Speer, S. P., Smidts, A., & Boksem, M. A. (2022).
NeuroImage, 246, 118761.
https://doi.org/10.1016/j.neuroimage.2021.118761
Abstract
Measurement of the determinants of socially undesirable behaviors, such as dishonesty, are complicated and obscured by social desirability biases. To circumvent these biases, we used connectome-based predictive modeling (CPM) on resting state functional connectivity patterns in combination with a novel task which inconspicuously measures voluntary cheating to gain access to the neurocognitive determinants of (dis)honesty. Specifically, we investigated whether task-independent neural patterns within the brain at rest could be used to predict a propensity for (dis)honest behavior. Our analyses revealed that functional connectivity, especially between brain networks linked to self-referential thinking (vmPFC, temporal poles, and PCC) and reward processing (caudate nucleus), reliably correlates, in an independent sample, with participants’ propensity to cheat. Participants who cheated the most also scored highest on several self-report measures of impulsivity which underscores the generalizability of our results. Notably, when comparing neural and self-report measures, the neural measures were found to be more important in predicting cheating propensity.
Significance statement
Dishonesty pervades all aspects of life and causes enormous economic losses. However, because the underlying mechanisms of socially undesirable behaviors are difficult to measure, the neurocognitive determinants of individual differences in dishonesty largely remain unknown. Here, we apply machine-learning methods to stable patterns of neural connectivity to investigate how dispositions toward (dis)honesty, measured by an innovative behavioral task, are encoded in the brain. We found that stronger connectivity between brain regions associated with self-referential thinking and reward are predictive of the propensity to be honest. The high predictive accuracy of our machine-learning models, combined with the reliable nature of resting-state functional connectivity, which is uncontaminated by the social-desirability biases to which self-report measures are susceptible, provides an excellent avenue for the development of useful neuroimaging-based biomarkers of socially undesirable behaviors.
Discussion
Employing connectome-based predictive modeling (CPM) in combination with the innovative Spot-The-Differences task, which allows for inconspicuously measuring cheating, we identified a functional connectome that reliably predicts a disposition toward (dis)honesty in an independent sample. We observed a Pearson correlation between out-of-sample predicted and actual cheatcount (r = 0.40) that resides on the higher side of the typical range of correlations (between r = 0.2 and r = 0.5) reported in previous studies employing CPM (Shen et al., 2017). Thus, functional connectivity within the brain at rest predicts whether someone is more honest or more inclined to cheat in our task.
In light of previous research on moral decisions, the regions we identified in our resting state analysis can be associated with two networks frequently found to be involved in moral decision making. First, the vmPFC, the bilateral temporal poles and the PCC have consistently been associated with self-referential thinking. For example, it has been found that functional connectivity between these areas during rest is associated with higher-level metacognitive operations such as self-reflection, introspection and self-awareness (Gusnard et al., 2001; Meffert et al., 2013; Northoff et al., 2006; Vanhaudenhuyse et al., 2011). Secondly, the caudate nucleus, which has been found to be involved in anticipation and valuation of rewards (Ballard and Knutson, 2009; Knutson et al., 2001) can be considered an important node in the reward network (Bartra et al., 2013). Participants with higher levels of activation in the reward network, in anticipation of rewards, have previously been found to indeed be more dishonest (Abe and Greene, 2014).
