Deliberation is (probably) triggered and sustained by multiple mechanisms

Abstract

De Neys proposes that deliberation is triggered and sustained by uncertainty. I argue that there are cases where deliberation occurs with low uncertainty – such as when problems are excessively complicated and the reasoner decides against engaging in deliberation – and that there are likely multiple factors that lead to (or undermine) deliberation. Nonetheless, De Neys is correct to surface these issues.

De Neys argues – lucidly, in my view – that dual-process theories have been (largely) ill-focused. Indeed, it seems that the core aspect of any dual-process model is not simply how it describes the two types of processes (or “systems”), but rather how describes the interaction between the processes. To this end, De Neys proposes that uncertainty is the key explanatory factor that determines when “type 2” deliberation is triggered (i.e., the “switch feature”).

Although many popular dual-process models have ignored the issue, as noted by De Neys, there is nonetheless a growing body of work that has focused on understanding what triggers deliberation. If intuitive (“type 1”) processing is autonomous (i.e., it is triggered automatically from some stimulus or thought process), as myself and others have argued (e.g., Pennycook, 2017; Thompson, 2013), then deliberative processes must be triggered by some underlying cognitive factor or factors. That is, when explaining the progeny of the process, intuitive processes can be explained by simple stimulus–response pairings. However, one needs to posit additional factors that would then lead to subsequent deliberative processes.

In my past work, I have focused on the potential role of response conflict in triggering deliberation (Pennycook, 2023; Pennycook, Fugelsang, & Koehler, 2015) – that is, cases where the system detects a conflict between intuitive outputs lead to subsequent deliberation. The three-stage dual-process model separates the initial “intuition” stage (where processes are initiated autonomously) from the subsequent “metacognition” stage (where conflicts between outputs of stage 1 are monitored). The presence or absence of conflict then determines the extent of deliberation in the final “reason” stage. This matches well with the sort of tasks that are common in the literature because they typically involve a salient (but incorrect) intuition that conflicts with some other relevant (and more normatively accurate) factor. However, as De Neys notes, a model that focuses solely on intuitive conflict does not help explain why deliberation occurs in the absence of intuitions.

To solve this problem, De Neys argues that general uncertainty casts a wider net and can explain both cases where conflict detection leads to deliberation but also cases where there are no apparent intuitions present. As conflict between intuitions leads to uncertainty, so to does a lack of clear answer. There is a lot to like about this proposal; however, I am hesitant to adopt uncertainty as the key explanatory factor.

One possibility is that “uncertainty” is merely correlated with the underlying causal feature (or features) that trigger deliberation. Indeed, there may be cases when uncertainty is high but deliberation is low. For example, when an individual is facing a problem that is prohibitively complicated they surely have a feeling of high uncertainty – nonetheless, the individual may decide to not engage in deliberation and to simply not bother attempting a solution (and, indeed, people do tend to prefer tasks that require less cognitive effort; Chong, Bonnelle, & Husain, 2016; Shenhav et al., 2017). Hence, in such a scenario uncertainty would not lead to deliberation.

De Neys has offered a strong framework for understanding how type 1 outputs contribute to the engagement of subsequent deliberation. He even mentions a similar case where deliberation does not lead to an answer and the individual decides to stop deliberating. This is accommodating by noting that opportunity cost could be factored into the uncertainty parameter. However, one may question at that point whether deliberation is prompted by several separate mechanisms and that uncertainty is just a reasonable proxy for many of them.

This issue is related to recent work in the cognitive control literature that has investigated what might lead individuals to engage in effortful processing across situations. For example, the expected-value-of-control model posits that people weigh the costs and benefits of exerting mental effort (Shenhav, Botvinick, & Cohen, 2013; Shenhav, Prater Fahey, & Grahek, 2021). Of course, it may be that this calculation only occurs after deliberative processes have been initiated by some other process and therefore that costs/benefits play a role in how much deliberation occurs, not whether it is triggered in the first place. Nonetheless, if one takes the view that dual processes can be considered as opposite poles of a single continuum (as opposed to two fundamentally separate processes; De Neys, 2021), then the expected value of control is a factor that may play a role in whether substantive deliberation occurs. This is noted by De Neys but the implications for the focus on uncertainty are not addressed.

Relatedly, a great deal of research has focused on apparent individual differences in the willingness to engage in deliberation (see Pennycook [2023] for a review). If one were to focus solely on uncertainty as the factor that triggers deliberation, this would presume that individual differences in uncertainty-related processes (e.g., sensitivity to uncertainty) would be central. However, individual differences in deliberation tend to focus on whether people are willing to engage in cognitive effort (e.g., in the “Need for Cognition” scale; Petty, Brinol, Loersch, & McCaslin, 2009) or in the willingness to question one's intuitions and prior beliefs (e.g., in the “Actively Open-minded Thinking” scale; Pennycook, Cheyne, Koehler, & Fugelsang, 2020). Here too, the key factors that explain whether people engage in a meaningful level of deliberation may not be explained very well by appealing to uncertainty.

To conclude, I agree wholeheartedly with De Neys's argument that the underlying assumptions of many dual-process theories are poorly conceived – particularly as it related to the exclusivity feature. However, I am less convinced that deliberation can be explained by a single underlying causal mechanism. It seems likely to me that there are multiple separate mechanisms that are relevant for understanding why deliberation occurs. In any case, even if I am uncertain about using uncertainty as a key mechanism in dual-process models, I am nonetheless convinced that De Neys's proposal is certainly a step forward.