Sleep deprivation alters choice strategy without altering uncertainty or loss aversion preferences

Abstract

Sleep deprivation alters decision making; however, it is unclear what specific cognitive processes are modified to drive altered choices. In this manuscript, we examined how one night of total sleep deprivation (TSD) alters economic decision making. We specifically examined changes in uncertainty preferences dissociably from changes in the strategy with which participants engage with presented choice information. With high test-retest reliability, we show that TSD does not alter uncertainty preferences or loss aversion. Rather, TSD alters the information the participants rely upon to make their choices. Utilizing a choice strategy metric which contrasts the influence of maximizing and satisficing information on choice behavior, we find that TSD alters the relative reliance on maximizing information and satisficing information, in the gains domain. This alteration is the result of participants both decreasing their reliance on cognitively-complex maximizing information and a concomitant increase in the use of readily-available satisficing information. TSD did not result in a decrease in overall information use in either domain. These results show that sleep deprivation alters decision making by altering the informational strategies that participants employ, without altering their preferences.

Neural Mechanisms of the Transformation from Objective Value to Subjective Utility: Converting from Count to Worth

“…In the losses domain, on average participants were risk neutral ( mean ± SD α = 1.04 ± 0.26). There was no significant correlation between individual risk preferences across the gains and losses domains ( r 28 = 0.26, p = 0.16), concurring with recent studies (Kurnianingsih and Mullette-Gillman, 2015; Kurnianingsih et al, 2015; Mullette-Gillman et al, 2015a,b). For further consideration of this and additional behavioral analyses we point the reader to Kurnianingsih and Mullette-Gillman (2015), which features a larger sample (including the individuals in this study) performing the same task with slightly more trials (30 additional trials by sampling two additional levels of rEV).…”

“…On average, participants were risk neutral ( mean ± SD α = 1.04 ± 0.26) (Figure 4A), with a range of preferences. There was no significant correlation between individual risk preferences across the gains and losses domains ( r 28 = 0.26, p = 0.16), concurring with recent studies (Kurnianingsih and Mullette-Gillman, 2015; Kurnianingsih et al, 2015; Mullette-Gillman et al, 2015a,b). We then replicated our analyses to identify the regions encoding the value-to-utility transformation, covarying the rEV value regressor by individual preferences (Figure 4B).…”

“…Although these metrics have different theoretical assumptions, we find strong correlations between them in this study (gains: r 26 = −0.74, p < 0.0001; losses: r 27 = −0.45, p < 0.0001) and previous studies (Stanton et al, 2011; Kurnianingsih and Mullette-Gillman, 2015; Kurnianingsih et al, 2015; Mullette-Gillman et al, 2015a,b). …”

“…These results were obtained using a specific formulation for the value on each trial (rEV) and a common measure of individual risk preferences (α). To test the robustness and generalizability of the encoding of value subjectifcation in the daMCC, we repeated our analyses using an alternative formulation of value (chosen value, CV) and an alternative metric for quantifying individual risk preferences (risk premium, a linear formulation; Stanton et al, 2011; Kurnianingsih and Mullette-Gillman, 2015; Kurnianingsih et al, 2015; Mullette-Gillman et al, 2015a,b). These alternatives extend our analyses from simply across gains and losses to an additional 2 × 2 space across risk preferences (α and premium) and value formulations (rEV and CV).…”

“…We additionally quantified each individual's risk premium using psychophysical indifference point analyses (for details see Kurnianingsih and Mullette-Gillman, 2015; Kurnianingsih et al, 2015; Mullette-Gillman et al, 2015a,b), which measures the degree and direction of value subjectification in a zero-centered multiplicative/linear form (as compared to the one-centered power function form of the α risk preference metric). Although these metrics have different theoretical assumptions, we find strong correlations between them in this study (gains: r 26 = −0.74, p < 0.0001; losses: r 27 = −0.45, p < 0.0001) and previous studies (Stanton et al, 2011; Kurnianingsih and Mullette-Gillman, 2015; Kurnianingsih et al, 2015; Mullette-Gillman et al, 2015a,b).…”