Knowledge and Confidence

Why does our confidence so often exceed our competence? And what can we do about it?

March 23, 2022

I recently read Julia Galef’s The Scout Mindset - an excellent book packed with practical wisdom. In one section, she credits our desire to feel certain as one of our biggest weaknesses in truth-seeking. When we are channeling this state of mind, we will use words like absolutely, obviously, highly likely, no doubt, and so on. Galef includes a trivia game where you answer questions and report your confidence in your answers as you go. I created an online version of this game - you can play it here. Based on the limited results of this mock experiment, people tended to be overconfident in their guesses. See Figure 1.

Figure 1 The black line represents perfect calibration (e.g., participants get 55% of the questions correctly when they are 55% confident). When the red line falls below the black, we are overconfident. When the red line is above the black, we are underconfident.

Interestingly, when you flip the axes and let confidence be the dependent variable, you get something similar to the Dunning-Kruger effect. Compare Figure 2 with Figure 3.

Figure 2 The data here correspond with Figure 1, but the axes are flipped for comparison to the famous Dunning-Kruger effect.

Figure 3 The Dunning-Kruger effect

The general trend of this experiment has been consistent. When participants find questions difficult, they undersell themselves and perform better than they thought. As the questions got easier, they were generally overconfident in their knowledge. How does this compare with our current understanding of confidence research?

A quick search on google scholar reveals overconfidence bias to be quite inconsistent. Some studies equate overconfidence to a social bias: “Overconfidence is induced by the desire to send positive signals to others about one’s own skill.” Many other studies find significant overconfidence despite the absence of social cues. Findings like these can be made compatible with a more nuanced approach to understanding overconfidence.

Confidence can be complicated

There may be more than one way to think about overconfidence. Many studies have found that when we find something difficult, we remain overconfident in our performance. This is also called overestimation. Conversely, when we find tasks easy, we tend to underestimate our performance. Together, these tendencies make up the hard-easy effect, which states that we overestimate our performance more on hard than easy tasks. Paradoxically, we judge ourselves as much less competent than others on hard tasks and much more competent on easy ones. Exaggerating one’s superiority over others is called overplacement.

The interesting thing about overplacement is that we tend to exaggerate our performance relative to others, which exacerbates the effect of over/under-placement. We often don’t have great information about what we know, especially in games like trivia. We have even less information about what other people know.

Trivia games such as the one I developed from The Scout Mindset, address another form of overconfidence, overprecision, which relates to the accuracy of one’s beliefs. This is the more stubborn side of overconfidence, which isn’t affected much by the difficulty of a task. When isolating our analysis to overprecision alone, we have to be careful not to let the other faces of overconfidence confound the results. For example, we may be overconfident because 1. we are exaggerating our competence, 2. we are exaggerating our superiority, or 3. we are exaggerating our certainty. To tease these apart in our trivia game, we need to assess more than the participant’s item-level confidence (confidence while answering each question). This looks only at overprecision. To measure overestimation, we should ask them to guess how many correct responses they got out of 40. To measure overplacement, we should ask the participant whether they think they did better or worse than average.

·      Overestimation – Wishful thinking about one’s abilities. We overestimate our performance most on difficult tasks.

·      Overplacement – Exaggerating one’s own superiority over others. We overplace ourselves most on easy tasks.

·      Overprecision – Exaggerating one’s certainty. Task difficulty doesn’t have an effect on overprecision.

In sum, our current understanding of the three types of overconfidence is that task ease affects overplacement and overestimation in opposite ways (e.g., “As overestimation goes up, overplacement goes down”), and heavily depends on the difficulty of the task. Our overprecision is more stubborn, and we seem to remain overconfident regardless of the task’s difficulty. So, if participants found the trivia game difficult, we could expect them to overestimate their competence, underplace their performance, and poorly assess their confidence on each question. Turns out that the Dunning-Kruger effect may ignore some important nuance.

Why are we ever overconfident?

Overconfidence seems purely maladaptive. Imagine an athlete who thinks she will perform well no matter what, so she continually skips practice; or a person who views himself as invulnerable, so he engages in reckless behavior. For overconfidence to evolve, it would have to be a better default than underconfidence. Our confidence influences our behavior too much for evolution to have missed it.

There are a few reasons why overconfidence might have come along for the ride. Perhaps most simply, it feels good to be confident. If you pride yourself on your scientific knowledge, it becomes difficult to admit what you don’t know. Overconfidence can also provide a source of increased well-being, so long as you don’t ostracize yourself by becoming too delusional and narcissistic. It also allows you to speak with more conviction, and thus makes you seem more persuasive.

