Professor Nalini Ambady, Stanford University, is the senior author of the study I discuss below. Before you read on, I ask that you take a look at the website Nalini Needs You. Nalini has a rare form of cancer and needs to find a genetic match for a bone marrow transplant. If you are not a likely match, perhaps you know someone who is, and both of you will find it easy to register as donors. For more information, please also see this post by our colleague Sam Sommers.
Now to the topic of perceptions of (un)trustworthiness. Trust is the quintessential social dilemma. There is much to be gained from trusting others if they reciprocate, and much to be lost if they don’t. Never trusting is not a solution because it brings social isolation and economic hardship. While humans cannot solve the trust dilemma, they will always try to mitigate it. I have previously compiled a list of 10 cues people may look to for help. Two of these require further comment, namely, the idea that certain facial features signal trustworthiness and the idea that people who promise to be trustworthy are more likely to be trustworthy than people who do not.
A team of authors around Nalini Ambady (Rule et al., 2013) has conducted a series of elegant studies to see if people can tell from facial photographs alone whether a person has committed a crime or minor act of cheating implicating a lack of trustworthiness. Across all four studies, the answer was consistently ‘No.’ At the same time, however, there was significant agreement among the respondents as to who was trustworthy and who was not. Even more remarkably, a part of the limbic system responsible for social intelligence (the amygdala) was active when the respondent felt having detected either very high or very low trustworthiness. Since these feelings are essentially uncorrelated with actual trustworthiness, there will be two types of error. Sometimes, people trust when they shouldn’t, and sometimes they fail to trust when it would have been safe. Unfortunately, only the former error comes with clear negative feedback (betrayal), making it easier to learn how to distrust than to trust.
Rule et al. carefully consider the context of their finding. The hypothesis that trustworthiness is written on a person’s face assumes that trustworthiness is a stable personality trait that is expressed consistently across situations. Yet, when they compiled a list of trustworthy and untrustworthy individuals, they had to rely on information regarding criminal acts committed in specific situations. When some photos show individuals convicted of corporate fraud, whereas others show Nobel Laureates, a prominent act of untrustworthiness is known about the former. It is not known how many times or on what occasions the convicts were trustworthy, and it is not known how many times or on what occasions the laureates were untrustworthy. Perhaps facial expressions reveal a person’s intention to betray at the moment of decision. Rule et al. note that one study (Verplaetse et al., 2007) found evidence for this idea.
Rule et al. found that self-reports of past cheating were negatively (r = -.26) correlated with the magnitude of cheating in an experimental task. This finding was marginally significant (p = .09), but strong enough to be surprising. Participants responded to the statement “I have never cheated on a test” before they had an opportunity to do so, which amounts to an opportunity to make a promise rather than to rationalize past behavior. My earlier suggestion that promises are more likely to be kept than broken must therefore be qualified. A promise remains a valid, if imperfect, cue if it is made in the context of an impending exchange and if it is cast as a personal commitment to the trustor (Dawes et al., 1977). In contrast, individuals who broadcast their general trustworthiness (or dignity or integrity) may be hiding something.
I regard the primary finding reported by Rule et al. as important, although (or perhaps because) it is a repeated null result. Null results are often frowned upon because they raise suspicions that the researchers lacked the talent or the will to produce positive, or non-null effects. The Rule et al. finding that people cannot tell trustworthiness from a facial photo is important, though, because folk psychology suggests that they can. When there is no real (or positive) effect, a replicable null results is required to restore disbelief. Short of running Bayesian analyses, Rule et al. did everything possible to corroborate the view that there is no effect (i.e., a priori power estimates, replications, and a meta-analysis).
Now suppose for fantasy’s sake that there had been a positive effect. Suppose people could tell from a facial photo whether a person had committed a criminal offense reflecting a lack of trustworthiness. Also suppose (as I think was the case in the Rule et al. studies) that the photos were taken before the individual was convicted. This should allow you to expect that you, and everyone else, is able to tell if your boss (or anyone else) will commit a corporate crime (or anyone else will cheat someone somewhere). By being able to tell, I am referring to the modest statistical argument that you would be able to beat chance predictions significantly, if with a small effect size. The implications are considerable. Since you do not know for which occasion the person will plot betrayal, you will probably disengage from interaction more broadly. You will thereby reduce the number of Type 1 errors (betrayals), but inasmuch as betrayal is overall less likely than reciprocity, you will increase the number of Type 2 errors (missed opportunities) at a steeper rate. The result is that your predictive accuracy will end up nullifying itself.
Rule et al. note that people can make reasonably good predictions when distinguishing extroverts from introverts or the smart from the stupid. That makes sense because extroversion and intelligence do not hinge on individual acts, but instead reflect aggregates. You are not an introvert if you fail to show up at one party, and you are not (necessarily) stupid if you flunk one test. If you betray trust once in a dramatic and public way, the label of trustworthiness will stick. Trustworthiness is not a character trait or a talent, but a reputation. It is easily lost and hard to repair.
If god’s job is to tell the righteous from the wicked, a hard-shell interpretation is that one breach of one commandment (even if imagined) is enough to put a person in the sinners bin. A softer interpretation is that all acts matter so that good deeds can offset bad ones. Being omniscient, He will perform better at the last judgment than the participants in the Rule et al. studies, but if He otherwise reasons like humans, he will take the hard view. He will look for existence proofs of wickedness and not average. I find that somehow disheartening.
Back to donating blood and marrow. Donating is not an act of trustworthiness in the narrow sense. It is an act of loving-kindness, where you may not know who will benefit. In western parlance, it is also an act of indirect reciprocity. If you donate, you hope, but cannot know, that others, who do not know you, will donate so you can live. In the Indian worldview, I think this amounts to karma. So make it a good one.
Dawes, R. M., McTavish, J., Shaklee, H. (1977). Behavior, communication, and assumptions about other people's behavior in a commons dilemma situation. Journal of Personality and Social Psychology, 35, 1-11.
Rule, N. O., Krendl, A. C., Ivcevic, Z., & Ambady, N. (2013). Accuracy and consensus in judgments of trustworthiness from faces: Behavioral and neural correlates. Journal of Personality and Social Psychology, 104, 409-426.
Verplaetse, J., Vanneste, S., & Braeckman, J. (2007). You can judge a book by its cover: The sequel. A kernel of truth in predictive cheating detection. Evolution and Human Behavior, 28, 260-271.