Halo Effect

011 min

Key takeaways

• What it is: one good, or bad, impression spreads to unrelated judgments about the same target
• Origin: Edward Thorndike, 1920, from correlated ratings of military officers
• Where it bites: hiring interviews, performance reviews, and first impressions of a product's interface
• Guard against it: rate one trait at a time, before forming an overall opinion

021 min

Where the Halo Effect shows up

A hiring manager meets a candidate who arrives on time, dresses sharply, and gives a firm handshake, and within the first two minutes rates them as more competent, more intelligent, and more likely to succeed in the role, before the interview has covered a single technical question. Nothing about the candidate's actual skill has been assessed yet.

This is the halo effect: a single, often superficial, positive impression spreads outward and colours judgments of qualities that have not actually been observed. It works the same way for a product. A visitor who finds a website's landing page beautifully designed will tend to also rate the company's support as more reliable and its pricing as more fair, neither of which the landing page's visual design has any real bearing on. Run the interview or the product test with a plainer first impression, with no change to the technical answers or the product's actual features, and the ratings on those unrelated dimensions drop with it.

032 min

Why the Halo Effect happens

The halo effect works through a shortcut psychologists call attribute substitution: forming a considered judgment about someone's competence, honesty, or reliability takes real evidence and effort, so the mind substitutes an easier question it already has an answer to, such as whether this person is likeable or this page looks good, and reports that answer as if it had done the harder work. Thorndike first measured the pattern quantitatively when he had military officers rate soldiers on independent traits like physique, intelligence, leadership, and character, and found the ratings correlated far more than genuinely independent traits should. Thorndike wrote of the pattern: "The correlations are too high and too even. For example, for the three raters next studied, the average correlation for physique with intelligence is .31; for physique with leadership, .39; and for physique with character, .28." A rater who should have been scoring four separate qualities was, in practice, scoring one overall impression four times.

Nisbett and Wilson's 1977 experiment showed the substitution happens below conscious awareness. Two groups of students watched a video of the same lecturer, warm in one version and cold in the other, then rated his appearance, mannerisms, and accent, all of which were identical across both videos. Student's judgements of a lecturer's accent, mannerisms and physical appearance were all affected by his apparent likeability. But, again, when interviewed afterwards, the participants were completely unaware of what had influenced their judgements. They did not report liking his voice because he was warm; they reported his voice as objectively more pleasant.

041 min

Where the Halo Effect comes from

Edward Thorndike, a psychologist at Columbia University, published the finding in 1920 in the first volume of the Journal of Applied Psychology, in a paper titled "A Constant Error in Psychological Ratings." He was analysing ratings that commanding officers gave soldiers on a standard evaluation form covering unrelated qualities. Thorndike did not run a controlled experiment; he examined existing rating data and noticed that raters were, in effect, scoring one overall impression rather than several independent traits. He named the pattern after the religious image of a halo, a single glow that surrounds and flatters everything within it. The finding has held up well and is rarely disputed as a real effect, though later researchers, including Nisbett and Wilson, have refined the explanation for why it happens.

051 min

Individual effects

For one person doing the judging, the halo effect distorts a specific decision: how much weight to give evidence they have not actually seen. A manager who already rates an employee highly on communication will tend to also rate that same employee highly on technical skill, on a review form that asks for both separately, even when the manager has seen far more evidence of one than the other. The error runs in a single direction, toward inflating the unobserved trait to match the observed one, never toward correcting it. A recruiter who forms a good first impression from a polished resume and a confident phone screen carries that impression into the technical interview, and unconsciously grades borderline answers more generously than they would from a candidate whose first impression had been merely adequate.

061 min

Systemic effects

At the scale of a whole company or market, the same shortcut compounds through metrics that never separate the two things being confused. A company's best-known, best-loved product forms an overall impression of quality that competitors then have to overcome, even in categories where the company has no actual track record. Review aggregators and app-store ratings show the same compounding at a smaller scale: an app from a developer with several well-reviewed titles tends to collect more generous early reviews for a brand-new, unproven release than an identical app from an unknown developer, before enough people have actually used the new release long enough to judge it on its own merits.

