Failure or a Capacity for Knowing?

Geoffrey Chen

How a Person Builds a World · Season Two: “Others, Dialogue, and Shared Knowledge” · Article Seven

A person finds an article written several years earlier and realises that he no longer agrees with its judgement. He can quietly delete it. He can invent new defences for the old position. Or he can acknowledge that someone later offered better reasons and he was persuaded.

The third description feels most vulnerable. A publicly stated view becomes a trace of identity. Changing it can feel like admitting that one's earlier self was insufficiently intelligent or firm, or that another person won. In public spaces organised around the spectacle of debate, “You changed your position” is often an accusation rather than evidence of epistemic progress.

Yet if a person's view never changes after encountering better evidence and argument, “steadfastness” gives an earlier self a permanent veto over every future self. Being persuadable is not automatically admirable. Pressure, charisma and misinformation also change beliefs. The genuine capacity is to keep a view open to appropriate reasons without treating every new voice as a direction that must be followed.

Changing a view is not one kind of event

“I changed my mind” can describe at least three different transformations.

The first is factual correction. A person believed a figure, date or event, later discovered that the source was wrong, and replaced the belief. The main change concerns content.

The second is revision of scope. The central judgement remains, but its range narrows, conditions are added or previously excluded cases become visible. Many mature changes do not jump from yes to no. They turn an over-complete sentence into a more accurate set of qualifications.

The third is a change of frame. New reasons do not merely correct an answer; they alter how the question is posed. A person understood a conflict as a matter of individual character and later sees institutional incentives. They once asked only whether a technology works and later ask who bears its errors. Parts of the former answer may remain true, but they acquire a new place in a different world.

If every change is reported as a reversal, we miss common forms of learning: reweighting, reclassification and reconstruction of the problem.

Why revision can feel like a loss of self

Views are not floating sentences. They connect to earlier choices, group membership, public reputation and moral self-evaluation. Acknowledging error may also mean recognising that one supported a harmful decision or misunderstood another person for years. The cost is not merely logical.

We also tend to treat consistency across time as evidence of reliability. It is true that someone who changes with every audience is difficult to rely on. But consistency has value only when the material basis has not significantly changed. When evidence, concepts or an understanding of consequences have changed, repeating the old conclusion may be the deeper inconsistency—consistency with a former statement at the expense of consistency with reality.

The harder issue is that self-understanding is often built around a set of answers: I am a person who believes this principle. If the answer moves, the self appears to lose structure. A more resilient description is possible: I am a person who requires these kinds of reasons and who revises when evidence of this kind appears. One identity is maintained by a fixed conclusion; the other by a process of correction.

This does not eliminate conviction. It makes conviction serious without making it immovable: I commit to this position now for these reasons while acknowledging in advance that the world may require another judgement through new facts.

Humility does not make every view equally doubtful

Psychological research has recently given substantial attention to intellectual humility, commonly centred on metacognitive recognition of the limits of one's knowledge and beliefs. A broad review of the predictors and consequences of intellectual humility surveys its relationship to personal decisions, social relations, science and public judgement while noting limits in measurement and causal evidence.

Humility is not low confidence, nor is it the view that everything is only an opinion. A person can be highly confident in a well-tested conclusion while accurately representing residual uncertainty. Another person can speak modestly while never allowing evidence to alter anything.

One recent investigation of false-belief revision studied participants holding particular inaccurate beliefs about vaccines and the 2020 US election. Higher intellectual humility was associated with exploring more information contrary to those beliefs, which in turn predicted updating. Intellectual humility did not directly predict updating, however, so the researchers reported moderate rather than extravagant support. The two-study paper is valuable partly for its restraint. Epistemic character cannot be replaced by one scale or slogan. Revision also depends on the information environment, source evaluation and action.

Likewise, the “backfire effect”—the idea that correction causes people to believe a misconception more strongly—has often been repeated as a general law. Later research suggests that it is far less stable than the popular account implies. A study examining the relationship between backfire effects and measurement reliability is a reminder that even our belief that people do not revise must itself be revisable. People defend commitments, and they also update. An experimental phenomenon should not become a fatalistic theory of human nature.

How to keep change from becoming an untraceable swing

A reliable change should leave reasons behind. Not every everyday adjustment requires a report, but an important public or personal revision can usually state four things: what was believed before; what new evidence or distinction appeared; where the change applies; and what has not changed.

This record serves two purposes. Publicly, it lets others distinguish opportunistic repositioning from reasoned updating. Privately, it prevents the current view from rewriting the past into “I have always thought this”. People are motivated to preserve a sense of continuity and may underestimate changes in recollection. Keeping dated versions and the reasons that seemed adequate at the time allows an intellectual history to remain visible.

That is why deleting every old error is not always best. Some content should indeed be removed to prevent continuing harm or misinformation. Removal can be accompanied by an explanation rather than a pretence that the earlier view never existed. A correction need not preserve an error as permanent humiliation. It can show how knowledge is maintained.

Nor should “I am open to change” become a performance. The test is not how often a person declares openness, but whether they can identify evidence that would alter the view, actively encounter sources capable of challenging it, and adjust relevant action afterwards. Updating language without updating conduct often shows that the old commitment still controls the decision.

Another person's persuasion does not remove my responsibility

Being persuaded can tempt a person to transfer responsibility to the persuader: “He made me believe it.” But once reasons are accepted and enter one's actions, responsibility cannot be completely outsourced. The speaker is responsible for presenting the case honestly. The hearer remains responsible for considering source, consequence and scope.

This also distinguishes persuasion from manipulation. Good persuasion makes reasons more visible and allows the listener to reconstruct the judgement after leaving the persuader. Manipulation uses fear, identity pressure, controlled information or authority to produce a conclusion without giving the person an intelligible basis for it. The first strengthens epistemic capacity; the second bypasses it.

The distinction is important in AI dialogue. A system can rapidly generate forceful prose for almost any position. Fluency and completeness can move a user. When using AI to test a view, the question should not be only which side sounds more persuasive. Claims need sources; evidence must be separated from rhetoric; the strongest counterexample should be considered; and important matters require leaving the dialogue to inspect independent material.

If an answer changes my view, I should still be able to say in my own words which reason changed it. “The AI sounded certain” is not such a reason.

Epistemic integrity does not mean remaining unchanged

Revision exposes difference across time: the former self and present self do not entirely agree. A person need not deny the difference to remain whole. A more truthful continuity comes from connecting the versions—understanding why the earlier view was held, why the later one differs and what the change reveals about learning.

Being persuaded is therefore neither automatically failure nor automatically virtue. The criteria are whether the change responds to better evidence, reasons or framing; whether its extent is proportionate; whether responsibility and traceability are preserved; and whether the revision is reflected in action.

Under those conditions, change is not the breakdown of a capacity for knowing. It is that capacity at work. A world that cannot be revised may look stable but can preserve itself only by excluding reality. A world capable of explaining how it changes has a chance to persist through time without becoming rigid.

What needs protection is not every old opinion, but the channel through which belief remains answerable to reality, other people and counterevidence. A person's world does not necessarily collapse when an answer is replaced. Sometimes it becomes durable for the first time in the willingness to replace it.

Primary sources and further reading

Continue reading: Explore the How a Person Builds a World series.

Discover more from Geoffrey Chen

Subscribe to get the latest posts sent to your email.