Conscious versus unconscious thinking in the medical domain: the deliberation-without-attention effect examined

Introduction

Medical knowledge, intelligence, and reflective reasoning [1, 2] are considered to be the cornerstones of a doctor’s daily work. Yet where critical decisions are to be taken, many doctors are aware of the importance of ‘gut feelings’ and ‘educated intuition’ [3, 4], i.e. unconscious reasoning that may guide behaviour more than conscious knowledge does [5–8]. It is important to distinguish two ways in which the notion of ‘unconscious’ processes has arisen in the literature. The first is a contrast with conscious analytical processing as a component of a dual process model of reasoning [9] and is synonymous with non-analytical or so-called System 1 thinking. The ‘gut feeling’ alluded to above likely reflects this process, which occurs concurrent with or even ahead of analytical reasoning. A very different kind of unconscious thought arises from studies of the deliberation without attention (DWA) effect, wherein individuals may even profit from unconscious thought [10]—i.e. ‘active thought processes that take place while attention is directed elsewhere’ [11] occurring hours or even days after the event—and take better decisions than after conscious thought. [12–17] This effect [12] has, however, been challenged [18–25] (see Acker [26] for a meta-analysis and Baumeister et al. [27] for a review) and the findings are not uniformly supportive. In particular, in the medical domain, Mamede et al. [24] recently found that conscious thought resulted in better decisions with complex tasks and experts, contradicting the DWA effect, although it provided no advantage over a DWA condition with students. It is not clear why the effect of DWA would have different effects at different levels of expertise. Conversely, De Vries et al. [15] reported more correct diagnoses of psychiatric case vignettes after unconscious processing than after conscious processing.

Studies showing the DWA effect [12–17] have mainly focused on decisions that were of importance to the decision maker, such as ‘what is the best car to buy?’ (although often in hypothetical experimental situations). But, does the effect also occur when decisions are made that are important for others, which is daily practice for doctors? In particular, decisions related to prognosis and therapy need careful consideration, and this is exactly what one may expect from experienced physicians. If, in either Simple or Complex situations, delaying a decision to allow time for unconscious processes to intervene is superior to conscious deliberation, physicians would be advised to postpone a decision and ‘sleep on it.’ This would in many ways be at variance with common sense and daily practice. The claims that originate in studies that supported the DWA effect have, however, attained a large audience [e.g. 14–17, 28, 29], so we decided to test if these claims hold true for individuals who make far-reaching decisions daily, i.e. physicians. Likewise, if a DWA effect is present in clinical doctors, the medical educational system must incorporate its repercussions.

In the present study, we confronted physicians and medical students with fictitious patients having objectively determined different life expectancies. Medically relevant characteristics were provided in either a Simple or a Complex task, closely following the design of Dijksterhuis et al. [12]. We aimed to simulate clinical practice where doctors receive a host of information for their patients and then must decide on whom to take action most urgently. Our participants assessed the patients’ life expectancies after either conscious or unconscious thought. We included level of expertise because it is at the basis of optimal decision-making and has been largely ignored in most previous DWA studies. Given our previous findings [24] we expected that experienced physicians, as a result of their large knowledge base built up over the years, are not affected in their performance by mode of thought. In particular, we were interested to find out if a DWA effect occurred in the complex task, because this most closely resembled the task where positive effects had been demonstrated [12].

Methods

Results

For the 57 novices, years of experience was ‘none’ for 55, and ‘a few months or less’ for 2 (one Simple/UNC, one Complex/UNC). For the 86 experts, i.e. residents or consultants with at least some experience, the levels of experience (‘a few months or less’, 6–12 months, 1–5, >5 years) were distributed as follows over the four conditions: Simple/CSC (N = 1, 3, 5, 5, respectively), Simple/UNC (N = 1, 2, 7, 3, respectively), Complex/CSC (N = 1, 2, 13, 12, respectively), Complex/UNC (N = 1, 2, 17, 11, respectively).

As a check upon our manipulation in the Simple and Complex versions, we compared mean ranks for percentages given for patients Lo, Med1, Med2, and Hi, respectively. With minor exceptions for novices in both Complex conditions and for experts in the Simple/UNC condition, mean ranks were in line with our manipulation: lowest for patient Lo, highest for patient Hi, and in between for the two other patients. We considered these results to be confirmation that our manipulation had created fictitious patients that clearly differed in life expectancy in the intended direction (Table 3).

In an initial analysis of the Rho coefficients, the difference in mean scores between experts and novices (Expert, M = 0.44, Novice, M = 0.26) did not quite reach statistical significance (F(1,142) = 3.20, p = 0.08). However, a simplified analysis omitting Mode of thought and Condition did result in a significant difference (p = 0.04). We found no significant main or interaction effects for any of the other factors in the ANOVA (Mode of thought: F(1,142) = 0.23, p = 0.64; Condition: F(1,142) = 0.15, p = 0.70; interactions: all ps > 0.1). When focusing on the DWA effect, we did not find a significant main effect of DWA (F(1,142) = 0.23, p = 0.64 for all cases, F(1,86) = 0.19, p = 0.67 for complex cases). Average rank correlations for all conditions are presented in Table 4.

Discussion

The present study investigated task performance in medical doctors and students in evaluating fictitious patients with different life expectancies. Medically relevant characteristics were provided in either a Simple or a Complex task, closely following the design of Dijksterhuis et al. [12]. We expected doctors not to be influenced by mode of thought (i.e. conscious vs. unconscious). As predicted, the deliberation-without-attention effect was absent. Contrary to what this effect predicts, task performance after unconscious thought was not better than after conscious thought [12–17].

We verified that our fictitious patients indeed differed in the assumed direction: both the objective statements by the two board-certified internists and the mean ranks over all participants (Table 3) supported our intended manipulation. Compared with Studies 1 and 2 by Dijksterhuis et al. [12], in which such verification was lacking, our double manipulation check can be considered a strength of our study. Also, our participants had a specific expertise in the domain in which they were tested and the chosen age and gender of our fictitious patients made assessments of the life expectancies relevant. Lastly, we carefully selected characteristics that were objectively relevant for a patient’s life expectancy, which resulted in combinations of characteristics that could go together well in real patients.

It would be interesting to see whether a DWA effect occurs when medical decisions are directly important for decision-makers themselves. On the other hand, the present study reaffirms previous research showing that expert doctors do not benefit from unconscious thought while making decisions in complex situations. [24].

The important lesson that can be learned from the present study, and one that opposes the central claim of the DWA framework, is that there is no reason to assume that practising physicians do well to distract themselves with other tasks before making important decisions for patients.

A possible weakness in our study is the artificiality of the experimental setup. Doctors in clinical practice who receive relevant patient information and medical characteristics seldom have to rely on their memory only before deciding what action to take.

Conclusion

In a relevant medical decision task (Simple or Complex), unconscious thought did not lead to better performance than conscious thought. Our findings seem to indicate that the DWA effect, if theoretically valid, does not take place in physicians or in medical students. Our decision task, although theoretically similar to previous studies, failed to reveal a superior effect of unconscious thinking.

Essentials

• The basic design for the deliberation-without-attention effect was successfully extrapolated to a decision task in the medical domain, where doctors and medical students assessed the life expectancies of virtual patients as a computerized medical decision task.

• The deliberation-without-attention effect could not be replicated.

• Unconscious thought did not lead to better performance of experienced physicians or medical students than conscious thought.

• Experts made more accurate assessments of life expectancies than novices.