Assessing social anhedonia in a transdiagnostic sample: Insights from a computational psychiatry lens

Abstract

Anhedonia, a reduced ability to experience positive affect and seek rewards, is present across many psychiatric disorders, notably among individuals who experienced trauma. Within the social domain, anhedonia manifests as an altered sense of belonging and social isolation and is associated with poorer clinical outcomes. Yet, mechanistic operationalizations of social anhedonia are lacking, limiting our understanding of the relationship between these mechanisms and affective symptoms. To address these questions, we developed a social reward exploration task which was administered to a transdiagnostic sample of trauma-exposed Veterans (N = 33) while they underwent functional magnetic resonance imaging. The goal was to maximize compliments from two unknown partners, as participants were told these partners selected compliments based on seeing their picture. A Bayesian reinforcement learning modeling approach was used to extract cognitive and neural markers of compliment (reward) exploration. To address these questions, we developed a social reward exploration task which was administered to a transdiagnostic sample of trauma-exposed Veterans (N = 33) while they underwent functional magnetic resonance imaging. The goal was to maximize compliments from two unknown partners, as participants were told these partners selected compliments based on seeing their picture. A Bayesian reinforcement learning modeling approach was used to extract cognitive and neural markers of compliment (reward) exploration. Higher social connectedness (β = 0.51; 95 % CI=[0.11,0.94]) and anxiety (β = 0.57; 95 % CI=[0.13,1.00]) were independently associated with more model-based choices of the partner they anticipated to be most complimenting. In the dorsal anterior cingulate cortex (ACC; z = 3.89, p .001) and left inferior parietal lobule (z = 3.96, p .001), neural responses to reward prediction errors (RPE) were more positive in response to compliment relative to non-compliment outcomes. Greater positive RPE ACC activation was associated with lower anxiety (β = -0.51; 95 % CI=[-0.99,-0.10]. Computational approaches to social reinforcement learning can help identify important neurocognitive differences in social reward sensitivity among individuals with complex affective profiles, such as trauma-exposed individuals. Understanding these differences may help develop new prediction and treatment tools for social anhedonia.

Keywords: Anhedonia; Anxiety; Social anhedonia; Social connectedness; Social reinforcement learning.

Fig. 1

Task, Behavior, and Model Comparison. A. FLIRT trial timeline. Participants completed 20 games, each with 15 trials. On each trial of each game, participants had to assign one token (stacked horizontally at the top of the screen) to one of the two partners. After placing each token, they either earned 1 compliment if the token turned green or zero points if the token turned red. Each trial lasted about 2 s, including participants’ trial reaction time to assign a token and a 800 ms outcome phase showing the outcome once assigned, i.e., a compliment (in green) or a simple “hello” (in white). At the end of each game, participants saw a brief fixation screen (6 s), and the next game followed. At the beginning of the task, participants were instructed to try earning as many compliments as possible in the task. The 40 compliment rates for each partner (2 for each of the 20 games) were experimentally set and sampled from a Beta distribution (α = β = 2). The rates’ order of presentation and assignment to partners picture was randomized across participants. B. Model-agnostic behavior as a function of time across games and participants: rate of choosing the most complimenting partner (based on experimentally set reward rates; left y axis, black line) and rate of choosing partner least explored (right y axis; blue line). C. Individual BIC as a function of tested model. Lower values indicate greater model fit with behavior. Robust estimates (posterior median) and 95 % credible intervals (CI) are represented with black dots and line range, respectively;⋆ = 95 % CI of the difference not including zero, ⋆⋆ = 99 % CI of the difference not including zero. WSLS=Win-Stay/Lose-Shift, DBM=Dynamic Belief Model, FBM=Fixed Belief Model, RW=Rescorla-Wagner.

Fig. 2

Relationship between affective measures and model-based performance. A. Robust estimates (posterior median) and credible intervals (CI) of the independent effects of anhedonia, anxiety, and social connectedness on individual-level model parameters and predictive accuracy (from multivariate Bayesian regression analysis). Anhedonia, anxiety, and social connectedness were measured with reverse-coded Snaith-Hamilton Pleasure Scale (SHAPS), the short Depression Anxiety and Stress Scale (DASS-21) Anxiety scale, and Social Connectedness Scale, respectively. Both 95 % (thinner line) and 90 % (thicker line) are represented. Higher social connectedness (β = 0.51; 95 % CI=[0.11,0.94]) and higher anxiety (β = 0.57; 95 % CI=[0.13,1.00]) were associated with higher DBM predictive accuracy. B. Scatter plots representing the correlation of DBM accuracy with social connectedness (left) and anxiety (right) are provided for illustrative purposes. Affective measures are revisualized with respect to the two other ones. C. Participants’ rate of choosing the partner estimated to be most complimenting as a function of dichotomized social connectedness (blue) and anxiety (green) levels (darker/lighter shade=higher/low level, respectively) as a function of time.

Fig. 3

Regions of activation correlating with positive reward prediction errors. Axial slices (numbers represent z coordinate). R=Right, L=Left.

Fig. 4

Differential activation to positive vs negative reward prediction errors. A. Activation clusters in the dorsal ACC (Peak Voxel MNI Coordinates: x = −1, y = −1, z = 49) and left IPL (Peak Voxel MNI Coordinates: x = −41, y = −34, z = 52) associated with significant difference in RPE activation as a function of outcome type (compliment vs no compliment). B. Activation patterns (individual coefficient z statistic) in these 2 clusters as a function of outcome type; significantly greater activations were observed in response to compliment relative to no compliment (“pass”) outcomes in the ACC (Cohen’s d =0.75) and left IPL (Cohen’s d = 0.91); brown dots and lines represent robust posterior estimates and 95 % credible interval (CI); ACC=anterior cingulate cortex; IPL=inferior parietal lobule; MNI=Montreal Neurologic Institute; RPE=reward prediction error.

Fig. 5

Relationship between affective measures and positive RPE activation in ACC and left IPL. A. Robust estimates (posterior median) and credible intervals (CI) for the independent effects of anhedonia, anxiety, and social connectedness on individuals’ average positive RPE activation (from multivariate Bayesian regression analysis). Anhedonia, anxiety, and social connectedness were measured with reverse-coded Snaith-Hamilton Pleasure Scale (SHAPS), the short Depression Anxiety and Stress Scale (DASS-21) Anxiety scale, and Social Connectedness Scale, respectively. Both 95 % (thinner line) and 90 % (thicker line) are represented. Higher anxiety was associated with lower positive RPE activation in the dACC (β = −0.51; 95 % CI=[−0.99,−0.10]). B. For visualization purposes, we provide a scatterplot of ACC positive RPE activation (Z score) as a function of anxiety (residualized with respect to the two other affective measures). Note: After removing one potential influential outlier with elevated Cook’s distance of.79 (lower right corner of this graph), the multivariate regression effect of anxiety on dACC positive RPE activation was reduced but remained within credible interval (β = −0.36; 95 % CI=[−0.72,−0.01]).