The researchers who defined selfishness as a measurable trait did not anchor it to any behaviour: their definition turns on maximising your own gain at others' expense while holding beliefs that justify it, which is the part a list of behaviours cannot see

There is a durable genre of article that lists the behaviours giving away a selfish person. Interrupting. Never asking questions. Ordering first. Taking the better chair. The lists are satisfying to read because everyone can attach a face to them within about four seconds.

The awkwardness is that the researchers who have spent the last decade building a measurable definition of exactly this trait did not build it out of behaviours like these, and their reasons are worth understanding.

How the trait is actually defined

The most developed attempt is the Dark Factor of Personality, proposed by Morten Moshagen, Benjamin Hilbig and Ingo Zettler in Psychological Review in 2018. Their argument is that the aversive traits studied separately for decades, including Machiavellianism, narcissism, psychopathy, sadism, egoism, greed and spitefulness, share a common core.

The definition they give, set out on the research group’s own site, is precise: the general tendency to maximise one’s individual utility while disregarding, accepting, or malevolently provoking disutility for others, accompanied by beliefs that serve as justifications.

Read that again and notice what is missing. There is no behaviour in it.

What the definition contains is an orientation, a disregard for the cost to others, and a set of beliefs doing the work of making that acceptable. Those beliefs are not decoration. The researchers describe high-D individuals as holding convictions that rationalise the harm: that they are superior, that others are inferior, that dominance is legitimate, that everyone behaves this way underneath.

The older and better-established measure points the same direction. Honesty-Humility is the sixth dimension in the HEXACO model developed by Kibeom Lee and Michael Ashton, and it is the one the Big Five does not capture. It concerns manipulation, entitlement, materialism and exploitation as tendencies, assessed across many items and situations, not identified from a single observed act.

Why one behaviour carries so little information

The deeper problem with the checklist format is not that its items are wrong. It is that a single observation is close to worthless as evidence, and this is one of the better-replicated findings in personality psychology.

In 2001, William Fleeson published a paper in the Journal of Personality and Social Psychology that had people report their own behaviour repeatedly over two to three weeks. What he found is that people vary enormously from occasion to occasion. Within-person variability across those weeks was near the high end of what was possible, close to the total variability in the whole sample.

Fleeson’s summary is that individuals differ from themselves over time at least as much as they differ from each other on average, and that the typical person routinely displayed nearly all levels of nearly all traits within the observation window.

The same person is generous on Tuesday and grasping on Thursday. Not as a contradiction, but as the normal shape of a trait.

Seymour Epstein made the methodological version of this argument in the Journal of Personality and Social Psychology in 1979. A single behavioural observation is unreliable in the technical sense, because so much of it is occasion-specific noise. Aggregate enough observations and the signal appears clearly. The trait is real. It is just not visible in any one moment, which is precisely the moment the checklist genre asks you to judge from.

This matters most where the stakes are highest. The colleague who took credit in one meeting, the friend who did not offer to pay once, the founder who was short with a junior employee on a bad afternoon: each is one draw from a distribution, and one draw is not the distribution.

The justifying belief is the part that travels

If there is a usable signal in this research, it is not on the behaviour list. It is in the second half of the D definition.

Behaviour varies by occasion. The belief that makes the behaviour acceptable is more stable, because it has to be reconstructed each time and tends to be reconstructed the same way. Someone who takes the larger share and is faintly embarrassed about it is doing something different from someone who takes the larger share and explains why they were owed it.

The explanation is the more informative object. Not because one instance of it settles anything, but because it is the component the researchers identified as travelling with the trait rather than with the situation.

We would put this no more strongly than that. It is an inference from how D is defined, not a validated detection method, and no serious researcher in this area claims you can read the trait off a stranger.

What this does not settle

A fair amount, as it happens.

D and Honesty-Humility are both assessed largely through self-report questionnaires, which have obvious limits when the trait in question includes a tendency towards self-serving presentation. The field is aware of this and has worked on it, but it is a real constraint and not a solved one.

Fleeson’s 2001 work is one line of research, influential and much cited, though the broader person-situation question it addresses has been argued over for fifty years and is not closed. Nothing here says traits are illusions or that behaviour is random. The point is narrower: that the diagnostic value of any single act is low, not that character does not exist.

These are also research constructs and not clinical categories. D is a dimension along which everyone sits somewhere, described by researchers as a continuous trait. It is not a label to apply to a colleague, and the paper does not offer it as one.

There is one more limitation worth naming plainly. Almost all of this work rests on Western university and online samples, which is the standing problem with personality research generally, and it is not obvious that the same structure would appear everywhere.

What we find useful about the research is how badly it fits the format it keeps getting written up in. The trait is real, measurable and consequential, and the honest description of how to detect it in another person is that you mostly cannot, at least not from anything you could put in a list.