A common feature of many models of willpower is effortful moment-by-moment inhibition of desire for immediately attractive but ultimately inferior rewards, a process linked to brain regions including dorsolateral prefrontal cortex (DLPFC) (e.g., McClure, Laibson, Loewenstein, & Cohen, Reference McClure, Laibson, Loewenstein and Cohen2004). Ainslie conceptualizes this suppression primarily in terms of inhibition of tempting impulses or attentional filtering of distracting stimuli, and contrasts it with more stable, lasting intentions based on self-prediction, which are termed resolve. Yet, growing evidence suggests that normative choices need not arise from the suppression of impulses toward tempting rewards. Instead, neuroscientific data and computational modeling approaches suggest that decision-makers can differentially weight relevant stimulus attributes of available options in order to arrive at optimal outcomes (Berkman, Hutcherson, Livingston, Kahn, & Inzlicht, Reference Berkman, Hutcherson, Livingston, Kahn and Inzlicht2017; Rangel & Clithero, Reference Rangel and Clithero2014).

Attribute weighting models have successfully explained decision behavior for a range of tasks, including dietary and intertemporal choice. In dietary choice, individuals select between options varying in taste and health attributes. Dietary self-control has been associated with changes in the relative weights associated with the perceived tastiness and healthiness of different food options, both behaviorally and at the neural level (Bhanji & Beer, Reference Bhanji and Beer2012; Hare, Camerer, & Rangel, Reference Hare, Camerer and Rangel2009; Harris, Hare, & Rangel, Reference Harris, Hare and Rangel2013; Tusche & Hutcherson, Reference Tusche and Hutcherson2018). However, this attribute weighing process does not appear to require explicit suppression of taste representations for effective self-control. For example, Harris et al. (Reference Harris, Hare and Rangel2013) used electroencephalography (EEG) to measure changes in neural signals associated with subjective value for participants during natural responding versus dietary self-control. Contrary to Ainslie's description of this study as demonstrating “suppression of reward center activity” during self-control (sect. 4), participants showed similar neural responses to taste attributes of the foods across both natural and self-control sessions. In contrast, neural activity associated with the perceived healthiness of the foods increased dramatically under the self-control conditions, consistent with a greater weighting of this attribute in this decision context (Harris et al., Reference Harris, Hare and Rangel2013), and in line with previous observations (Hare et al., Reference Hare, Camerer and Rangel2009). Furthermore, studies that trace the decision process using mouse-tracking have shown that healthy dietary choices depend not only on the relative weights of taste and health attributes, but also the timing with which they are incorporated into the decision process (Lim, Penrod, Ha, Bruce, & Bruce, Reference Lim, Penrod, Ha, Bruce and Bruce2018; Maier, Raja Beharelle, Polania, Ruff, & Hare, Reference Maier, Raja Beharelle, Polania, Ruff and Hare2020; Sullivan, Hutcherson, Harris, & Rangel, Reference Sullivan, Hutcherson, Harris and Rangel2015). These different aspects of weighting strength and timing can be captured by a time-varying drift diffusion model (Maier et al., Reference Maier, Raja Beharelle, Polania, Ruff and Hare2020).

Similarly, attribute weighting can better account for choice behavior in intertemporal choice paradigms, in which participants select among options varying in reward amount and time delay. Individual variation in intertemporal choice appears to reflect differential focus on reward amount versus time information, as well as when these attributes are incorporated into the decision process (Amasino, Sullivan, Kranton, & Huettel, Reference Amasino, Sullivan, Kranton and Huettel2019; Reeck, Wall, & Johnson, Reference Reeck, Wall and Johnson2017). Specifically, more patient individuals focus on directly comparing between reward amounts, whereas comparatively impatient decision-makers integrate across amount and time information within options. Similarly, patient individuals incorporate amount information into the decision process earlier and have shorter response times (Amasino et al., Reference Amasino, Sullivan, Kranton and Huettel2019), effects that may seem paradoxical from an effortful suppression perspective. Experimental manipulations that promote a comparative, attribute-based strategy can also causally shift decision-making toward greater patience (Reeck et al., Reference Reeck, Wall and Johnson2017). Although broadly congruent with the results from dietary choice tasks above, these data are also in line with the observation that manipulations such as displaying zero-pay events (Magen, Dweck, & Gross, Reference Magen, Dweck and Gross2008) can influence intertemporal choice by making the trade-off between time and money attributes more explicit (Lempert & Phelps, Reference Lempert and Phelps2016).

Thus, the attribute weighting framework provides a powerful explanation for normative choice behavior across a variety of tasks. Moreover, by combining this approach with physiological data, researchers have shed light on the neural correlates of shifts in attribute weighting. In particular, converging evidence suggests that selective attention plays a key role in attribute weighting, both through eye-tracking of endogenous attentional shifts (e.g., Amasino et al., Reference Amasino, Sullivan, Kranton and Huettel2019) and exogenous cueing to specific attributes (Hare, Malmaud, & Rangel, Reference Hare, Malmaud and Rangel2011). Recent data suggest that attentional cueing can change both the strength and timing of attribute weighting (Maier et al., Reference Maier, Raja Beharelle, Polania, Ruff and Hare2020), supporting an attentional mechanism for previously observed effects of attribute integration timing (Lim et al., Reference Lim, Penrod, Ha, Bruce and Bruce2018; Sullivan et al., Reference Sullivan, Hutcherson, Harris and Rangel2015). Under conditions of simple choice, this process is likely driven by salience and/or highest value (Busemeyer, Gluth, Rieskamp, & Turner, Reference Busemeyer, Gluth, Rieskamp and Turner2019; Krajbich, Armel, & Rangel, Reference Krajbich, Armel and Rangel2010; Rangel & Clithero, Reference Rangel and Clithero2014). However, in decision contexts where there is a conflict between choice-relevant attributes (i.e., requiring willpower), attentional shifts to resolve this competition likely arise from activity in the DLPFC, which has known links to executive function and attentional selection (Kastner & Ungerleider, Reference Kastner and Ungerleider2001; Miller & Cohen, Reference Miller and Cohen2001). Consistent with this idea, new analyses suggest that increased DLPFC activity is associated with choices that conflict with current objectives; for example, the DLPFC shows an increased response to unhealthy choices when participants are actively focusing on health (Hutcherson, Rangel, & Tusche, Reference Hutcherson, Rangel and Tusche2020).

Finally, attribute weighting can provide an alternative way to think about the issues raised by Ainslie's model of willpower. Existing models of attribute weighting seem analogous to suppression, reflecting an effortful momentary shift in the representation of stimulus properties such as healthiness or reward amount. However, attribute weighting could conceivably extend to the types of recursive self-predictions described by Ainslie as resolve. Self-referential processes such as prospection, imagination, and memory have already been implicated in patient intertemporal choice (Jenkins & Hsu, Reference Jenkins and Hsu2017; Lempert & Phelps, Reference Lempert and Phelps2016; Lempert, Speer, Delgado, & Phelps, Reference Lempert, Speer, Delgado and Phelps2017). The finding that delay-of-gratification failures are more common when the arrival of a reward is uncertain (Kidd et al., Reference Kidd, Palmeri and Aslin2013; McGuire & Kable, Reference McGuire and Kable2013) also points to the role of internal beliefs in maintaining resolve. Although these types of introspective processes have thus far received comparatively little attention in computational models of self-control, individual variation in attribute weighting strategies likely reflects larger differences in life experience, temperament, and personal beliefs. By adding self-referential processes to existing models of attribute weighting, future work may better characterize the computational and neural correlates of willpower.