Trust Puzzle

Highly trustworthy individuals think others are like them and tend to form beliefs that are too optimistic, causing them to assume too much social risk, to be cheated more often and ultimately perform less well than those who happen to have a trustworthiness level close to the mean of the population.  On the other hand, the low-trustworthiness types form beliefs that are too conservative and thereby avoid being cheated, but give up profitable opportunities too often and, consequently, underperform.  Our [empirical] estimates imply that the cost of either excessive or too little trust is comparable to the income lost by foregoing college. [emphasis added]

More here.  How much to trust others does sound like a very important parameter, and it makes sense that errors on this could have large consequences.  But it is puzzling that we make such large errors, apparently unable to learn much useful from those around us about how much to trust.  I could see how a lack of trust might prevent one from learning from others that one should trust more, but how does trusting too much prevent one from learning from others to that one should trust less?

GD Star Rating
loading...
Tagged as:
Trackback URL:
  • mitchell porter

    “how does trusting too much prevent one from learning from others to that one should trust less?”

    It’s not always obvious that someone is ripping you off. They may have an excuse or a cover story, and if you trust too much, you’ll believe it, and never even notice what actually happened.

    • Michael Turner

      Yes, but time after time? If somebody’s an easy mark, it gets around and the easy mark attract swindlers, so learning opportunities will eventually be abundant. Why does somebody getting cheated more often than most people continue to be an easy mark?

      Most likely, it’s not one single factor, but some combination of independent personality traits not easily changed, like (1) innately blunted perception of typical “tells” of dishonesty, (2) not being intelligent enough to correlate “tells” with being cheated, and (3) being easily intimidated (or otherwise manipulated, e.g. by a cheater’s sob story) even when someone suspects they are dealing with a cheat. Just being somewhat more impaired than most in two of these categories might be enough.

  • Sebastian Franck

    Maybe trust isn’t something learned – and thus cannot be unlearned. It may be a kind of innate social cognition bias. If being trusting is a fundamental character trait, there is only so much you can do to change it, even when you change your beliefs. Your weights are permanently screwed.

    • http://hanson.gmu.edu Robin Hanson

      Do you think that you cannot choose to trust people overall more vs. less?

      • Sebastian Franck

        Kind of, yeah. Just like I cannot choose to be colorblind or not. Trustingness might be a kind of pattern blindness. A case of simply not being able to recognize the signs. So time and time again my expectations of trustworthiness are thwarted. And I’m surprised every time. (I think I myself may be suffering from a slight case of this).

      • Jess Riedel

        I think it would be very difficult to reliably modify your overall trusting. When we decided to trust or not trust, we do it based on intuition and not on a calculated decision. Since we don’t have access to the “intuiting” process, it can’t be easily adjusted; it’s a black box.

  • Peter Twieg

    I think being trusting is an attribute which isn’t completely hidden from others, or at least it correlates with a lot of non-hidden attributes in such a way that if one became less trusting, one would become less appealing to others whether intended or not. And this appeal isn’t merely to those who would exploit one’s trust, but those who simply would prefer to be trusted easily. Several charismatic personality types simply would seem odd if their possessors were really inwardly jaded – and I imagine that part of the reason for this is genuinely because they’re usually not.

  • Ben

    There’s more to it than being trusting or not trusting. There’s exercising good judgment in whom to trust. Perhaps the reason people aren’t better at determine whether to trust someone is that it’s really quite hard, and that trustworthy people aren’t all that easy to identify.

  • http://www.bayesianinvestor.com Peter McCluskey

    I expect that there are benefits to signalling that you are more trusting and trustworthy than the average person. These benefits may be nonfinancial, such as having more likable friends. So I don’t find being too optimistic about trust any more surprising than the well-known forms of overconfidence.

  • Jess Riedel

    This is pure speculation, but maybe we gauge the likelihood of betrayal by other not based so much on any past data we’ve collected, but by simulating in our minds what we would do in their situation. I think that is roughly we how we make many of our social decision: by imagining what we would do in each person’s shoes, and playing out the simulation.

    That would suggest that honest people would model others as overly honest (and likewise dishonest people would model others as overly dishonest). I suspect that it would be very difficult for us to incorporate the small amount of honesty data we acquire in our lives into this black-box simulation. Rather, we would tend only to make correction based on very similar past experiences. In other words, we might learn that we shouldn’t trust pan handlers when the tell us they just need bus fare to get home, and yet be unable to use that data to correctly adjust our other estimations of honest.

    (I think I may just be rehashing some points made by Eliezer Yudkowsky awhile back on this blog.)

  • Curt Adams

    I am looking to adopt and one of the things taught in the classes I have had to take is that our attitude towards trust is mostly learned as infants and extremely difficult to change afterwards. It’s not based on rational thinking and the rational actor model just doesn’t apply.

  • Dagon

    I’m not sure that a one-dimension measure of “amount of trust in others” is sensible. I try (and succeed, I think) to be trustworthy, and I tend to trust by default on some topics and amounts and not others. The level of trust I feel varies widely among individuals and groups I interact with even for similar transactions.

    I pretty strongly believe that justified trust reduces transaction costs by a whole lot, and this is an underappreciated part of wealth disparity. But I don’t think it can be averaged very well over all transactions for an individual, let alone any sizable group.

    • Michael Turner

      One dimension isn’t enough, I agree. I’m not sure of the study’s methodology, but does it control for “trust traps”? A society can settle at a low equlibrium of generalized trust, or at a higher one. At the lower ranges, it tends to lock in poverty.

      The amount of trust that’s optimal for an individual isn’t necessarily optimal for the society or subsociety of which that individual is a member. Democracies with the rule of law and protections for the rights of minorities are a way to substitute trust in government (to behave reliably, at least, and relatively fairly) for generalized community trust. Government using such principles scales pretty well, and leads to better economic average outcomes over the long run. Communitarian trust is nice, but it doesn’t scale well at all.

      I’m wondering if there are localized trust traps even in places with relatively high generalized trust. (This study was conducted in Europe.) Living in a ghetto would mean you’re quite rational to be not very trusting of people around you; and living in a ghetto is a poor predictor of material success later on.

  • fenn

    Maybe the rubes accept/believe the excuses given by those that dupe them.

  • http://www.weidai.com Wei Dai

    Suppose there is a signal, which takes a random walk over time, but slowly. And every time you measure the signal, you get signal + a large amount of random noise. How quickly should you change your estimate of the signal based on recent observations? Clearly, the more slowly the signal changes, and the higher the noise, the slower you should change your estimate.

    Given the above, I suspect that in our EEA, our social environment changed very slowly, so we are programmed to not learn how much to trust others too quickly. This is now maladaptive in today’s world, which changes much more quickly.

  • hayden

    May I join the convo?

  • Clinton McMurray

    Perhaps the level of trust to assign another individual is a socially learned skill. As children most of us spend most of our time with family and for the most part our trust in them not only works well but is essential for survival. As we age we encounter people farther and farther removed from family. These strangers face smaller and smaller costs for ripping us off and vice versa. So, the people around we spend most time with whilst very young don’t teach us much through action about the skill of assesesing trustworthyness. The binary choice offered to children in the “stranger danger” lesson may actually exacerbate the problem.

    We are often quite emotionally damaged by our first encounters with people who take advantage of our trust. Do you remember how that feels? If we are seriously taken advantage of we might switch to mistrust on average – “once bitten, twice shy”. Perhaps the frequency and magnitude of initial trustingness and eventual switching has a good deal to do with the social networks we move in. This all seems very testable.