A probabilistic map of emotional experiences during competitive social interactions

Statistical learning across cognitive and affective domains: a multidimensional review

Statistical learning (SL) is a fundamental cognitive ability enabling individuals to detect and exploit regularities in environmental input. It plays a crucial role in language acquisition, perceptual processing, and social learning, supporting development from infancy through adulthood. In this review, we adopt a multidimensional perspective to synthesize empirical and theoretical findings on SL, covering experimental paradigms, developmental trajectories, and neural mechanisms. Furthermore, we extend the discussion to the emerging intersection between SL and affective processes. Although emotional factors have recently been proposed to modulate SL performance, this area remains underexplored. We highlight current insights and theoretical frameworks addressing the SL-emotion interaction, such as predictive coding theory, and propose directions for future research. This review provides a comprehensive yet focused overview of SL across cognitive and affective domains, aiming to clarify the scope and future potential of this growing field.

Dynamic brain connectivity predicts emotional arousal during naturalistic movie-watching

Human affective experience varies along the dimensions of valence (positivity or negativity) and arousal (high or low activation). It remains unclear how these dimensions are represented in the brain and whether the representations are shared across different individuals and diverse situational contexts. In this study, we first utilized two publicly available functional MRI datasets of participants watching movies to build predictive models of moment-to-moment emotional arousal and valence from dynamic functional brain connectivity. We tested the models by predicting emotional arousal and valence both within and across datasets. Our results revealed a generalizable arousal representation characterized by the interactions between multiple large-scale functional networks. The arousal representation generalized to two additional movie-watching datasets with different participants viewing different movies. In contrast, we did not find evidence of a generalizable valence representation. Taken together, our findings reveal a generalizable representation of emotional arousal embedded in patterns of dynamic functional connectivity, suggesting a common underlying neural signature of emotional arousal across individuals and situational contexts. We have made our model and analysis scripts publicly available to facilitate its use by other researchers in decoding moment-to-moment emotional arousal in novel datasets, providing a new tool to probe affective experience using fMRI.

Distinct neural computations scale the violation of expected reward and emotion in social transgressions

Traditional decision-making models conceptualize humans as adaptive learners utilizing the differences between expected and actual rewards (prediction errors, PEs) to maximize outcomes, but rarely consider the influence of violations of emotional expectations (emotional PEs) and how it differs from reward PEs. Here, we conducted a fMRI experiment (n = 43) using a modified Ultimatum Game to examine how reward and emotional PEs affect punishment decisions in terms of rejecting unfair offers. Our results revealed that reward relative to emotional PEs exerted a stronger prediction to punishment decisions. On the neural level, the left dorsomedial prefrontal cortex (dmPFC) was strongly activated during reward receipt whereas the emotions engaged the bilateral anterior insula. Reward and emotional PEs were also encoded differently in brain-wide multivariate patterns, with a more sensitive neural signature observed within fronto-insular circuits for reward PE. We further identified a fronto-insular network encompassing the left anterior cingulate cortex, bilateral insula, left dmPFC and inferior frontal gyrus that encoded punishment decisions. In addition, a stronger fronto-insular pattern expression under reward PE predicted more punishment decisions. These findings underscore that reward and emotional violations interact to shape decisions in complex social interactions, while the underlying neurofunctional PEs computations are distinguishable.

Language-specific representation of emotion-concept knowledge causally supports emotion inference

Humans no doubt use language to communicate about their emotional experiences, but does language in turn help humans understand emotions, or is language just a vehicle of communication? This study used a form of artificial intelligence (AI) known as large language models (LLMs) to assess whether language-based representations of emotion causally contribute to the AI's ability to generate inferences about the emotional meaning of novel situations. Fourteen attributes of human emotion concept representation were found to be represented by the LLM's distinct artificial neuron populations. By manipulating these attribute-related neurons, we in turn demonstrated the role of emotion concept knowledge in generative emotion inference. The attribute-specific performance deterioration was related to the importance of different attributes in human mental space. Our findings provide a proof-in-concept that even an LLM can learn about emotions in the absence of sensory-motor representations and highlight the contribution of language-derived emotion-concept knowledge for emotion inference.

