The part you play

Most people who know you know a version of you with fewer sides than you have. At work you are the calm one; among friends, the planner; online, perhaps a single opinion at different volumes. It is tempting to call these characters fake, as if a fuller and truer person waited behind them. That framing gets the economics wrong. A simplified character is often the only version of a person that others can actually use, and once performed in front of them it becomes partly true. Getting out of one depends less on consistency than on telling people, early, what else you are.

Start with why anyone would want you simpler. The sociologist Ezra Zuckerman describes a rule of thumb in labour markets, one he and his colleagues documented in film casting: candidates with experience in one category of work are screened out when they apply for another.¹ The rule matters most, he argues, when the quality of someone's work is hard to judge and credentials say little.¹ An employer who cannot evaluate an actor's range directly reads the record instead, and a run of thrillers says "thriller actor".

This is the trap for anyone who really can do several things. In a market that judges by category, a person who works across many categories looks very like a person who keeps moving because they keep failing.¹ Zuckerman's word for the risk is "non-entities": people no audience recognises as fit for any particular job.¹ Range, which ought to be an asset, reads as a warning sign, and the person who has it does better to hide most of it.

The natural assumption is that this is a problem of power: studios told contract actors what to play, so freedom would let them play everything. Zuckerman tested something close to that. Hollywood moved from a studio system, in which actors were under long contracts and assigned their parts, to a market in which actors can turn roles down. That change did not bring a fundamental decline in how concentrated actors' careers were in particular genres, and most of the modest decline he did find came with the decline of the Western.¹ He cautions that the two eras do not compare cleanly. But the right to refuse work did not, by itself, produce a generation of generalists.¹ Once others are reading you by category, the lock is in their reading, not in who assigns the work.

The performance writes back

In 1992 the psychologist Dianne Tice asked introductory psychology students to answer interview questions as though they were either extraverted or introverted. Some believed they were being watched and identified by a graduate student behind a mirror; others answered anonymously. Afterward, the students who had publicly played extraverts rated their "true" selves as considerably more extraverted than those who had publicly played introverts, and when they were sent to a waiting room with a stranger they placed their chairs closer to that person.² Students who gave the same performance anonymously showed no significant difference on either measure.² Choosing the portrayal, drawing on one's own past, and expecting to meet the audience again all strengthened the effect.²

The study is small and old, with twenty people in each condition, and everything was measured in the same session.² But its direction matters for this question. The audience was not incidental to the change; it was what made the performance stick. A character played in front of people who will remember it is being written partly by them. That is why "fake" is the wrong word for most of these characters. They are not masks over an unchanged face but a joint account, kept by the person and the people who watch them, and both sides add to it.

What audiences actually punish

If that account is hard to change, the obvious reason is that people punish inconsistency. Someone known as the careful one who suddenly takes a reckless bet looks unreliable; someone who preached one thing and did another looks like a hypocrite.

A set of experiments by Jillian Jordan, Roseanna Sommers, Paul Bloom and David Rand suggests that is not quite what is punished. Participants read short scenarios about people who condemned a behaviour such as illegally downloading music. Hypocrites who condemned it and did it were judged harshly, more harshly even than people who lied outright about their own good conduct.³ But "honest hypocrites", who condemned the behaviour while admitting that they sometimes did it themselves, were judged no worse than people who simply did it without comment.³ Admitting to some unrelated failing did not help.³ The authors' explanation is that condemnation works as a signal of one's own conduct, and what people resent is being given a false signal.³

Those experiments were about moral condemnation, not personality, so carrying them over to character is a step beyond the evidence. If the mechanism does carry over, the cost of stepping outside a role comes less from the inconsistency than from the earlier performance looking like a claim that turned out to be false. A person known only as calm, who turns out to have a temper, has in effect been signalling something untrue; a person who has said all along that they are calm at work and short-tempered in traffic has not.

That leaves an unglamorous route out. Consistency will not provide it, since consistency is what built the character. A sudden display of everything else will not either, because in a market that reads by category, range looks like failure. What remains is to keep the other sides on the record before they are needed, so that when one of them shows up it confirms what people were told instead of contradicting it.