Be warned: The troubling expectation of regret

Many actions are hardly better than leaps in the dark, and we are only too familiar with the experience of regret when we realize that we made the wrong decision (see Frith and Metzinger, 2016). [Regret is not to be confused with disappointment. We feel disappointment when the outcome of a decision is not as good as we expected. We feel regret when we realize that, had we made a different choice, we would have achieved a better outcome.] This experience is explicit, as we are fully aware of it and can talk about it. But what value does it have for us? It always seems to be too late to do anything about it. But is it? Regret might first appear as a shadow in the twilight zone, and you may have to learn to recognize it. It is a shadow that can warn us before we commit to a decision. Once we have learned to heed this signal, we can anticipate the regret that we would feel if it turned out that we made the wrong decision.

Learning about this shadow can be quite rapid, as was demonstrated in an experiment with children aged six to seven years using a variant of the classic marshmallow test (McCormack et al., 2019), in which the children had to decide whether to wait for a short delay to win two sweets, or a longer one to win four. Those who chose the short delay were shown that they would have won more if they had waited longer and then were asked whether they regretted not waiting longer. The next day, the children were presented with the same choice again. Only those children who said they regretted choosing the short delay on day one changed their behavior on day two and tolerated the longer delay. Presumably, this was the effect of the anticipated regret that was invoked by the experimenter's question.

If you ever put in a bid at an auction, you will know that bidding is a rich source for feelings of anticipated regret. Imagine yourself bidding £100 for a rare book that you would dearly love to add to your collection. Your bid did not succeed, and someone else managed to get the book. You immediately wonder how much it went for. You will groan if you find that it went for £105. Next time, you will likely bid £110, or maybe would otherwise. even £120, for a similar book. In contrast, if the winning bid was £500, you will probably not bid at all for another one like it next time. The coveted book is clearly out of your price range.

There is another type of auction, in which you are not told the value of the winning bid. Here, there is less scope for anticipated regret, and you tend to bid less than you would otherwise. This is not just speculation. Experiments have shown that there is a clear difference: People tend to make higher bids in auctions where they know that they will be told the winning bid. They can factor in the painful regret that they might feel if they lost out by a small margin (Filiz-Ozbay and Ozbay, 2007). No wonder that most auction houses advertise in advance that they will communicate the winning bids as soon as they are known.

A toe in the water: The vexing feeling of uncertainty

We make decisions under uncertainty all the time. Was our last decision a good one? Will anybody tell us? It turns out that we can be guided to make our next decision even without feedback, via signals from the depths (Folke et al., 2016). For example, a feeling of high uncertainty creates low confidence. This feeling makes it very likely that you will change your mind when the same choice is presented again. People who are responsive to these signals are generally more accurate with regard to matching up their subjective uncertainty and their actual behavior. They are also better at knowing which task they should take on when given a choice. They can judge in advance whether they will be able to do the task well (Carlebach and Yeung, 2020).

What is the effect of feeling very uncertain? After making an error, we slow down a little, and are not even aware of it, as in the case of the typists. Slowing down gives us a chance to collect more evidence – that is, alter our speed/accuracy trade-off (van den Berg et al., 2016). In an experiment that involved making a decision between two essentially similar stimuli, the uncertainty was artificially manipulated by presenting variable evidence: that is, the stimuli were of different strengths and reliabilities. In this experiment, the participants tended to seek more information – they asked to see a stimulus again – even when their accuracy was high (Desender, Boldt, and Yeung, 2018).

This finding suggests that it is the subjective feeling of certainty, or confidence, more than the objective level of accuracy, that guides our behavior. Of course, in everyday tasks, the feeling of uncertainty tracks the accuracy of our performance. Hence, it is a good idea to seek more information when we are uncertain. When making difficult decisions, it is a good idea to confer with others when possible. In these cases, we are automatically drawn to taking advice from people who display more confidence. We say more on this topic in chapter 15.

One valuable effect of signals from the deep that feed our feelings of certainty (or rather uncertainty) is that they help us learn a new skill, even when we get no feedback about whether we were correct or not. For this reason, we can learn a difficult perceptual skill (in the lab, this often means detecting the orientation of stripes embedded in noise) through repeated presentations without ever being told whether we are right or wrong (e.g., McKee and Westhe, 1978; see also Zylberberg, Wolpert, and Shadlen, 2018). It is the feeling of certainty that shows us when we are on the right track, even if nobody is there to tell us.

This learning does not happen by magic. In fact, it is the feeling that bolsters our confidence, and that provides the necessary feedback. We note that this feedback is internal and based on translating a valuable message that monitors signals sent up from the depths of the information-processing system. The level of subjective certainty or uncertainty that is associated with each trial provides the signal that indicates whether a response was probably correct. Just as with external feedback, which may be derived from reward, regions deep in the brain (the nucleus accumbens and ventral tegmental area; see figure 12.2) encode prediction errors that are derived from this feeling of confidence (Guggenmos et al., 2016).

The relationship between the feeling of confidence and actual competence is fragile and can be biased toward either underconfidence or overconfidence. We will revisit this problem in the next chapter. For those who want to know how exactly metacognitive confidence is measured, we say the magic words "signal detection theory (STD)" (Swets, Tanner, and Birdsall, 1961). This method makes a distinction between sensitivity and bias. Bias here refers to being either overconfident or underconfident in our decisions. In contrast, sensitivity refers to how good we are at predicting whether our decisions will be good or bad. Do we know, or are we just guessing? Our sensitivity to the likely outcome of our decisions is an aspect of metacognition and is independent of our general state of confidence (Fleming, 2021). It is also independent of our ability to make decisions. In theory, you could be a good decision-maker but have no idea which decisions are going to be good or bad or why. Such a person would be difficult to work with, despite their ability, because we could not rely on their confidence reports.

- Find out more about the book.