The association between social rewards and anxiety: Links from neurophysiological analysis in virtual reality and social interaction game

Individuals' affective experience can be intricate, influenced by various factors including monetary rewards and social factors during social interaction. However, within this array of factors, divergent evidence has been considered as potential contributors to social anxiety. To gain a better understanding of the specific factors associated with anxiety during social interaction, we combined a social interaction task with neurophysiological recordings obtained through an anxiety-elicitation task conducted in a Virtual Reality (VR) environment. Employing inter-subject representational similarity analysis (ISRSA), we explored the potential linkage between individuals' anxiety neural patterns and their affective experiences during social interaction. Our findings suggest that, after controlling for other factors, the influence of the partner's emotional cues on individuals' affective experiences is specifically linked to their neural pattern of anxiety. This indicates that the emergence of anxiety during social interaction may be particularly associated with the emotional cues provided by the social partner, rather than individuals' own reward or prediction errors during social interaction. These results provide further support for the cognitive theory of social anxiety and extend the application of VR in future cognitive and affective studies.

An evaluation of the Ultimatum Game as a measure of irritability and anger

The Ultimatum Game is an effective tool for understanding how social decision-making is influenced by emotions in both research and clinical settings. Previous findings have shown that the Ultimatum Game can evoke negative emotions, especially anger and aggression. In a sample of non-clinical adults (N = 143) we evaluated the sensitivity of an anger-infused version of the Ultimatum Game to individual differences in anger and irritability. Findings showed significant relationships between anger and aggressive behaviors in the Ultimatum game, but no association between irritability and aggressive behavior were observed. This indicates that the anger-infused Ultimatum Game is a promising method for studying individual differences in trait anger and anger expression. However, the relationship between decision-making in the anger-infused Ultimatum Game and irritability is less straight forward and needs further investigation. Therefore, when studying the behavioral responses of irritability, it would be beneficial to capture other behaviors beyond aggressive responses.

Prosocial emotions predict individual differences in economic decision-making during ultimatum game with dynamic reciprocal contexts

Social decision-making is known to be influenced by predictive emotions or the perceived reciprocity of partners. However, the connection between emotion, decision-making, and contextual reciprocity remains less understood. Moreover, arguments suggest that emotional experiences within a social context can be better conceptualised as prosocial rather than basic emotions, necessitating the inclusion of two social dimensions: focus, the degree of an emotion's relevance to oneself or others, and dominance, the degree to which one feels in control of an emotion. For better representation, these dimensions should be considered alongside the interoceptive dimensions of valence and arousal. In an ultimatum game involving fair, moderate, and unfair offers, this online study measured the emotions of 476 participants using a multidimensional affective rating scale. Using unsupervised classification algorithms, we identified individual differences in decisions and emotional experiences. Certain individuals exhibited consistent levels of acceptance behaviours and emotions, while reciprocal individuals' acceptance behaviours and emotions followed external reward value structures. Furthermore, individuals with distinct emotional responses to partners exhibited unique economic responses to their emotions, with only the reciprocal group exhibiting sensitivity to dominance prediction errors. The study illustrates a context-specific model capable of subtyping populations engaged in social interaction and exhibiting heterogeneous mental states.

Hidden Reward: Affect and Its Prediction Errors as Windows Into Subjective Value

Scientists increasingly apply concepts from reinforcement learning to affect, but which concepts should apply? And what can their application reveal that we cannot know from directly observable states? An important reinforcement learning concept is the difference between reward expectations and outcomes. Such reward prediction errors have become foundational to research on adaptive behavior in humans, animals, and machines. Owing to historical focus on animal models and observable reward (e.g., food or money), however, relatively little attention has been paid to the fact that humans can additionally report correspondingly expected and experienced affect (e.g., feelings). Reflecting a broader "rise of affectivism," attention has started to shift, revealing explanatory power of expected and experienced feelings-including prediction errors-above and beyond observable reward. We propose that applying concepts from reinforcement learning to affect holds promise for elucidating subjective value. Simultaneously, we urge scientists to test-rather than inherit-concepts that may not apply directly.

Naturalistic reinforcement learning

Humans possess a remarkable ability to make decisions within real-world environments that are expansive, complex, and multidimensional. Human cognitive computational neuroscience has sought to exploit reinforcement learning (RL) as a framework within which to explain human decision-making, often focusing on constrained, artificial experimental tasks. In this article, we review recent efforts that use naturalistic approaches to determine how humans make decisions in complex environments that better approximate the real world, providing a clearer picture of how humans navigate the challenges posed by real-world decisions. These studies purposely embed elements of naturalistic complexity within experimental paradigms, rather than focusing on simplification, generating insights into the processes that likely underpin humans' ability to navigate complex, multidimensional real-world environments so successfully.

