Metacognition in human decision-making: confidence and error monitoring.
Yeung N, Summerfield C. · 2012
Read Metacognition in human decision-making: confidence and error monitoring. by Yeung N, Summerfield C. (2012) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc3318764.txt.
Introduction
Yeung and Summerfield reviewed how people judge their own decisions, in a 2012 article on the science of self monitoring. People often sense their confidence in a choice and notice their own mistakes, even with nobody telling them the right answer. This self knowledge, called metacognition, helps people avoid repeating errors and stops them wasting effort on shaky choices.
The review brings together two research lines that grew up apart, one on confidence judgments and one on error spotting. Both, the authors argue, rest on extra processing that happens after a decision, inside the same systems that made it. That shared machinery explains why confidence and accuracy usually rise and fall together. The authors note that nearly all the evidence comes from quick, simple lab choices, and question how well the ideas stretch to long, real world decisions.
Catalog record
| CREATOR | Yeung N, Summerfield C. |
|---|---|
| YEAR | 2012 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 70 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Metacognition in human decision-making: confidence and error monitoring By Nick Yeung, Christopher Summerfield Published in: Philosophical Transactions of the Royal Society B: Biological Sciences DOI: 10.1098/rstb.2011.0416 ABSTRACT Abstract People are capable of robust evaluations of their decisions: they are often aware of their mistakes even without explicit feedback, and report levels of confidence in their decisions that correlate with objective performance. These metacognitive abilities help people to avoid making the same mistakes twice, and to avoid overcommitting time or resources to decisions that are based on unreliable evidence. In this review, we consider progress in characterizing the neural and mechanistic basis of these related aspects of metacognition—confidence judgements and error monitoring—and identify crucial points of convergence between methods and theories in the two fields. This convergence suggests that common principles govern metacognitive judgements of confidence and accuracy; in particular, a shared reliance on post-decisional processing within the systems responsible for the initial decision. However, research in both fields has focused rather narrowly on simple, discrete decisions—reflecting the correspondingly restricted focus of current models of the decision process itself—raising doubts about the degree to which discovered principles will scale up to explain metacognitive evaluation of real-world decisions and actions that are fluid, temporally extended, and embedded in the broader context of evolving behavioural goals. Keywords: decision-making, confidence, metacognition, error monitoring INTRODUCTION Metacognition in perceptual choice (a). Imagine yourself cycling along a narrow, winding country lane on a summer's day. Successfully negotiating each twist and turn in the road requires the interpretation of a variety of subtle sensory cues and their conversion into appropriate motor commands. For example, incoming visual signals that reveal the curvature of the road, or motion parallax signals arising from the trees beyond, allows you to gently adjust the handlebars towards the left or right to steer the bicycle smoothly round each bend. Over past decades, psychologists and neuroscientists have devoted substantial effort to understand the neural and computational mechanisms by which actions are selected on the basis of a stream of incoming sensory information [1]. Because this sort of sensorimotor control often requires an observer to commit to one discrete action from among several possible candidates, this literature has been informed by computational models that offer a formal account of how categorical decisions are made. These models, described in more detail below, have provided an excellent account of the decisions and decision latencies exhibited by observers selecting actions on the basis of ambiguous sensory information [2–5]. However, now imagine that as you are cycling, the sky darkens and it begins to rain—visibility is reduced, and the road becomes slippery and wet. The approach to each bend still requires a deft angling of handlebars, but now you are less certain about whether each chosen action is optimized to steer the bicycle in a fashion that appropriately meets the gradient of the curve and the camber of the road. Action selection proceeds as before, but you are unsure whether the actions chosen are the most appropriate ones. In fact, you might even experience an immediate sense that you have just made a poor choice. In short, you are less confident about your decisions. Curiously, despite universal agreement that an accompanying sense of confidence is a subjectively salient property of almost all our decisions, there is currently little consensus about how we might incorporate decision confidence into formal models of choice behaviour or explore its biological substrates. Fundamental questions remain unanswered. For example, is the information that gives rise to the ‘second-order’ estimate of confidence in a choice identical to that determining the ‘first-order’ choice itself? Why are we generally more sure that we are correct than that we have made an error, even for difficult choices? Why do we sometimes appear to ‘change our mind’ after a motor programme has been initiated? Are these changes of mind necessarily accompanied by awareness that the initial choice was incorrect? In what follows, we review the literature that has posed these and related questions about decision confidence in perceptual choice tasks. Subsequently, we highlight links between this work and a literature that has considered how people monitor whether they have made an error under conditions of uncertainty or conflict. Finally, we propose some potentially fruitful avenues for research that draw upon common themes in these two literatures, building on their shared strengths and addressing their shared limitations. Formal models of perceptual choice (b). Even under good viewing conditions, visual information is corrupted by multiple sources of noise and uncertainty, arising both in the external world and in the dynamics of neural processing. One sensible way to increase the signal-to-noise ratio of visual information is to sample the external world repeatedly and integrate this information over time, making a decision only when the information is considered to be of sufficient quality [5]. This idea forms the basis of a class of model in which binary choices are described via an accumulation-to-bound mechanism, with successive samples of information totted up until they reach a criterial level or ‘bound’, upon which a commitment to action is made (figure 1). [Figure: The drift-diffusion model. Accumulating evidence (the decision variable, y-axis) over time (x-axis) is shown for two illustrative trials (marked a and b, grey and black lines), one on which the choice θ is made and the other on which the choice –θ is made. A decision is triggered when evidence reaches θ or –θ. Grey line, trial 1; black line, trial 2.] Different versions of this model make divergent assumptions about exactly what quantity is integrated en route to a decision—i.e. about precisely how the ‘decision variable’ (DV) is composed. To simplify matters, in this review, we focus on one variant of this model—the ‘drift-diffusion’ model, or DDM—in which the DV corresponds to the relative likelihood of the two choices being correct, given the stimulus [3]. In the DDM, the DV on sample vt is updated on each sample t with an increment composed of two quantities: δ, a linear drift term that encodes the rate of evidence accumulation, and cW, Gaussian noise with a mean of zero and a variance of c2: [formula: 1.1] Decisions are made when the DV exceeds a fixed deviation from zero, θ, such that during evidence accumulation: [formula: 1.2] This simple model has much to recommend it both as a normative and descriptive account of categorical choice. The DDM has been widely and very successfully applied to decision-making in a range of cognitive domains—from low-level perceptual decisions to retrieval of facts from semantic memory, to economic decision-making under uncertainty [6–8]—and accounts neatly for several empirical characteristics of reaction times (RTs) observed for binary choices in these tasks. Firstly, the natural geometry of the accumulation-to-bound process predicts the observed rightward skew in RT distributions. Secondly, varying the bound offers an elegant account of the economy of speed and accuracy that characterizes mental chronometry tasks. Thirdly, by allowing drift rate and origin of accumulation to vary across trials, the model accounts for the relative RT of correct and error trials under conditions where speed or accuracy are emphasized [8]. Moreover, the model implements a sequential probability ratio procedure that optimizes speed for a given error rate, and thus takes its place