What is confidence for?
A general audience summary of our recent article in Nature Communications: “Confidence controls perceptual evidence accumulation” Tarryn Balsdon, Valentin Wyart, and Pascal Mamassian
For most of us, knowing what we see feels simple and automatic, but it actually involves a lot of complex processing within our brains.
Imagine you have a pile of puzzle pieces in front of you, face down and there is no box. The task is to guess the puzzle picture by turning over pieces. Each piece you turn is more evidence for guessing the picture, and you keep adding to this evidence until at some point you can guess the picture.
This process of collecting evidence is one way we think about vision, in the field of computational neuroscience this is called ‘evidence accumulation’. Like with the puzzle, our brains take pieces of evidence one by one until we can decide what we are seeing.
Through this understanding of how we see, we are then able to explain why we see some things better and also faster than others. For example, with the puzzle, some pieces you turn over may be more helpful, so you turn over fewer pieces and take less time to make an informed decision. For seeing, this can be a difference of milliseconds, but still a measurable difference.
Ultimately if you had the time, to know for certain you are correct about the puzzle image you would turn over all the puzzle pieces and put them together. But vision needs to happen quickly, so that we can interact with our environment. We adjust how much evidence we collect based on this time pressure. When we need to decide quickly, we turn over fewer puzzle pieces, speeding up our decision at the risk of being less accurate. The point at which we are ready to decide is called the ‘bound’, and in the lab we have shown that people are able to adjust this bound according to different time pressures, effectively turning over fewer puzzle pieces to make decisions faster.
How do we know if we have made the right decision? With the puzzle, you may look back at the pieces you have turned over. If you have a lot of good puzzle pieces, you should feel confident you made a good guess, but if you only have a few pieces, or even a lot of pieces with not much on them, then you should feel worried you may have guessed incorrectly. By looking back at the evidence, you develop a feeling of confidence about your decision.
People can readily report their feelings of confidence. In the lab, we often ask people to report their confidence on a rating scale, where 1 represents low confidence (or a random guess) and 4 high confidence (certain you were correct). People can be quite good at this, reporting low confidence more frequently for incorrect decisions, and high confidence more frequently for correct decisions. Which makes us think that this feeling of confidence might be useful for learning from our past mistakes and communicating our decisions to others.
In a recent experiment we discovered two interesting things about confidence that suggest that confidence plays a far greater role in our everyday lives. First, after deciding what they have seen people will use more evidence to decide their confidence than they used to decide on the picture. Whilst the visual decision is hasty, people will readily look for more evidence to work out if that decision was correct. So, people don’t just look back over the puzzle pieces they have already turned over, they can turn over more after they have made their decision.
Second, we found that people could use their feeling of confidence to decide when to decide. To go back to the puzzle metaphor, people use their feeling of confidence to decide whether they need to turn over more pieces, or if they could just guess the picture with what they have. In order to do this, confidence must be established throughout the process of collecting evidence. People aren’t making a decision and then looking back at how much evidence they have; they are assessing their confidence as they go.
These findings force us to rethink both how we see, and how we feel confident, in the field of cognitive neuroscience. Confidence doesn’t only reflect of a past decision, but confidence can influence how we make decisions, by informing us when we have enough evidence. Even if you had plenty of time, you wouldn’t waste your effort looking at more puzzle pieces if you already feel confident guessing the picture correctly. Confidence enable us to see more efficiently, and not just efficiently in terms of time pressure, but also the effort of processing all the information. Vision needs to be fast and effortless.
In this way, feelings of confidence play a large role in our everyday lives. Whilst here we are studying vision, we think that these findings about confidence could apply in all sorts of decision making, from whether you should add sugar to your tea to who to vote for.
