You Dislike Most Folks

You might love humanity in the abstract, but given greater exposure to any one person, the odds are pretty good that you’d soon dislike them, often intensely.  JPSP ’07:

Although people believe that learning more about others leads to greater liking, more information about others leads, on average, to less liking. Thus, ambiguity—lacking information about another—leads to liking, whereas familiarity—acquiring more information— can breed contempt. This “less is more” effect is due to the cascading nature of dissimilarity: Once evidence of dissimilarity is encountered, subsequent information is more likely to be interpreted as further evidence of dissimilarity, leading to decreased liking. …

We do not argue that increased information leads to less liking in every case, but rather that this is the case on average. Individuals may feel overly positively toward their significant others, but these are the rare exceptions who were liked enough to stimulate efforts to acquire more information.

Why do we so consistently misjudge here?  The paper suggests:

Given the ultimate goal of finding a mate, it might be adaptive to start with a positive bias to generate many new options from which to choose; given limited capacity, however, in both available time and cognitive capacity, it may be adaptive to switch to a negativity bias while screening to eliminate poor options quickly. In fact, the robustness of optimism prior to first dates may be essential in motivating people to persevere in a long and arduous screening process.

This doesn’t make much sense to me.  Instead, let me suggest that “nice” people, who tend like more others, also tend to be more liked by others.  So we are built to try to appear nicer by appearing to like others more.  Our initial attitude toward strangers is more visible to most folks than our later dislike for the few we come to know better. Hat tip to Rob Wiblin.

More evidence that people just aren’t as nice they seem:

We introduce the joy-of-destruction game. Two players each receive an endowment and simultaneously decide on how much of the other player’s endowment to destroy. In a treatment without fear of retaliation, money is destroyed in almost 40% of all decisions.

In the hidden treatment … on average, 39.4% of all decisions involve the destruction of at least some of the partner’s endowment. … in the hidden treatment substantial amounts are burned (in total 20.4% of the maximum allowed).

GD Star Rating
Tagged as:
Trackback URL:
  • michael vassar

    But that last datum only implies that around 20% of people are malicious to strangers, even given no risk to themselves, and when the idea of doing so is suggested to them by authorities. That’s about what I would have predicted.

    The first finding is odd given the ‘mere exposure effect’. Might a simple diminishing marginal utility model work well here though? People like others in small doses? I know I was always baffled by the idea that solitary confinement could possibly be worse than sharing a cell with a random prisoner, but this seems to be the empirical consensus, unless ‘solitary confinement’ is simply a euphemism.

  • Stuart Armstrong

    Somewhat weak evidence, I feel.

    I’d require more before I change my opinion on this to a large extent.

  • Ungated version of “The Pleasure of Being Nasty”.

    This is an interesting study in the strain of pure darkness that exists in the human spirit. Apparently at least 20% of 40 undergraduate students at Tilburg University were evil.

    • Doug S.
    • Thanks for the paper link. Looking at the experiment setup, it doesn’t say what “destruction of income” means exactly. If it means that the University simply keeps the money, the outcome would just imply that some students prefer the University to have that money instead of the other participants.

      To really show “evilness” the setup has to be that the “destroyed income” is used to buy some costly goods, which then get physically destroyed. (Just burning the money doesn’t work either, because in that case the value of money goes up slightly and you’re just spreading the wealth amongst all money holders.)

      • Seems unlikely to me.

        What I am bothered by more is the small amounts. People behave differently – with less thought – when the amounts at stake are small, as opposed to when they are high.

        Still, looking at the graphs in the study article, it certainly does appear that evil is what’s being demonstrated.

      • BavarianTarzan

        […] subjects were asked to indicate what part of
        the income of the other subject they wished to destroy.

        No actual resources are being destroyed! And I suppose that they not even destroyed the money, because DESTROYING MONEY IS ILLEGAL, so whenever somebody asked me this kind of question, I would simply laugh at them and tell them “Sure, go ahead, destroy the money”. I think that a lot of the subjects in this study knew this, so linking saying “Yes” to destruction to anything like “evil” is just utter nonsense to me.

