Paper Title: heuristics in risky decision-making relate to preferential representation of information

Research Background

When making choices, individuals not only differ from each other but also deviate from normative theoretical recommendations. One explanation for this difference is that individuals have unique information representation preferences when evaluating choice options. The authors of this paper hypothesized that during the choice evaluation process, individuals’ reliance on different information sources reflects a preferential representation of the most informative stimuli.

In recent years, decision theory has suggested that individuals should evaluate the value of choices by calculating expected utility. However, psychologists have long observed that participants typically use heuristic methods rather than strictly following the expected utility computation strategy. These heuristic methods include inappropriately weighting utility or probability information. Although some models have parameterized people’s heuristic dependency on types of information, the specific neurocognitive mechanisms remain unknown. This paper investigates whether individuals’ preferential representation of certain stimuli when using specific information is related to their behavioral characteristics through magnetoencephalography (MEG).

Research Source

This paper is authored by Evan M. Russek, Rani Moran, Yunzhe Liu, Raymond J. Dolan, and Quentin J. M. Huys, associated with University College London, Queen Mary University of London, and Beijing Normal University. The paper was published in the journal “Nature Communications” and accepted on May 3, 2024.

Research Objective and Methods

This paper aims to explore the relationship between individuals’ behavioral weighting of probability and reward information during risky decision-making and their neural representation of relevant stimuli. The research includes two experiments:

1. Primary Experiment: Using MEG to record neural data from participants (n=19) during a risky decision-making task. Participants had to choose to accept or reject a gamble in each trial, with outcomes being either a safe result or one of two gambling results. This experiment seeks to decode the neural representation of stimuli during the decision-making process using multivariable methods.
2. Validation Experiment: Using behavioral priming to measure stimulus representation based on perception detection latency. This experiment verifies the primary experiment’s results by introducing perception detection tasks during the evaluation period.

By analyzing how individuals behaviorally weight probability and reward information during the choice process, this study examines whether these behavioral characteristics are linked to individuals’ neural representation preferences for related outcome stimuli.

Experimental Procedure

1. MEG Decision-Making Task

   • Participants completed a task involving multiple bet options, choosing between a known fixed outcome and a probabilistic gamble in each trial.
   • Neural data was simultaneously recorded using MEG to decode participants’ neural representations during risky decision-making.
   • Utilizing recently developed multivariable methods, MEG signatures of visual stimuli during decision-making were identified and analyzed.

2. Perception Detection Task

   • Participants performed a decision task with additional perception detection tasks in some trials.
   • After the disappearance of the probabilistic stimulus, one of three outcome stimuli was displayed with a probe, requiring participants to report the direction of the probe as quickly as possible.
   • The influence of choice preferences on stimulus possibility and absolute reward was analyzed.

Data Analysis and Algorithms

Multivariable methods were used to decode MEG data, training several classifiers to identify neural representations at each moment and verifying these representations’ reproducibility during the choice evaluation period. A linear model was adopted to predict reactivated outcome representations, and a robust linear regression model was used to handle individual differences.

Results

Primary Experiment Results: MEG Decision-Making Task

• Using MEG decoding methods, it was found that individuals’ neural responses to high probability and high reward outcomes during the decision-making process were related to their behavioral characteristics.
• Comparative analysis of reactivated outcomes indicated that individuals who preferred probability information showed higher variability.
• Neural responses at specific time points were significantly correlated with individuals’ probability and reward information weighting during behavioral choices.

Auxiliary Experiment Results: Perception Detection Task

• Behavioral experiments showed that individuals who used probability information in decision-making had faster response times to high probability outcomes in the perception detection task. Similarly, those who relied on reward information had faster response times to high reward outcomes.
• These results further validated the findings from the MEG experiment.

Conclusion and Significance

This study indicates that the way individuals filter and select information during risky decision-making is related to their heuristic decision-making patterns. This finding reveals the connection between individuals’ neural representation priorities and behavioral tendencies when evaluating choices. Specifically, different outcome representations reflect individuals’ heuristic reliance on specific source information.

Research Highlights

• The study reveals differences in individuals’ neural representation of probability and reward information during risky decision-making.
• By integrating behavioral and neural data, the study verifies and extends previous theories on heuristic decision-making and information selection tendencies.
• The research provides a potential neurological mechanism explanation for the widespread behavioral differences in individuals’ decision-making processes.

These findings not only enhance our understanding of the decision-making process but also offer potential directions for improving treatment strategies for mental health issues. By understanding individuals’ tendencies to select and prioritize certain information, more effective intervention methods can be designed to improve decision-making abilities.

Research Methods and Data Sources

The raw and pre-processed MEG data used in the study are stored in the OpenNeuro database, and the behavioral data are stored on Zenodo. The analysis code for the study has also been made public.

Overall, this study investigates individuals’ information selection and usage tendencies during risky decision-making from both neural and behavioral perspectives through multi-angle data analysis and experimental validation. It provides significant theoretical support and empirical evidence for further understanding the decision-making mechanism.