071 min

Examples

A notable example is the manner in which the popularity of Apple's iPod generated enthusiasm for the corporation's other products. Consumers who had a good experience with one product assumed the company's unrelated products, from laptops to phones, would be similarly well made, before trying them. The term halo effect is used in marketing to explain consumer bias toward certain products because of favorable experience with other products made by the same company. That is why an established company's new product category gets a head start that an identical product from an unknown company would not, and why a second, unrelated product line can coast for years on a reputation earned by the first.

An illustrative case shows the same pattern in hiring: a hiring panel spends the first five minutes of an interview building rapport with a candidate who happens to share the panel's alma mater. The panel later scores the candidate's answers to a set of standardised technical questions more generously than an outside reviewer, given only the transcript, scores the same answers. Nothing about the transcript changed between the two scorings; only the panel's prior warmth toward the candidate did.

081 min

How the Halo Effect shows up in product, design, and AI

Nielsen Norman Group ran a study testing websites for both visual appeal and task usability, and found a split: the websites that had high visual-appeal ratings were then tested for usability, and on average, participants' task-failure rate on these sites was over 50%, yet despite this atrocious failure rate, participants' satisfaction ratings remained high. A designer reading only the satisfaction scores would conclude the interface worked; the task-completion data said otherwise. This is the halo effect operating directly on a product metric that teams routinely use to make ship decisions.

The same distortion appears when a product manager judges a new feature by how polished its demo looked rather than by what it actually measured in a controlled test, and when a team evaluating a large language model's output judges an answer as more accurate because it is written fluently and confidently, when fluency and confidence are properties of the writing, not evidence that the underlying fact is correct. A grader scoring model outputs without a rubric that separates "sounds right" from "is right" will systematically over-trust a well-written wrong answer over an awkwardly-written correct one. The fix in every one of these cases is the same one Thorndike's data points to: separate the dimension being judged from the dimension actually creating the impression, and score them with different evidence.

091 min

How to guard against the Halo Effect

Two techniques have direct evidence behind them. The first is scoring one trait at a time across every candidate or item before moving to the next trait, rather than forming a full impression of one candidate before moving to the next; this is why standardised interview panels ask every candidate the same question in the same order and record the answer before discussing impressions. The second is blinding the rater to the information that triggers the halo in the first place, such as grading a written test without a name attached, or reviewing a pull request's logic before reading anything about the author's usual reputation.

Neither technique works by making the rater try harder to be objective. Awareness of the halo effect does not reduce it, because the substitution happens before conscious judgment starts. What works is removing the order and the information that lets one trait leak into another.

101 min

Common misunderstandings

The popular shorthand for the halo effect is that attractive people get treated better, which is true but incomplete: the effect works on any salient positive trait, not only physical appearance, and it works just as strongly in the negative direction. A single poor first impression, sometimes called the horn effect, spreads to unrelated judgments exactly the way a good one does, dragging down ratings of qualities the observer has not actually seen. This is different from a fundamental attribution error, which explains someone's actions by their character; the halo effect instead lets one known trait bleed into judgments of a completely different, unobserved trait, before any action has even been explained.

The effect is also sometimes treated as something a rater can simply correct by trying harder to be fair. Thorndike's original finding, and every replication since, shows the substitution happening below the level raters can consciously inspect or override, which is why structural fixes such as blind review and one-trait-at-a-time scoring work far better than an instruction to be more objective.

111 min

Halo Effect vs. nearby concepts

The halo effect forms the first impression; confirmation bias then defends it, and anchoring is the closest neighbour when the trait being distorted is a number rather than a character judgment. Telling the three apart matters in practice: fixing a halo problem means changing who rates what and when, while fixing a confirmation-bias problem means changing what evidence gets sought out afterward.

?6 questions

Questions people ask