Format

Send to

Choose Destination
Neuroscience. 2018 Aug 10;385:25-37. doi: 10.1016/j.neuroscience.2018.05.052. Epub 2018 Jun 8.

Resting-State Functional Connectivity Underlying Costly Punishment: A Machine-Learning Approach.

Author information

1
College of Information Science and Technology, Beijing Normal University, Beijing, China; State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, China; Key Laboratory of Affective and Social Cognitive Science, Shenzhen University, Shenzhen, China.
2
College of Information Science and Technology, Beijing Normal University, Beijing, China.
3
Key Laboratory of Behavioral Science, Institute of Psychology, Chinese Academy of Sciences, Beijing, China; University of Chinese Academy of Sciences, Beijing, China.
4
College of Information Science and Technology, Beijing Normal University, Beijing, China. Electronic address: wuxia@bnu.edu.cn.
5
Key Laboratory of Affective and Social Cognitive Science, Shenzhen University, Shenzhen, China; Medical School, Kunming University of Science and Technology, Kunming, China; Center for Emotion and Brain, Shenzhen Institute of Neuroscience, Shenzhen, China. Electronic address: luoyj@szu.edu.cn.
6
School of Systems Biology, George Mason University, Fairfax, VA, USA; Department of Psychology, University of Mannheim, Mannheim, Germany.

Abstract

A large number of studies have demonstrated costly punishment to unfair events across human societies. However, individuals exhibit a large heterogeneity in costly punishment decisions, whereas the neuropsychological substrates underlying the heterogeneity remain poorly understood. Here, we addressed this issue by applying a multivariate machine-learning approach to compare topological properties of resting-state brain networks as a potential neuromarker between individuals exhibiting different punishment propensities. A linear support vector machine classifier obtained an accuracy of 74.19% employing the features derived from resting-state brain networks to distinguish two groups of individuals with different punishment tendencies. Importantly, the most discriminative features that contributed to the classification were those regions frequently implicated in costly punishment decisions, including dorsal anterior cingulate cortex (dACC) and putamen (salience network), dorsomedial prefrontal cortex (dmPFC) and temporoparietal junction (mentalizing network), and lateral prefrontal cortex (central-executive network). These networks are previously implicated in encoding norm violation and intentions of others and integrating this information for punishment decisions. Our findings thus demonstrated that resting-state functional connectivity (RSFC) provides a promising neuromarker of social preferences, and bolster the assertion that human costly punishment behaviors emerge from interactions among multiple neural systems.

KEYWORDS:

costly punishment; cross validation; fairness; machine learning; support vector machine; ultimatum game

Supplemental Content

Full text links

Icon for Elsevier Science
Loading ...
Support Center