Our findings differ considerably from those reported by Dijksterhuis et al. (2006, 2009). In experts, conscious deliberation was superior to unconscious deliberation and in line with our hypothesis the effect was strongest for complex cases. Dijksterhuis and colleagues found conscious deliberation to be poorer for complex cases. We found the opposite. For simple cases, reasoning mode did not affect experts’ decisions, again differently from their findings. Among students, mode of reasoning did not make a difference, as long as the participants were required to delay their response. Interestingly, among students and with complex cases, performance became poorer when allowed think about the decision. For simple cases, we found a deliberation-without-attention effect, but only when that treatment was compared with the immediate-decision condition.
We offer here an alternative explanation for both our findings and those of Dijksterhuis and colleagues. We distinguish between two cognitive processes that may underlie diagnostic decision-making. The first is a fast pattern-recognition process in which the cues of a case are associatively matched against a pattern in memory (Moors & De Houwer, 2006; Norman & Brooks, 1997; Schmidt & Boshuizen, 1993). The second is a more elaborate, knowledge-based reasoning process, in which problem-solvers consciously activate relevant knowledge until the problem is understood in terms of its underlying structure (Rikers, Schmidt & Boshuizen, 2002; Stanovich & West, 2000). This distinction between pattern-recognition processes and analytical reasoning is in line with dual-process theories of reasoning and judgment (Evans, 2008; Sloman, 1996). In addition, we argue that a real difficulty often involved in diagnostic reasoning is to overcome the influence of salient, seductive cues that may lead one down the garden path. An example may clarify this. Most physicians, confronted with a patient having a history of coronary disease and high cholesterol levels, with severe pain behind the chest bone, would immediately consider a diagnosis of acute myocardial infarction. This fast, similarity-based, diagnostic approach leads to accurate decisions in routine situations and is viewed as an efficient strategy developed with experience (Norman, 2005; Norman & Eva, 2010; Schmidt & Rikers, 2007). Nevertheless, this approach may also fail (Elstein, 2009; Graber, Franklin & Gordon, 2005). When this initial diagnosis is wrong, it can only be “repaired” by further conscious analysis of the other, less salient features of the case, leading to the activation of alternative hypotheses and, eventually, recognition of the correct diagnosis. This process of conscious thinking about the case often requires the mobilization of deep, causal knowledge that helps integrate the salient cues with the less obvious ones into a new and better diagnosis, or to discard the salient cues as irrelevant. Doctors often have trouble overcoming the seduction of salient cues, for instance because they use shortcuts to quick decisions or because of premature closure, making it the number-one reason for diagnostic error (Graber et al., 2005; Klein, 2005; Redelmeier, 2005). How do these considerations relate to our experiment?
We assume that responses to the immediate-decision condition were dominated by pattern recognition, whereas the conscious-thinking condition induced the more analytical reasoning process. Let’s first look at the effect of conscious, analytical reasoning on complex cases. Among doctors, the activation of additional relevant knowledge led to a sizable effect over the immediate-decision condition. In fact, consciously thinking about the problem led to a 50% gain in diagnostic accuracy! The validity of this effect of overcoming initial bias was further strengthened because we were able to show that, within the conscious-thinking condition, doctors were very well able to recover from an initially wrong diagnosis as witnessed by the difference between initial and final diagnosis. In line with the so-called default-interventionist form of dual process theories, a rapid intuitive judgment was apparently repaired by effortful analytical reasoning (Evans, 2008). Among students, conscious reasoning had a reverse effect. Their initial diagnosis, although already poor, became worse when they were allowed to think consciously about it and considered alternatives. Since these students had little and largely incorrect knowledge about these complex cases to begin with (as witnessed by their already poor performance in the immediate-decision condition), this suggests that the additional activation of that little knowledge apparently only confused them further by increasing the number of incorrect diagnoses to choose from.
Consciously deliberating about the simple cases displayed a very different pattern. Since the fast, pattern-recognition response in doctors already generated a correct diagnosis in most cases, nothing was gained by further analysis. Hence no differences arose. However, students profited from further analysis, suggesting that sufficient knowledge was available to diagnose these easy cases but this knowledge had to be activated first in order to help students overcome initial bias produced by salient cues, similar to what happened to the doctors confronted with complex cases.
What role is there to play for deliberation-without-attention in diagnostic decision-making? A limited one, it seems. In most conditions, there was no significant difference with the immediate-decision response, similarly to what has been shown in other recent studies (Newel, Wong, Cheung & Rakow, 2009; Rey, Goldstein & Perruchet, 2009), suggesting that unconscious thinking did not really facilitate problem solving. There seems to be, nevertheless, instances in which it helps. In our study, only in students confronted with easy cases there was an effect of deliberation-without-attention. Based on Dijksterhuis et al.’s (2006) assumption that unconscious thinking allows for an associative process in which all cues are weighted equally, one may assume the following: the students had sufficient knowledge to solve these simple cases, but were initially seduced to some extent by some salient (and irrelevant) cues. The associative process postulated by Dijksterhuis and colleagues may have weakened the influence of the salient cues, allowing other cues to equally influence the diagnostic decision. Hence a better performance.
Our explanation for the findings is to some extent speculative, as further research is required to provide additional information on the nature of errors committed under the various conditions. We investigated diagnostic decisions under experimental conditions, which may restrict generalization to real settings and to other types of decisions. The conscious thought condition in our study comprised an elaborate, structured approach to analyze case features, and it may be argued that this helps more than the deliberation-with-attention condition in Dijkterhuis et al.’s (2006, 2009) studies. On the other hand, it allowed checking whether participants were in fact thinking about the case. Furthermore, one may reasonably argue that the procedure closely replicates how doctors reason while thoroughly analyzing a patient’s problem.
A more accurate picture of decision-making seems to be drawn, when expertise is taken into account and problems replicate real situations encountered in professional life in complex domains. In line with recent studies (Acker, 2008; Lassiter, Lindberg, González-Vallejo, Bellezza & Phillips, 2009), our findings cast doubts on the statement that unconscious thought is the superior mode of making complex decisions. Experts confronted with complex problems can make better decisions after consciously thinking about their choices, which is a conclusion with significant implications for (clinical) practice. If one does not have knowledge in a domain, then relying on immediate decisions would be a wiser choice for solving complex problems. Deliberation-without-attention apparently only helps beginners faced with simple problems.