Why does the mind wander? Maybe when the agent’s current goal is deemed insufficiently rewarding, the cognitive control system initiates a search for a new, more rewarding goal

Why does the mind wander? Joshua Shepherd. Neuroscience of Consciousness, Volume 2019, Issue 1, October 22 2019, niz014, https://doi.org/10.1093/nc/niz014

Highlights
. Makes a novel and empirically tractable proposal regarding why the mind wanders
. Offers novel explanations of data on mind wandering
. Offers predictions for future work on mind wandering
. Integrates literature on cognitive control with the literature on mind wandering
. Discusses implications for a philosophical account of the nature of mind wandering

Abstract: I seek an explanation for the etiology and the function of mind wandering episodes. My proposal—which I call the cognitive control proposal—is that mind wandering is a form of non-conscious guidance due to cognitive control. When the agent’s current goal is deemed insufficiently rewarding, the cognitive control system initiates a search for a new, more rewarding goal. This search is the process of unintentional mind wandering. After developing the proposal, and relating it to the literature on mind wandering and on cognitive control, I discuss explanations the proposal affords, testable predictions the proposal makes, and philosophical implications the proposal has.

Predictions

The cognitive control proposal makes predictions. Confirmation of these would be good news; disconfirmation would be bad news.

First, given the explanation offered for the initiation of mind wandering episodes, the proposal predicts that increases in reward for satisfying an occurrent goal would correlate with decreases in propensity to mind wander. It is well-confirmed that increasing reward leads to boosts in performance level, and to overcoming any purported “ego-depletion,” even for very boring tasks. Paradigms that have established this result could be used to test for the place of mind wandering in the behavioral data.

Second, the proposal predicts that increases in reward for non-occurrent goals the agent possesses would increase mind wandering. We have already seen that reminding agents of goals they possess, or of goals they will soon need to attempt to satisfy, leads to more mind wandering in the direction of these goals. The prediction here is more specific. If one were to, e.g., notify participants that they were soon to perform a task associated with some level of reward, and then to put participants through a low reward task, the prediction is that tendency to mind wander towards this task would be associated with the discrepancy in reward between the current and upcoming task.

Third, this proposal draws upon a view of the cognitive control system on which the learning of values associated with goals, and the learning of values associated with stimuli features predictive of goals, is crucial. So the proposal, plus plausible assumptions about reinforcement learning processes, predicts that it is possible to train participants to associate stimuli with certain goals, and that registration of such stimuli would generate mind wandering to the degree that the associated goal is rewarding. Very costly goals would produce little mind wandering. Cheap but rewarding goals would produce more.

And it may be possible to extend this result. It depends on what the agent associates with rewarding goals. Above I suggested that the system need not always compare value between explicit goals, and that the value computation might include an association between expected levels of reward and particular environments. If so, simply placing an agent in such environments would manipulate levels of unintentional mind wandering.

It may be useful to distinguish predictions this proposal makes from a related proposal: the current concerns hypothesis. The current concerns hypothesis (for which, see Klinger et al. 1973; Smallwood and Schooler 2006) has it that mind wandering is caused by a shift in salience—when one’s current goals (or concerns: here I use these terms interchangeably), become more salient than the external environment, one’s mind begins to wander. As Smallwood explains the view, “attention will be most likely to shift to self-generated material when such information offers larger incentive value than does the information in the external environment” (2013, 524). This proposal is distinct from mine in the following ways. First, I propose a specific mechanism, connected with recent modeling work in cognitive control, to explain the onset of mind wandering. Thus far, of course, the proposal can be seen as a specification of the current concerns hypothesis. Second, this mechanism initiates mind wandering not by turning attention to one’s current concerns, but by directed thought to search for a more valuable goal than the present one. So the cognitive control proposal makes predictions the current concerns hypothesis does not. For example, the cognitive control proposal predicts that propensity to mind wander could be increased by devaluing the present goal, independently of the salience of any of one’s current goals. That is, no matter how much one’s current goals or concerns lack salience, once could increase mind wandering by devaluing the occurrent goal. And it predicts that mind wandering will not turn directly to one’s other goals—the mind may wander to the environment, rather than to internal concerns, since this is one way the agent may attempt to find a more rewarding task. So we should, e.g., be able to find episodes of more intense environmental scanning as a part of the mind wandering episode. Indeed, if the environment is expected to contain valuable options, one would predict that this is where attention will go, rather than to any internal space of concerns.

This is not to deny that mind wandering represents a failure in some sense. McVay and Kane (2010b) have argued that mind wandering represents an executive control failure. What fails is a process of goal maintenance: “we suggest that goal maintenance is often hijacked by task-unrelated thought (TUT), resulting in both the subjective experience of mind wandering and habit-based errors” (324). The possibility I am raising is that failures of goal-maintenance could in another sense be successes of a different process. Indeed, perhaps processes of goal-maintenance are closely related to the value-based process of estimating the expected value of continuing on some task, or of searching for a new task, that I propose underlies unintentional mind wandering.

In sum, the proposal is plausible on its face. If correct, it promises to explain a range of data regarding mind wandering, and to explain the—from the agent’s conscious perspective very puzzling—initiation of mind wandering episodes. The proposal may also contribute to explanations of the dynamics of mind wandering. The predictions this proposal makes are testable, and work in this direction might take steps towards further integrating knowledge of how cognitive control works with knowledge of how mind wandering works.