    • I added to the quote on that – 40% were evil!

      • But 20% could have been retaliating against a first strike or probable first strike.

      • In each round choices were simultaneous. First and last round destruction was not less than average rounds.

      • I thought they were iterating with the same partners on each round (so might know about a previous strike). Also, I thought there was a decided spike upward in last-round destruction.

  • Is there necessarily a misjudgment going on here? There is no particular reason to “get it right” on your average level of liking someone. If the variability of likability is maintained even after you get to know people better then there is no mistake.

    For example, we could think of a simple model where having more people in your life as aquantiances is good. You have established that these people are not outright enemies and at a minimum there are likely to serve as watch dogs for danger.

    Having lots of close friends, however, would not necessarily be good. You are likely to devote time and energy towards these people, even giving up some of your own resources to help them. If there goals are not compatible with your long term success then this is counterproductive.

    So we have an emotional mechanism that draws people close but counteracts the natural bonding process with strong dislike for people not similar enough.

  • Some of this could be a mechanism to avoid incest.

  • jonathan

    A contrary view to Swift’s, that he could hate the human race as he loved a Tom, Dick, etc. Or is it? I think at least two threads are involved: personal contempt and global contempt. For example, Nazi Germany (and now some Muslim states) dehumanized Jews and spoke of them as having inherent characteristics. The commander of Auschwitz in the psychological memoir compiled by his examining (French) doctors noted that it was this labeling which enabled his guards to kill people who looked their own mothers and sisters.

    On a more mundane level, in Detroit, we used to speak of racism in broad terms as “the blacks” and “the whites,” as the races were the Sharks and Jets fighting for territory. Again, group labeling.

    That doesn’t mean on the personal level you don’t dislike more. Ask anyone who dates. Feelings at best get more nuanced, not less. But that dislike also seems to translate into less group labeling.

  • Psychohistorian

    Wow. Just wow. The results of this study are really nowhere near what you’d get reading this post. Two of the studies use evaluations of people generated by random grouping of traits, i.e. not lists of traits based on actual people, and the third study is based purely on dating impressions, where standards for liking and similarity are likely to be highest, and has a massive survivorship problem – people who were right when they thought they’d have a great date would have left the dating site. More detail follows…

    First, for two of the studies showing this effect, real people were not evaluated. They took lists of traits (one from a study, one from actual people) and then made person-lists of random length from random traits, then asked people how much they liked those person-lists. This does not demonstrate any clear relations with actual people, because we have no idea how well these person-lists correspond to actual person-correspondence frequencies.

    As an exaggerated demonstration of this point, suppose descriptions are randomly drawn from the list: “friendly, courteous, serial killer, cat-lover, energetic, outgoing, shy, funny, puppy-torturer, polite, reliable, resourceful.” You’re going to like people you know less about, because the odds of seeing the two horrible traits increase as list length increases. However, of course, the odds of an actual person you meet having either of these two traits is phenomenally low. Because the person-lists generated from the source-list do not create a population that is similar to the actual population, your reaction to person-lists says very little about how you would react to actual people.

    The third study, that does use actual people, is based on dating web sites and impressions before and after dates. They admit the serious problem that people who had great dates may not have returned to the site, thus biasing their result negatively. Their way around this is arbitrary, hand-wavy, and largely a straw man – I’ll explain it in at the end of this comment, as it is quite complex – or just read page 5 of the study. Worse, the study focuses quite specifically on people who are using a dating site. These people have very obvious and strong incentives to be picky and to reject bad matches. The cost of having an unduly high opinion of a romantic partner is extremely high; the cost of having an unduly high opinion of a potential friend or coworker is much, much lower. Thus, the results don’t generalize to, “You’d like most people a lot less than you’d think,” so much as, “You’re more optimistic about people’s romantic compatibility with you before you date them compared with after.” They do provide great evidence of this, minus the survivorship bias problem, which could be large. But an accurate generalization would not have made for nearly as catchy of a title or abstract as an incredibly overbroad one.

