Information from Lay-Language Summaries is Embargoed Until the Conclusion of the Scientific Presentation
292—Neural Mechanisms: Processing Uncertainty and Risk
Sunday, November 10, 2013, 1:00 pm - 5:00 pm
292.02: Multiple neural circuits predict different types of financial risk taking
Location: Halls B-H
*C. C. WU, K. KATOVICH, B. KNUTSON; Psychology, Stanford Univ., Stanford, CA
Abstract Body: To account for human financial risk taking, traditional finance models have focused on statistical moments of mean and variance. Higher order statistical moments such as skewness, however, can also influence choice. We have proposed that part of this influence may result from peoples’ affective reactions to large but unlikely potential outcomes (e.g., wins in the case of lotteries, damage in the case of insurance). In this study, healthy human subjects (n=19) chose between risky gambles and certain gambles (i.e., 100% $0.00) for real money while being scanned with functional magnetic resonance imaging (FMRI). Gambles shared equal mean and variance, but varied in terms of skewness (i.e., positive skew: 25% +$5.25, 75% -$1.25, symmetric: 50% +$3.05 / 50% -$3.05; negative skew: 75% +$1.25, 25% -$5.25). Anticipatory neural activity (i.e., which occurred while subjects viewed gambles but before they selected their choice) was used to predict individuals’ risky financial choices in both whole brain and volume of interest analyses. Whole brain analyses utilized GraphNet, a multivariate pattern analysis method optimized for FMRI, which implements a combination of regularization parameters designed to yield generalizable yet interpretable solutions (Grosenick et al., 2013), while volume of interest analyses used logistic regressions to regress localized brain activity on choice. Across all gamble types, the whole brain GraphNet solution predicted risky vs. certain choices at 65.8% (leave one subject out cross validation, p<.001). Visualization of features selected by the GraphNet model revealed that positive features in the nucleus accumbens (NAcc) and negative features in the anterior insula predicted risky choice in general. Volume of interest analyses confirmed these associations over all gamble types, but additionally indicated that NAcc activity more robustly predicted risk seeking in positive skew trials, whereas anterior insula activity more robustly predicted risk avoidance in negative skew trials -- consistent with an anticipatory affect account (Wu et al., 2012). Together, these findings suggest that different neural circuits promote different kinds of financial risk seeking and imply that neurofinance models may account for choices that transcend the boundaries of traditional finance models.
Lay Language Summary: Why do people take risks? And why are people attracted to different kinds of risks? For instance, what drives purchases of lottery tickets versus insurance policies? New developments in functional magnetic resonance imaging (FMRI) allow investigators not only to identify neural correlates of but also neural predictors of choice. Novel statistical methods further facilitate identification of where and when brain activity can predict a person’s next choice. Applying these techniques to financial risk taking, we asked human subjects to choose whether to accept or reject different types of gambles as they were scanned with FMRI. Statistical analyses indicated not only that we could predict subjects’ choices to take a risk with brain activity seconds before choice, but also that activity in different circuits pushed subjects towards or away from different types of risk. Specifically, increased activity in a region associated with gain anticipation and feelings of excitement (the nucleus accumbens) predicted that people would be more likely to accept gambles involving high magnitude but low probability gains (e.g., like purchasing lottery tickets). However, increased activity in a different region associated with loss anticipation and feelings of anxiety (the anterior insula) predicted that people would be more likely to reject gambles involving high magnitude but low probability losses (e.g., like purchasing insurance policies). By applying neuroeconomic methods, these findings go beyond traditional economic models to suggest not only that researchers can use neural activity to predict whether subjects will take a financial risk prior to choice, but also which kind of risk they are most likely to accept. The results thus may have implications for understanding how people allocate their financial assets between high and low risk options, as well as for predicting what kinds of risks people are willing to bear.
Neuroscience 2013 (43rd annual meeting of the Society for Neuroscience)Exit