P055 Risk sensitivity modulates impulsive choice and risk aversion behaviors
Rhiannon L. Cowan*1, Tyler S. Davis1, Bornali Kundu2, Ben Shofty1, Shervin Rahimpour1, John D. Rolston3, Elliot H. Smith1
1Department of Neurosurgery, University of Utah, Salt Lake City, United States
2Department of Neurosurgery, University of Missouri – Columbia, Missouri, United States
3Department of Neurosurgery, Brigham & Women’s Hospital, Boston, United States
*Email: rhiannon.cowan@utah.edu
Introduction
Impulsivity is a multifaceted psychological construct that may impede optimal decision-making. Impulsive choice (IC) is the tendency to favor smaller, immediate, or more certain rewards over larger, delayed, or uncertain rewards [1]. A strategy such as risk aversion allows individuals to avoid potential loss of reward and gain instant gratification [2,3]. Risk sensitivity (RS) is defined as the variance associated with an outcome [4] and, therefore, may be examined via positive and negative prediction error (PE) signals, a canonical signal of reinforcement learning [5-7]. We posit that more impulsive individuals will exhibit risk-aversive tendencies, which will be observed via suboptimal performance and neural encoding of negative PEs.
Methods
71 neurosurgical epilepsy patients underwent implantation of electrodes into the cortex and deep brain structures. The Balloon Analog Risk Task (BART) is a useful paradigm to measure impulsivity and reward behaviors by conceptualizing the probability of potential reward [8]. Subject IC level was calculated as the difference between passive and active trial inflation time (IT) distributions. Outcome-aligned broadband high frequency (HFA; 70-150Hz) activity was modeled as a linear combination of temporal difference (TD) variables [5,9,10]. We compute the neural correlates of reward in a trial-by-trial manner from TD models with optimal learning rates [11] and RSTD models, which account for positive and negative PEs [12].
Results
MI choosers were more accurate than LI choosers (Z=2.04, p=.041), notably for yellow balloon trials (Z=4.09, p<.0001), yet LI choosers overall gained more points (Z=-3.57, p=.00036) primarily from yellow balloons (Z=-3.58, p=.00036; Fig.1). We observed no differences in optimal learning rates for reward or risk models between groups (p’s>.05) but saw increased RS was correlated with impulsivity (t(69)=-2.17, p=.03). We observed greater encoding of negative PEs (11.46%) than positive PEs (25.06%; χ2=159, p<.001). However, a group-level dichotomy revealed that MI choosers encoded significantly more negative RPEs (MI=11.42%, LI=9.45%; χ2=5, p=.025), whereas LI choosers encoded more positive PEs (χ2=4, p=.039).
Discussion
We utilize a dataset of 7000 intracranial electrodes to model RS and the neural underpinnings of IC. During BART, we found that LI choosers took more risks, leading to more optimal performance, while MI chooser’s accuracy-point tradeoff suggests a risk-aversion strategy, that aligns with the IC definition. Neurally, MI choosers encoded more negative PEs, and LI choosers encoded more positive PEs, which, in tandem with the differential behavioral strategies exhibited, suggests RS drives reward-seeking and may be modulated by impulsivity. This supports previous studies showing associations of positive PEs to risk-seeking behavior and negative PEs to risk-aversion behavior [13]. These findings have implications for decision-making, RS, and IC.
Figure 1. Figure 1. A. BART schematic B&C. IC scatter & histogram using Z-Value difference between active & passive ITs (apZVals) D. Accuracy by color E. Points by color F. LI & MI point distributions G-I. Regression plots: performance vs IC J. Glass brain of electrodes K&L. LI & MI regions encoding negative PE & positive PE M&N. LI & MI risk PE signals by trial category O. Risk sensitivity vs impulsivity.
AcknowledgementsThis research was supported by funding: R01MH128187
References
1. https://doi.org/10.1097/01.wnr.0000132920.12990.b9
2. https://doi.org/10.3389/fpsyg.2015.0051
3. https://doi.org/10.1016/j.bbr.2018.10.008
4. https://doi.org/10.1523/JNEUROSCI.5498-10.2012
5. https://doi.org/10.1016/j.neuron.2006.06.024
6. https://doi.org/10.1038/s41586-019-1924-6
7. https://doi.org/10.31887/DCNS.2016.18.1/wschultz
8. https://doi.org/10.1037//1076-898X.8.2.75
9. https://doi.org/10.1523/JNEUROSCI.2041-09.2009
10. https://doi.org/10.1523/JNEUROSCI.2770-10.2010
11. doi.org/10.1109/TNN.1998.712192
12. https://doi.org/10.1023/A:1017940631555
13. https://doi.org/10.1371/journal.pcbi.1009213
