Conversations with people you disagree with

Presenting evidence to someone who holds a different view frequently fails, and understanding why suggests better approaches than presenting more evidence.

Why argument underperforms

Motivated reasoning. People evaluate evidence differently depending on whether it supports what they already believe — scrutinising disconfirming evidence more critically than confirming evidence.

Which means the same argument is processed differently by two people, and the effect is stronger for beliefs connected to identity or group membership.

Beliefs are frequently social. A position held by everyone in your community is expensive to abandon, and the cost is not epistemic.

Which is why arguments that address only the factual content miss what is holding the belief in place.

Backfire effects, carefully. The claim that correction strengthens false belief was widely reported and has replicated poorly.

Larger studies generally find that corrections do work, modestly, and that the backfire pattern is rare rather than typical.

Which is genuinely encouraging: correcting misinformation is not counterproductive, it is merely less effective than one would hope.

What has better evidence

Deep canvassing. Extended, non-judgemental conversation involving exchange of personal narrative rather than argument.

Field experiments on this approach have reported durable attitude change on contested social issues, with effects persisting over months — an unusual result in a literature where effects generally decay quickly.

The mechanism proposed involves perspective-taking and the absence of pressure, rather than transmission of information.

Asking about mechanism. Asking someone to explain in detail how something works frequently reduces the extremity of their position.

The proposed explanation is an illusion of explanatory depth: people believe they understand complex systems better than they do, and attempting to explain reveals the gap.

The effect appears for causal understanding rather than for values, which limits where it applies.

Finding the crux. Identifying the specific belief on which the disagreement turns, rather than exchanging conclusions.

Establishing shared ground first. Which is not a rhetorical technique but a factual step — the shared ground is usually larger than either party assumes.

The conduct that helps

Ask more than you assert. Genuine questions, not rhetorical ones.

The distinction is whether you would be interested in an answer you did not expect.

State their position back until they agree you have it right, before responding.

Concede what is true. Most positions contain something correct, and acknowledging it costs nothing and substantially changes how the rest is received.

Do not attack the identity. An argument framed so that accepting it requires admitting a moral failing will not be accepted, regardless of its merits.

Leave room. People rarely change position during a conversation. They change afterward, in private, if the conversation left them able to.

Which means the objective of a single conversation is not agreement but making a later change possible, and pressing for agreement in the moment reliably prevents it.

The limits

Worth being honest about.

Some disagreements are about values, and no amount of evidence resolves them. Recognising this early saves everyone time.

Some conversations are not in good faith, and the techniques above assume good faith. Where they are not reciprocated, disengaging is reasonable.

And most beliefs do not change quickly. Expecting a conversation to shift a long-held position is a poor prediction, and the disappointment produces the escalation that makes things worse.

The reason to bother

Two, and they are worth separating.

The first is that you might be wrong, which is worth genuinely holding rather than saying. Conversations conducted by someone open to being persuaded go differently, and the difference is detectable.

The second is that the alternative — sorting into groups that agree — has costs that are visible everywhere, and that individuals can only address one conversation at a time.

Knowing when to stop

A useful signal, since these conversations rarely end cleanly.

When the same point is being made for the third time by either party, nothing further is being exchanged, and continuing converts a disagreement into a dispute.

Stopping while relations are intact preserves the possibility of a later conversation, which is where most actual change happens. Winning the current exchange generally forecloses it.