In this piece, by Jérémie Beucler and Wim De Neys, they revisit the received view of dual-process theory and some of its simplifying assumptions. Drawing on recent advances in reasoning research, they argue that the classical roles assigned to intuition and deliberation are more complex than they first appear, as intuition can sometimes be sound, while more deliberation does not always help.
Sometimes thinking is easy, such as when we are asked to multiply 3 by 4; at other times, it is hard, such as when we are asked to multiply 34 by 27. This simple contrast captures the core intuition behind dual-process theories: human thought can take two broad forms, one fast, the other slower. These theories have become one of the most influential frameworks for explaining reasoning and decision-making. They have been especially influential in explaining reasoning biases, where people systematically deviate from normative standards such as logic or probability principles. Originating in research on heuristics and biases by Kahneman and Tversky, as well as in early reasoning research by scholars such as Wason and Evans, these theories distinguish between fast, effortless intuitive processes and effortful, deliberative processes. These two processes were later labelled “System 1” and “System 2” by Keith Stanovich, labels subsequently popularised by Kahneman. However, their success also means that they have sometimes been oversimplified and have lagged somewhat behind the advances made in reasoning research in recent years, which we will introduce here.
A tale of two systems
Dual-process theory is often thought to imply a normative hierarchy between the two systems: ‘System 1’ is treated as the source of reasoning biases, whereas ‘System 2’ is assigned the role of correcting them. Consider, for instance, Nudge, by Nobel laureate Richard Thaler and Cass Sunstein, which helped popularise behavioural public policy. Early in the book, they distinguish between “two kinds of thinking, one that is intuitive and automatic, and another that is reflective and rational.” This framing implicitly portrays intuition as error-prone and deliberation as corrective.
Why, then, are we biased? The answer is simple: humans are “cognitive misers” who seek to minimise mental effort, and therefore seldom engage ‘System 2’ to override their faulty intuitions, resulting in irrational reasoning and poor decision-making. A logical consequence of this view is that reasoning biases could be corrected simply by making people deliberate more. This has important practical implications: intuitive thinking should be minimised and deliberation encouraged, especially when the stakes are high and mistakes may be costly, as in major life choices, such as getting married, changing careers, or giving up coffee.
However, this popular reception of dual-process theories is a simplification, and sometimes differs substantially from what dual-process theorists themselves have proposed. What is more, research in the reasoning field over the past twenty years has shown that the reality may be more complex than it seems.
Deliberation is not a silver bullet
As a start, deliberating more is often of little help. If biased responding were simply due to a lack of deliberation, then more deliberation should do the trick. Yet instructing participants to slow down to “stop and think” provides only weak performance gains. Paying them to give the correct response, even with very high amounts, does not help much, and can sometimes hurt depending on the type of reasoning problem. Even worse, deliberation appears to be least effective precisely when it is most needed. In a large meta-analysis, we recently showed that thinking longer did not substantially increase error correction among highly biased reasoners (Figure 1). In other words, the reasoners most prone to biased responses were also the least likely to benefit from additional time spent thinking. This makes deliberation not only an unreliable corrective tool, but also a potentially costly waste of cognitive effort for (some) reasoners. Deciding whether to deliberate is thus something of a gamble, as the extra effort does not always pay off.
Figure 1. Probability of correcting an initial biased response as a function of deliberation time and bias level. Longer deliberation mostly benefits reasoners with lower levels of bias. Adapted from Beucler et al. (2026).
Intuitions can be smart
Beyond showing that deliberation is not a silver bullet, a large body of research suggests that intuitions can often be correct from the start. First, biased reasoners who give an incorrect response to a reasoning problem are often sensitive to their mistakes, at least implicitly. Consider the following question: “How many animals of each kind did Moses take on the Ark?” Most people answer “Two,” failing to notice that it was Noah, not Moses, who took the animals onto the Ark. In a recent paper, we showed that people who fell for this illusion were not completely blind to their mistake: even when they answered “Two,” they were less confident than when answering ordinary questions correctly. Importantly, this drop in confidence appeared even at the intuitive stage, when deliberation was minimised using a concurrent cognitive load and a strict response deadline. Such intuitive sensitivity to the error suggests that the intuitive response itself may partially integrate the correct answer. This phenomenon, where biased reasoners often display implicit sensitivity to the conflict between the intuitive, tempting response and the competing correct response, is known as conflict detection.
