TL;DR
Shown ambiguous actions under time pressure, people judge significantly more of them to be intentional — and asked to describe accidental-looking events freely, they still generate intentional readings.1 Intention is the default interpretation; concluding “that was an accident” costs additional cognitive work.1 Layered on top is the older finding that observers infer stable character from behaviour even when explicitly told the situation forced it.2 Together they explain a specific everyday error: crediting people with deliberate strategy they never had. Both effects are real; both are also smaller and more conditional than the textbook version claims.2
Something watching, and presumed to be deciding. Usually it is only recording. Photo: Michał Jakubowski on Unsplash.3
The Assumption Hiding in Ordinary Complaint
Listen to how people describe each other’s behaviour and notice how much planning gets attributed. He said that to undermine me. She scheduled it that way so that I couldn’t attend. That comment was designed to remind everyone who’s senior.
Sometimes this is right — people do scheme. But the striking thing is how automatically the calculated reading arrives, and how rarely the alternative gets serious consideration: that the person was tired, or reaching for the nearest available phrasing, or following a habit, or had simply not thought about it at all. The interpretation attributes a level of self-monitoring that most people, most of the time, are demonstrably not doing.
That asymmetry isn’t a moral failing of the suspicious. It’s a documented feature of how the interpretation machinery runs.
Intention Is the Default, Not the Conclusion
The cleanest evidence comes from Evelyn Rosset’s work on what she called the intentionality bias, across three studies.
In the first, participants read ambiguous action sentences — the type where either reading is available, like “he set the house on fire” — under normal conditions or under time pressure. Those under time pressure judged significantly more of them intentional.1 In the second, asked to describe ambiguous actions spontaneously with no prompt about accidents, participants generated significantly more intentional interpretations — even for prototypically accidental actions.1 The third study compared processing: people who had judged sentences as intentional-or-not later remembered more of the unintentional ones than a control group did, consistent with the unintentional judgement having required extra work to reach.1
The conclusion Rosset draws is that adults hold an implicit bias to infer intention in all behaviour, and that this supports a dual-process account: intention is attributed automatically, and the accidental reading is a deliberate override applied afterwards — when there’s time and capacity to apply it.1
Which means the everyday version follows directly. Under load — busy, stressed, mid-argument, scrolling — the override doesn’t run. What’s left is the default, and the default says they meant it.
Why the Default Might Be There
There’s a plausible reason this asymmetry exists rather than the reverse, and it isn’t flattering to our accuracy but it is flattering to our survival. The error costs aren’t symmetric: treating an accident as deliberate wastes some social energy, while treating a genuine threat as accidental can be considerably worse. A system tuned to over-detect intention fails safely. It just fails often.
That framing also predicts where it should go badly wrong — and there’s evidence it does. Research on conspiracist ideation found that people biased toward inferring intentional explanations for ambiguous actions were more likely to endorse conspiracy theories, with intentional inference and anthropomorphism both independently predicting conspiracism.4 The correlations were modest — around r = .22 for intentional inference in one study — but the logic is tidy: a conspiracy theory is precisely an account in which nothing is accident, coincidence, or drift. It’s the intentionality default with no override ever applied.
The Older, Bigger Version of the Same Error
Stacked on top of reading intention into acts is reading character out of them, and the canonical demonstration is nearly sixty years old.
In Jones and Harris’s 1967 study, participants encountered essays or speeches taking a pro- or anti-Castro position. Half were told the writer had freely chosen the position; half were told it had been assigned. Participants then estimated the writer’s actual attitude. Even in the assigned condition — where the behaviour was explicitly, fully explained by the situation — participants still inferred that pro-Castro writers genuinely held pro-Castro views.2 The situational information was available, acknowledged, and largely ignored.
The mechanism mirrors the intentionality finding: dispositional judgement comes first and automatically, situational correction is effortful and second, and when people are cognitively busy they skip the correction.2 So both errors share a shape. The fast reading treats behaviour as a window onto a stable inner cause — a trait, a plan, a motive. The slow reading remembers that circumstances produce behaviour too, and the slow reading is the one that gets dropped when attention is scarce.
Where the Textbook Version Overreaches
Now the part that usually gets left out, and it matters because this is a field with a real history of over-claiming.
The famous actor-observer asymmetry — the tidy story that we explain our own behaviour situationally and everyone else’s dispositionally — has been substantially cut down. A meta-analysis by Bertram Malle found the asymmetry to be much smaller and more conditional than textbooks usually claim, holding mainly for negative or intentional behaviour rather than as a general law of social perception.2 So “we judge others harshly and ourselves leniently” is not a universal effect; it’s a narrower one that shows up in specific conditions.
The intentionality bias has drawn published challenges of its own, with later work disputing whether intentional inferences really do outrun unintentional ones and revisiting the paradigm — so its honest status is a real effect with a live methodological argument attached, not a settled law. I’d treat the direction as well supported and the size as unresolved.
The deepest complication is cultural, which for this site’s readers is the most interesting part. Miller’s 1984 comparison found American adults emphasising dispositional explanations while Indian adults favoured situational ones — and, tellingly, the gap widened with age, which points toward the bias being culturally learned rather than innate.2 The broader pattern is that collectivist cultures show weaker dispositional biases.2
Which makes the name almost funny. The “fundamental” attribution error may be substantially a habit of individualist cultures — the very cultures that named it and then assumed it was universal. Worth sitting with if you grew up somewhere that explains behaviour by circumstance first: the bias described here may be less yours than the literature’s.
What This Actually Licenses
Not much charity by default, and that’s worth being clear about. Nothing here says people never scheme; some do, deliberately and often, and a framework that rules that out is worse than the bias it’s correcting. The claim is narrower.
It’s that the calculated reading arrives free and the uncalculated reading costs effort, so the ratio of scheming you perceive to scheming that exists should be expected to run high — particularly when you’re tired, threatened, or arguing. And it’s that the perceived deliberateness of an act is partly a fact about your processing conditions rather than entirely a fact about the actor.
The practical version is a single question, asked at the moment the motive seems obvious: would I still read it this way if I weren’t busy? If the answer is no, the reading is telling you about your own bandwidth. Most people, most of the time, are not running the strategy you’ve assigned them. They’re improvising, same as you.