Another interesting feature of our psychology is that we are averse to losses – most people wouldn’t stake $1000 on a 50/50 chance of winning another $1000, though the loss would be equivalent to the gain. Overconfidence may serve as a counterbalance to our loss aversion system, and vice versa. For if one is too loss averse, they will be paralyzed with fear of even unlikely losses; and if one is too confident, they are likely to expend all their resources even if the likelihood of a payout is low.

Before we fully embrace our overconfidence, consider again what we are missing out on by having our confidence well-calibrated to reality. We are making ourselves more susceptible to falsehoods. We are either underselling ourselves or basking in our arrogance. If we have ever made a display of our overconfidence to others, we are bound to disappoint.

How to overcome overconfidence

We now know that overconfidence is partly due to imperfect information about our own and others’ knowledge. It always helps to remain skeptical of your confidence in situations where we have little information about what we, and others, know.

You can also counteract overconfidence with your propensity toward loss aversion. If you frame your decisions in terms of hypothetical losses, you are likely to make wiser choices. Some studies have found that the effects of overconfidence are reduced in economic settings. A common way of using this hack is the ball bet, which Galef also mentions in her book. The ball bet allows us to quantify our confidence by asking: would you rather pick out of a bucket with four balls, one of which pays $10,000 and the others zilch, or would you rather stake $10,000 that you are making the right decision? If you would rather take the ball bet, then you must be less than 25% confident in your answer.

Another, perhaps simpler, way of avoiding optimistic biases is to ask yourself how you would advise a friend. This allows us to separate our decisions from our identities. We get excited when situations address our identities, but this is precisely when pride and reputation biases are most likely to creep in. Lastly, studies have shown that our overconfidence can be diminished by considering all possibilities of an event. If you are confronted with a choice that will have a considerable impact on your life, resist the temptation to base your decision on wishful thinking, identity preservation, “following your gut,” and social pressure. It is almost always better to use your head and rely on careful deliberation.

Though tactics like these aren’t silver-bullets, they are better than nothing. If anything else, it helps to know that our confidence is context-dependent and often ill-calibrated with reality. A common thread between the remedies for overconfidence is healthy self-cynicism and critical thinking. A basic understanding of these facets of our confidence equips us to be less judgmental, less likely to fool ourselves, and more likely to make better decisions when they count.

References

Burks, Stephen V. and Carpenter, Jeffrey P. and Goette, Lorenz F. and Rustichini, Aldo, Overconfidence is a Social Signaling Bias. IZA Discussion Paper No. 4840, Available at SSRN: https://ssrn.com/abstract=1575893 or http://dx.doi.org/10.2139/ssrn.1575893

Burson, K.A., Faro, D., Rottenstreich, Y., ABCs of principal–agent interactions: Accurate predictions, biased processes, and contrasts between working and delegating, Organizational Behavior and Human Decision Processes, Volume 113, Issue 1, 2010, Pages 1-12, ISSN 0749-5978, https://doi.org/10.1016/j.obhdp.2010.05.002.

Erev, I., Wallsten, T. S., & Budescu, D. V. (1994). Simultaneous over- and underconfidence: The role of error in judgment processes. Psychological Review, 101(3), 519–527.

Galef, Julia. The Scout Mindset: See Things Clearly and Make Smarter Decisions, Portfolio Penguin, London, 2020, pp. 75–82.

Haran, U., Moore, D. A., & Morewedge, C. K. (2010). A simple remedy for overprecision in judgment. Judgment and decision making, 5(7), 467–476.

Kahneman, D., & Lovallo, D. (1993). Timid choices and bold forecasts: A cognitive perspective on risk and risk taking. Management Science, 39, 17–31.

Moore, DA, Schatz, D. The three faces of overconfidence. Soc Personal Psychol Compass. 2017; 11:e12331. https://doi.org/10.1111/spc3.12331

Moore, D. A., & Healy, P. J. (2008). The trouble with overconfidence. Psychological Review, 115(2), 502–517. https://doi.org/10.1037/0033-295X.115.2.502

Moore, D.A., Carter, A.B., Yang, H.H.J., Wide of the mark: Evidence on the underlying causes of overprecision in judgment, Organizational Behavior and Human Decision Processes, Volume 131, 2015, Pages 110-120, ISSN 0749-5978, https://doi.org/10.1016/j.obhdp.2015.09.003.

Proeger, T., Meub, L., Overconfidence as a social bias: Experimental evidence, Economics Letters, Volume 122, Issue 2, 2014, Pages 203-207, ISSN 0165-1765, https://doi.org/10.1016/j.econlet.2013.11.027.