    For completeness, here’s the problem with their dismissal of survivorship bias. Their claim is that, because there were more (6 vs 4) super-positive (10/10) responses in the post-date sample than in the pre-date sample, the lowered average response “did not originate from a gap in very positive ratings but rather from a redistribution of liking across the whole range of ratings”

    First, the numbers are dubious (no stat. significance is provided for 6 10’s vs. 4 10’s) and arbitrary (Were there more 9’s? More 8’s? Why is the number of 10’s the only relevant one?). Second, even if the gap originated from a redistribution, this does nothing to address the issue of people with positive responses not responding to the survey! It’s an interesting study, I admit, but they try really, really hard to stretch their data well beyond its breaking point.

    it involves an argument based of off the relative frequencies of highest (10 on a 10 pt scale) ratings between the pre-and-after group (6 10’s in the post, 4 10’s in the pre, absolutely no mention of p-value), and the belief that because there were more maximally positive answers – if you’re curious, it’s at the bottom of page 5) It finds, unsurprisingly, that women have much lower-because one thing about a lot more than people you know

    • Unless you find evidence that serial killer or puppy torturer was actually one of the traits, your skepticism about constructed trait combos seems a bit overblown.

      • Psychohistorian

        The specific traits aren’t the issue. The relative distribution is. It’s fine to include serial killing and puppy-torturing on the list, so long as the fictional people you generate have these traits with a frequency comparable to the actual population. If your fictional population does not have a trait distribution that resembles the real population, then the fact that you don’t like fictional people says little about your odds of liking real people. I used those two traits because they are real but rare, so the frequency problem is obvious. I suppose I could have used something less exaggerated, but the point works either way.

        “People like fictional people less the more they know about them” does not translate into “People like real people less the more they know them” unless there’s at least a tiny bit of evidence that the two groups are comparable in their distribution of traits. That issue isn’t even addressed in the study.

  • While it’s probably true that on average we tend to like people less as we learn more about them (compare the percentage of people you end up befriending to those you think are interesting on first impression) it will also be true that in the cases that matter the reverse tends to be true.

    I mean in both dating and friendship it seems to happen that once you overcome some threshold of attraction/interest learning new things about the person is almost always positive. This is probably just some evolutionary alliance building/pair bonding adaptations coming into play but it might also simply be that you are more likely to appreciate traits that are compatible with the things you value and are also more receptive of them when coming from such a source.

  • Noumenon

    It might not be a good study but it’s certainly true of me. I hate everyone the more I discover that they really aren’t like me.

  • Atavist

    An alternative explanation comes from another favorite of Hanson’s, construal-level theory. As social distance decreases between two people, the level of construal shifts down and people begin to focus less on the “superordinate” aspects of each others’ actions (ie, ends, ideals, etc. — all very good for impression management) and more on the “subordinate” aspects of each other’s actions (ie, habits, irksome means to ends, situational constraints). Perhaps what people initially like are each others’ idealized aspects represented at a high level of construal, and what people begin to dislike are the conflictingly base and all-too-common aspects represented at a low level of construal. If familiarity breeds contempt at least partly through a shift in construal level, increasing some other distance variable (eg, space, time, hypotheticality) between two familiarized individuals might partly reverse the dislike effect.

    • If I understand what you’re saying, it sounds like this could explain something that’s always seemed a little odd to me. An example is that if I’m away on vacation, and bump into somebody there who is (say) barely an acquaintance back at the office, then there is a tendency to 1) interact more extensively with the person than I ever did or would before, and 2) up-level the relationship even when back in the old context.

      Then again, this phenomenon can probably be explained just by 1) the effect of shared history on relationship level, and 2) matching your interaction investments to the available opportunities.

      • Atavist


        I don’t quite see how CLT, which is a theory about how our mental representations of things differ in level of abstraction across conditions of distance/proximity, would account for that phenomenon. I think you’re right to look to other factors like the ones you suggest in the second paragraph.

  • Pingback: Dispatches from Arpanistan()

  • Pingback: links for 2009-10-16 « Fantasising Zombies()