Test-retest reliability of decisions under risk: Evidence from two behavioral and EEG experiments

894 

Abstract Submission 

Jia Jin¹, Qin Xiao¹, Qiang Shen¹

¹Shanghai International Studies University, Shanghai, China

Jia Jin  
Shanghai International Studies University
Shanghai, China

Qin Xiao  
Shanghai International Studies University
Shanghai, China

Qiang Shen  
Shanghai International Studies University
Shanghai, China

Decisions involving risk are ubiquitous, and understanding how individuals manage risk is crucial in everyday life (Korucuoglu et al., 2020). Consequently, accurately measuring individuals' risk-taking propensities and the stability of these preferences over time has become increasingly important in recent years. However, few studies have systematically examined the reliability of EEG responses during risk-taking activities, and the potential of EEG to act as a predictor of behavior remains elusive.

In this study, we recruited a sample of 41 healthy participants to undertake the same experimental tasks involving risk twice over a short-term interval of 7-14 days. Thirty-six subjects successfully completed both sessions. In each session, we simultaneously recorded their behavioral and EEG data for two risk-related tasks, which included a classical 5-25 monetary gambling task (Task 1) and a mixed gambling task (Task 2). We performed Intraclass Correlation Coefficient (ICC) analyses on both the behavioral and EEG data to assess the test-retest reliability.

In the 5-25 monetary gambling task, participants generally tended to choose options with larger outcomes, a tendency that became especially prominent after experiencing a large loss in a prior trial. This behavioral pattern was reflected by a good ICC index of 0.677. Regarding the Event-Related Potential (ERP) data, we observed a prominent Feedback-Related Negativity (FRN) amplitude and an attenuated P300 amplitude in loss conditions, compared to gain conditions. Moreover, our analyses revealed fair reliability for FRN (0.440-0.599), while the reliability for P300 was relatively poor (0.258-0.617). Notably, single-trial EEG analyses suggested that these feedback-sensitive FRN and P300 could predict risk propensity in subsequent trials, a finding that remained robust in the retest phase.
In the mixed gambling task, subjects consistently demonstrated a preference for riskier choices, with no significant loss aversion detected in either session. While there was a general tendency to accept gambling opportunities, their inclination to do so was modest, yet exhibited good reliability across sessions (ICC = 0.701). EEG patterns were consistent with those observed in Experiment 1. The trial-wise analysis further supported these findings; specifically, outcomes eliciting increased FRN and decreased P300 amplitudes were linked to the magnitude of feedback. Additionally, more pronounced fluctuations in both FRN and P300 were observed when the Expected Value (EV) was negative, indicating a stronger neural response to anticipated losses. The reliabilities of FRN (0.707-0.719) and P300 (0.624-0.654) in response to gains and losses were good. Notably, the individual-level analysis suggested that the feedback-sensitive P300 amplitude in the 5-25 gambling task could predict the risk preference displayed in the mixed gambling task.

The study concludes that subjects' propensities for risk-taking remained consistent, even following a one to two-week break. The good ICC for both FRN and P300 amplitudes, in addition to their significant correlation with risk preference behaviors, suggests the potential of these electrophysiological markers to characterize risk preferences. Thus, these measures may serve as viable biomarkers to distinguish individual variations in risk preferences, applicable in both controlled laboratory conditions and real-world environments.

Decision Making ¹

EEG/MEG Modeling and Analysis

EEG ²

Electroencephaolography (EEG)