Author Reply: Affect,Value,Uncertainty,and Action

I’d like to thank the commentators for their thoughtful replies to our collection of articles on emotion and action. And I’m especially grateful to Gregor Hochstetter and Hong Yu Wong (2017) for their extended discussion of my article, “At the Core of our Capacity to Act for a Reason.” Hochstetter and Wong underscore the important point that action, like most higher level psychological phenomena, involves a layered array of systems and processes, central and peripheral, direct and indirect. No single system is likely to suffice, and I certainly didn’t take my claim that the affective system is “at the core” of our capacity to act for a reason to imply the affective system suffices to account for the complex phenomenon of action. For example, consider the complex “job description” for the circulatory system—to deliver oxygen, nutrients, antibodies, hormones, etcetera to the body, while removing various wastes and toxins. To say that the heart is at the core of the circulatory system is not to say that the heart alone fills this “job description,” only that it plays a central, indispensable role in it. Why, if the affective system cannot by itself fill the “job description” for acting for a reason, might one say that it nonetheless lies at the core of that phenomenon?

The key idea has to do with the role of the affective system in evaluation and control. While complex, action is also a highly integrated form of activity. Acting for a reason is not a mere concatenation of mental events that results in an outcome that happens to cohere with a goal, but a suite of processes—attention, perception, memory, inference, decision, motivation, decision, and monitoring—actively orchestrated around the attainment of an aim. We can speak of this suite of processes as acting for a reason because it is attuned to and controlled by evaluative representations that encode this aim and associated chances. This idea does not depend upon any particular philosophical theory of action, but upon the most general framework we have for understanding rational action, which dates back at least as far as Aristotle (ca. 350 B.C.E./1993: 433). In this general framework, decision and action depend upon two principal components: assessment of the value of certain outcomes and of the likelihood of these various outcomes conditional upon action. This framework is ubiquitous today as the backbone of rational decision theory—which finds application not only in human action, but also in the actions of intelligent animals and even in the engineering of artificial agents built from scratch to be capable of autonomous action. We now have good neuroscientific evidence that value and risk are separately encoded in the affective system, and that these encoded values then feed directly into decisional and action-guiding processes (Fiorillo, Tobler, & Schwartz, 2003; Grabenhorst & Rolls, 2011; Quartz, 2007; Tobler, O’Doherty, Dolan, & Schultz, 2006). Because these are the agent’s evaluations, there is a clear sense in which, when they achieve the integrated control of action, the resultant action is appropriately seen as an expression of that agent and her reasons for acting.

As Andreas Eder (2017, p. 344) points out, this sort of affective attunement and control of action via forward evaluative modeling “provides an elegant solution to what Railton identified as the agency-without-regress problem,” since “[e]motional appraisals can direct actions through the interface of ideomotor mechanisms without invoking a homunculus that ponders about the benefits and feasibility of an action course.” Like Eder, I believe that many have failed to appreciate the active, real-time attunement and control of action by emotion because they fail to move beyond the “‘automatic,’ ‘stimulus-driven,’ ‘impulsive’” (2017, p. 344) view of emotion as a mere “input” into the processes governing action. Thus, while I welcome the contribution Hochstetter and Wong (2017) make in setting the core activities of action-attunement and action-control in a wider psychic context, I don’t see any conflict with centering these evaluative activities in the affective system, broadly construed.

Bence Nanay (2017) suggests yet another way in which affect might play a core role in the generation of action. He notes that, in addition to calculations representing the expected value of potential actions, agency appears to involve a further element that actually translates this expected value into action, for example, by selectively lifting the inhibition against performing the relevant action. The only brain processes we know that reliably influence this kind of selective release, Nanay argues, involve affective responses to stimuli. Thus, core affective involvement in action is “the norm,” not the exception.

Finally, Guido Gendolla (2017) argues that the most important contribution emotion makes to action is motivational. However, he does not mean to restrict such emotional motivation to the older picture of “‘automatic,’ ‘stimulus-driven,’ ‘impulsive’” processes that Eder (2017, p. 344) criticizes. Gendolla argues that emotion does not merely “trigger” or “amplify” action, but also “provid[es] useful feedback” about “progress during self-regulated goal pursuit” (Gendolla, 2017, p. 349) and can function as an “incentive” in the form of imagined, anticipated affective responses. The representation of anticipated evaluative outcomes is, I agree, part of the fundamental motivational function of emotion. This makes sense as a form of evolved design—often the most important consequences of actions (or inactions) lie in a remoter future—so if we are to be appropriately moved by them in the present, our affective system must do more than respond to current stimulation. Instead, our action-guidance system would have to function prospectively by simulating future states using our affective system as a “test bed.” Recent research suggests that off-line empathic simulation of this kind is indeed a regular part of our “default network” functioning—the interconnected patterns of activity that become most prominent in the brain when not engaged in a focused task (Buckner, Andrews-Hanna, & Schacter, 2008). The “default mode” is a central, recurrent activity of the brain, attesting to the importance of evaluating and reevaluating potential future paths as one goes through life. Given affect’s central role in attuning and controlling action, it will be important for us to pay attention to unaroused affect and “off-line” simulations of anticipated affect, not simply the aroused “emotions” like fear, anger, joy, and disgust, which have been so prominent in studies of emotion in the past.