On top of these conflict-detection findings, research also shows that intuitions can often be correct. Most evidence for this comes from the two-response paradigm, in which people first give an initial response under time pressure and cognitive load to minimise deliberation and then give a final response after being allowed to deliberate (see Figure 2 for an example). Strikingly, across many classic reasoning problems, good reasoners often give the correct answer already at the initial response stage. In other words, when intuitions are measured before deliberation has had much time to intervene, they are not always biased or erroneous: under the right conditions, they can be “sound” from the start. For instance, in a recent paper, we used the two-response paradigm on Compound Remote Associates problems, where people must find a word linking three cues (for example, “cottage, swiss, cake” → “cheese”). When people solved these problems, they had already found the correct answer at the initial response stage in more than 70% of cases.
Figure 2. Time course of a two-response paradigm trial.
All of these findings suggest that System 1 can do more than previously thought. On top of faulty intuitions which sometimes lead us astray in reasoning, it can also possess, to some extent, sound intuitions. But where do these come from then? The current hypothesis is that they have been automated through practice and education. Think of chess experts, for instance. When faced with a chessboard, they can instantly relate it to previous plays or moves they studied. This pattern may hold in reasoning too.
The “smart intuitor” idea
At the level of reasoners, it appears that smarter people may also have better intuitions. Good thinking may begin before deliberation even starts. This is also what we found in our study on the Compound Remote Associates, where we estimated the semantic networks (think of it as the organisation of concepts in your mind) of our participants using computational modelling techniques. As shown in Figure 3, participants with better intuitions had denser and more interconnected semantic networks. Their concepts were closer and better connected, enabling activation to spread more easily between them – effectively making it easier for them to find the solution without further reflection.
Figure 3. Semantic networks of high intuitor participants (left) and low intuitor participants (right). One dot corresponds to one concept. (Adapted from Beucler and De Neys, 2026)
Cognitive gamblers
Where does this leave us? With a more nuanced view of dual-process theories. Intuitions can be smart under certain conditions, and deliberation will not always improve decisions. Reasoners may therefore be better described as cognitive gamblers than cognitive misers: the challenge is not simply to think more, but to know when deliberation is worth the effort, given one’s intuitive capacities and the problem at hand.
This also changes what we should aim for: not simply encouraging people to deliberate more, but helping them develop and organise the knowledge and competencies that support better intuitive judgements. This is close to what behavioural scientists call a “boost”: an intervention that strengthens people’s decision-making competencies, for example by providing simple decision rules, or reinforcing basic principles of probability and logic until applying them becomes more intuitive. It’s time to branch out from ‘thinking slow’ and embrace new understanding.
This post draws on the following papers, should you want to read more on the topic:
Beucler, J., Purcell, Z. A., Dr, Charles, L., De Neys, W., & Desender, K. (2026). Thinking in Vain: An Evidence Accumulation Account of Biased Reasoning. Retrieved from osf.io/preprints/psyarxiv/ze2ku_v2
Beucler, J., Voudouri, A., & De Neys, W. (2025). Moses Illusions, Fast and Slow. Journal of Experimental Psychology: Learning, Memory, and Cognition
Beucler, J., & De Neys, W. (2026). Intuitive Insight: Fast Associative Processes Drive Sound Creative Thinking. Cognition, 271, 106422.
Jérémie Beucler is a PhD candidate in Cognitive Science at LaPsyDÉ (CNRS & Université Paris Cité), supervised by Wim De Neys, Zoe Purcell and Lucie Charles. He also collaborates with Kobe Desender. He investigates the cognitive mechanisms underlying human reasoning, with a particular focus on reasoning biases using behavioural experiments, computational modelling, and large language models.
Wim De Neys is a cognitive psychologist and CNRS research professor at the University of Paris. He studies how people reason and make decisions. His research focuses on the interplay between intuition and deliberation, with particular attention to reasoning bias and intelligence. He combines experimental, developmental and neuroimaging approaches to understand when intuitive thinking goes wrong and right – and what this reveals about human (and artificial) rationality.