Toward Better Signals

While we tend to say and think otherwise, in fact much of what we do is oriented toward helping us to show off. (Our new book argues for this at length.) Assuming this is true, what does a better world look like?

In simple signaling models, people tend to do too much of the activities they use to signal. This suggests that a better world is one that taxes or limits such activities. Say by taxing or limiting school, hospitals, or sporting contests. However, this is hard to arrange because signaling via political systems tends to create the opposite: subsidies and minimum required levels of such widely admired activities. (Though socializing such activities under limited government budgets is often effective.) Also, if we put most all of our life energy into signaling, then limits or taxes on just signaling activities will mainly result in us diverting our efforts to other signals.

If some signaling activities have larger positive externalities, then it seems an obvious win to use taxes, subsidies, etc. to divert our efforts into those activities. This is plausibly why we try to praise people more for showing off via charity, innovation, or whistleblowing. Similarly, we tend to criticize activities like war and other violence with large negative externalities. We should continue to do these things, and also look for other such activities worthy of extra praise or criticism.

However, on reflection I think the biggest problem with signals today is the quality of our audience. When the audience that we want to impress knows little about how our visible actions connect to larger consequences, then we also need not attend much to such connections. For example, to show an audience that we care enough about someone via helping them to get medicine, we need only push the sort of medicine that our audience thinks is effective. Similarly for using charity to convince an audience we care about the poor, politics to convince an audience we care about our nation, or using creative activities to convince an audience we promote innovation.

What if our audiences knew more about which medicines helped health, which charities helped the poor, which national policies help the nation, or which creative activities promoted innovation? That would push us to also know more, and lead us to choose more effective medicines, charities, policies, and innovations. All to the world’s benefit. So what could make the audiences that we seek to impress know more about how our activities connect to these larger consequences?

One approach is make our audiences more elite. Today our efforts to gain more likes on social media have us pandering to a pretty broad and ignorant audience. In contrast, in many old-world rags-to-riches stories, a low person rose in rank via a series of encounters with higher persons, each of whom was suitably impressed. The more that we expect to gain via impressing better-informed elites, the better informed will our show-off actions be.

But this isn’t just about who we seek to impress. It is also about whether we impress them via many small encounters, or via a few big ones. In larger encounters, our audience can take more time to judge how much we really understand about what we are doing. Yes risk and randomness could dominate if the main encounters that mattered to us were too small in number. But we seem pretty far away from that limit at the moment. For now, we’d have a better world of signals if we tried more to impress via a smaller number of more intense encounters with better informed elites.

Of course to fill this role of a better informed audience, it isn’t enough for “elites” to merely be richer, prettier, or more popular. They need to actually know more about how signaling actions connect to larger consequences. So there can be outsized gains from better educating elites on such things, and from selecting our elites more from those who are better educated on them. And anything that distracts elites from performing well in this this crucial role can have outsized costs.

Of course there’s a lot more to figure out here; I’ve just scratched the surface. But still, I thought I should plant a flag now, and show that it is possible to think more carefully about how to make a better world, when that world is chock full of signaling.

GD Star Rating
Tagged as: ,
Trackback URL:
  • Joe

    A possible concern is that elites might like to signal their eliteness, and do so by favouring displays that non-elites find harder to understand, rather than judging solely on usefulness factors. Your own descriptions of evaluations within academia seem to portray this. And I think the same thing occurs more generally with the concept of “taste”: you can show off how much of, say, a film buff you are by picking an obscure movie as your favourite, and dismissing anything too popular.

    My point is that merely educating elites may not work if hidden signalling biases pervade not just the work they’re judging, but their own judgements too.

  • I’d avoid attempting to eliminate signaling, as that’s hopelessly baked deep into human nature. So to your point, channeling signaling effort in socially beneficial directions seems the best way to go.

    It’s far less clear to me that relying on elites to sift signals is a viable. That seems somewhat the system we already have, and it tends to naturally be gamed. Rather, a transparency of consequences for your signaling efforts seems more achievable. And making transparency a more common norm. So make it more clear what the effects of particular kind of medicine achive, or what different groups who perform medicine achieve. Even if people don’t care about making the effects of medicine honest, or a norm for transparency, once that data is provided clearly, it will be hard to ignore and force more effort into medicine the benefits health.

    Of course this is the effective altruism model. And obviously that kind of model can also easily turn inward into rewarding status against it’s own members rather than focused outwards.

    But nothing is perfect against the gravitational pull of signaling.

    So I think pushing for norms of transparent effectiveness is the primary model to pursue. Even if it can be quixotic at times.

    • “Transparent connections between acts and larger consequences” seems the same as making everyone better educated on such connections. Yes of course when possible, but the question is how to make this possible.

      • In health, Medicare and Medicaid spending by hospital on standard procedures is sometimes publicly made available, but it is not a standard report which comes out, say, on a quarterly basis. Just publishing those numbers consistently could create a lot of pressure to be more effective. Of course given the elephant in the brain, no one directly involved feels motivated to do this. But that doesn’t mean it can’t happen.

        So to answer your question on how to make it possible, I’d suggest that in particular government spending on health be made public by hospital on standard procedures. And more generally publicize costs, which are easier to know, then the benefits, which are harder. But if we at least have a norm of publishing costs, especially for government spending, that could help a lot. People want to show they care, and if it’s blatantly obvious money is being poorly spent, that shifts incentives in the right direction.

