Title：Attention Utility: Theory and Evidence
Speaker：Prof. Chew Soohong
Date：22 May 2021
Time: 16:00PM-18:00PM
Location：HongYuan Building 311
Abstract：
We offer an attention utility model for decision making under risk and uncertainty from the perspective of our mind as a cognitive miser. For uniform finite lotteries, our axiomatization delivers attention-dependent decision weights which resemble the divisive normalization form in the predictive brain literature and can accommodate behavior often associated with consequentialist choice under high attentiveness. In addition to the commonly discussed Allais paradoxes and attitude towards four-fold risks, the model can account for recent evidence on the continuity of decision weights from probability weighting and the asymmetric loss-gain arousal response being linked to aversion towards even-chance mixed risk around the status quo. The descriptive validity of our model is further demonstrated in terms of how it can account for a rich range of choice anomalies with shared roots in the volatile nature of the attentional process. These anomalies include the discontinuity in certainty equivalent from event splitting, disjunction effect, uncertainty effects, and source preference encompassing ambiguity aversion and familiarity bias. Under low attentiveness, in contrast with random utility, attention utility yields an alternative account of stochastic choice. Building on the attention utility (AU) model proposed in Chew, Wang, and Zhong (2021) for decision making under risk and uncertainty, we offer a neurochemical model of attention and its role in decision making through a triad of neurotransmitters – dopamine (DA), serotonin (5HT), and norepinephrine (NE) – which modulate respectively gain, loss, and attention processing. Building on Zhong et al.’s (2009) application of DA and 5HT tones to model the loss-gain differentiation in risk attitude, we propose an NE tone, comprising the elements of the attention information messenger system concentrated in the locus coeruleus (LC), as a modulator of the attention function in AU. The LC-NE system facilitates switching between two primary modes of attentiveness: an exploitative mode which is engaged and focused versus an exploratory mode which is disengaged and unfocused. Correspondingly, the NE neuron switches between a rapid firing phasic exploitative mode involving higher energy expenditures and a slow-firing tonic mode with lower energy expenditures. We link NE phasic and tonic firing to properties of the attention function and derive testable implications linking the decision maker’s genetic makeup, LC brain activation, skin conductance response (SCR), and eye movement to observable choice behavior. Notably, attentional loss aversion in the attention function is supported by evidence linking asymmetric loss-gain SCR and behavioral loss aversion, i.e., aversion to symmetric risks around the status quo. We further discuss how to test the prediction of our model on the impact of NE tone on the fourfold pattern of risk attitude – an elevated level of caffeine/nicotine reduces aversion (tolerance) for even-chance gain (loss) and increases preference (aversion) for longshot gain (loss). Taken together, our LC-NE model of attention in decision making, involving an active interplay between phasic and tonic modes at the neuronal level, offers an neurobiological account of the cognitive miser underpinning the AU model.