        This is not an especially original idea, but I think it could be pushed harder than it has been.

  • lump1

    Signals need to be costly and hard to fake, but that still leaves a huge range of actions that are eligible to play the role of signals. Something narrows down that range to the types of signals that real people actually send. Conventions play an important role: We observe how others signal something desirable, realize that we too have the resources to also send such signals, and that’s how signaling conventions get rolling.

    If we want to get the most positive externalities from all that inevitable signaling, trend-setters would be smart to take account of all the things that people often want to signal, figure out all the eligible signals that would do the job (cost, unfakability) and identify the method with the best externalities.

    Academics aren’t the people with the social power to bend the conventions to something better, but maybe we can figure out which signaling regimes are worth promoting. Then the Oprahs will at least have some guidance about how to steer their zombie armies. So if you had Oprah’s ear, and you knew that signaling would not decrease but its execution was somewhat malleable, what would you ask her to promote? What currently used signaling strategies would it be a substitute for?

  • wiztwas

    Assuming this is true. Social media should be banned.

    • Or, taxed. Tax ads. That would be a tax on many kinds of social media, and a tax on another kind of signaling.

  • Mirabella

    You are assuming that the elites know more than the non elites. Recent political events would indicate this as a false premise.

    • Fortalez84

      Part of the way you become elite is by being good at the signalling game. Which means that elites on average are going to have a lot of crazy beliefs.

  • Dan Kelly

    All this seems to assumes that signalling is bad, that it should be discouraged or replaced with other activities that are somehow more real or productive. What if it isnt and it shouldnt?

    • Fortalez84

      Whether it’s bad or good, signalling is expensive, almost by definition. A cheap signal would lose its effectiveness as a signal absent some kind of collective norm.

      That said, I agree that signalling isn’t necessarily bad. If a cure for cancer is found, it’s likely to be by some researcher who is trying to impress other people with his intelligence, skill, and creativity. In a country that is trying to show its superiority in terms of quality scientific research.

      And it’s obviously dangerous to have some central authority deciding what signalling is good and what is counterproductive. In fact, part of the reason college is so expensive is the various loan programs passed by legislatures who want to demonstrate their support of education.

      • Joe

        I don’t think it’s true that signalling is expensive by definition. A signal just needs to be differentially expensive according to whether the individual does or doesn’t have the claimed attribute. The optimal signal is infinitely expensive (i.e. impossible) for those who don’t have the attribute, and infinitely cheap (i.e. free) for those who do.

      • Fortalez84

        In principle, I agree with you. But imagine a signal which approximates the ideal you describe. For example, consider my ATM card and PIN number. They are very cheap for me to obtain and remember; they demonstrate to the bank that I have a certain attribute (authority to withdraw money from my account); and they would be expensive for an imposter to obtain). A nearly ideal signal, and yet if someone saw me withdrawing money from my bank account , would they see it as a signal?

        I’m not sure, but I think the problem is my sloppy thinking (and perhaps that of others) about what a signal is.

  • Theresa Klein

    If you’re waiting for the intelligence of your audience to improve, you’re going to have a long wait.

  • Great post. I agree that it is key to improve the quality of the audience.

    People who explain sub-optimal outcomes by reference to signaling often leave their explanations incomplete. The mere fact that someone is trying to signal altruism, for instance, does not explain why so much altruism is ineffective. You also have to add another assumption, namely that the people you are trying to impress with your signals – the audience – either do not care about effectiveness, or are incapable of detecting it (or some combination of the two).

    This means that the audience is key both when it comes to explaining sub-optimal outcomes of signaling behavior, and when it comes to interventions that improve outcomes of such behavior.

  • David Condon

    “What if our audiences knew more about which medicines helped health, which charities helped the poor, which national policies help the nation, or which creative activities promoted innovation? That would push us to also know more, and lead us to choose more effective medicines, charities, policies, and innovations.”

    Yes, I agree this is a critical point.

    I’m not sure what to make of your third to last paragraph. By small vs large encounters, are you referring to the amount of time the audience devotes to attending to a particular signal? Is a more intense encounter the same as a larger encounter? Why is “many and small vs a few and large” a critical distinction?

    I like the way signal detection theory divides this topic up into four possible outcomes, so I’ll use that. You’re referring to situations when too many noise trials are presented, and it prevents the audience from responding when signal trials are presented. My impression is that your main goal is to increase the ratio of signal-to-noise, but then there’s this other argument you make with respect to making audiences more elite1:

    “The more that we expect to gain via impressing better-informed elites, the better informed will our show-off actions be.”

    Maybe, but it seems like that’s conditional upon a number of unstated claims. How often are better informed elites impressed by better informed actions? How reliably are audiences able to identify which elites are better informed?

    1 I presume making audiences more elite refers to increasing their hit probability on signal trials and their correct rejection probability on noise trials when the signal-to-noise ratio is held constant.

  • Robert Koslover

    Taxing Hollywood, in particular, sounds good to me. Or just repeal their existing tax cuts. See
    (Admittedly, Mr. Reynolds had a slightly different, yet not entirely different, objective in mind.)

  • Riothamus

    So we want to target more qualified elites with our signals, but we rely on elites to signal their qualifications. This seems to be a little circular, and I wonder if it could be smacked with something like PageRank. Scott Aaronson talked about using the idea for an eigenmorality here:

    If the signals we value can be put onto a directed graph, we could probably choose an elite audience effectively. If it works some elites may even have an incentive to optimize for it.