SOMAScience: A Novel Platform for Multidimensional, Longitudinal Pain Assessment

Chronic pain is one of the most significant health issues in the United States, affecting more than 20% of the population. Despite its contribution to the increasing health crisis, reliable predictors of disease development, progression, or treatment outcomes are lacking. Self-report remains the most effective way to assess pain, but measures are often acquired in sparse settings over short time windows, limiting their predictive ability. In this paper, we present a new mobile health platform called SOMAScience. SOMAScience serves as an easy-to-use research tool for scientists and clinicians, enabling the collection of large-scale pain datasets in single- and multicenter studies by facilitating the acquisition, transfer, and analysis of longitudinal, multidimensional, self-report pain data. Data acquisition for SOMAScience is done through a user-friendly smartphone app, SOMA, that uses experience sampling methodology to capture momentary and daily assessments of pain intensity, unpleasantness, interference, location, mood, activities, and predictions about the next day that provide personal insights into daily pain dynamics. The visualization of data and its trends over time is meant to empower individual users' self-management of their pain. This paper outlines the scientific, clinical, technological, and user considerations involved in the development of SOMAScience and how it can be used in clinical studies or for pain self-management purposes. Our goal is for SOMAScience to provide a much-needed platform for individual users to gain insight into the multidimensional features of their pain while lowering the barrier for researchers and clinicians to obtain the type of pain data that will ultimately lead to improved prevention, diagnosis, and treatment of chronic pain.

Resting-state network predicts the decision-making behaviors of the proposer during the ultimatum game

Objective. The decision-making behavior of the proposer is a key factor in achieving effective and equitable maintenance of social resources, particularly in economic interactions, and thus understanding the neurocognitive basis of the proposer's decision-making is a crucial issue. Yet the neural substrate of the proposer's decision behavior, especially from the resting-state network perspective, remains unclear.Approach. In this study, we investigated the relationship between the resting-state network and decision proposals and further established a multivariable model to predict the proposers' unfair offer rates in the ultimatum game.Main results.The results indicated the unfair offer rates of proposers are significantly related to the resting-state frontal-occipital and frontal-parietal connectivity in the delta band, as well as the network properties. And compared to the conservative decision group (low unfair offer rate), the risk decision group (high unfair offer rate) exhibited stronger resting-state long-range linkages. Finally, the established multivariable model did accurately predict the unfair offer rates of the proposers, along with a correlation coefficient of 0.466 between the actual and predicted behaviors.Significance. Together, these findings demonstrated that related resting-state frontal-occipital and frontal-parietal connectivity may serve as a dispositional indicator of the risky behaviors for the proposers and subsequently predict a highly complex decision-making behavior, which contributed to the development of artificial intelligence decision-making system with biological characteristics as well.

Differentially private knowledge transfer for federated learning

Extracting useful knowledge from big data is important for machine learning. When data is privacy-sensitive and cannot be directly collected, federated learning is a promising option that extracts knowledge from decentralized data by learning and exchanging model parameters, rather than raw data. However, model parameters may encode not only non-private knowledge but also private information of local data, thereby transferring knowledge via model parameters is not privacy-secure. Here, we present a knowledge transfer method named PrivateKT, which uses actively selected small public data to transfer high-quality knowledge in federated learning with privacy guarantees. We verify PrivateKT on three different datasets, and results show that PrivateKT can maximally reduce 84% of the performance gap between centralized learning and existing federated learning methods under strict differential privacy restrictions. PrivateKT provides a potential direction to effective and privacy-preserving knowledge transfer in machine intelligent systems.

Social preferences trump emotions in human responses to unfair offers

People commonly reject unfair offers even if this leaves them worse off. Some explain this as a rational response based on social preferences. Others argue that emotions override self-interest in the determination of rejection behavior. We conducted an experiment in which we measured responders' biophysical reactions (EEG and EMG) to fair and unfair offers. We measured biophysical trait anger using resting-state EEG (frontal alpha-asymmetry), state anger using facial expressions, offer expectancy processing using event-related EEG (medial-frontal negativity; MFN) and self-reported emotions. We systematically varied whether rejections led proposers to lose their share (Ultimatum Game; UG) or not (Impunity Game; IG). Results favor preference-based accounts: Impunity minimizes rejection despite increasing subjectively reported anger. Unfair offers evoke frowning responses, but frowning does not predict rejection. Prosocial responders reject unfair UG offers more often after unmet fairness expectations. These results suggest that responders do not reject unfairness out of anger. Rather, people seem motivated to reject unfair offers when they violate their behavioral code but only when rejection has payoff consequences for the proposer, allowing them to reciprocate and restore equity. Thus, social preferences trump emotions when responding to unfair offers.

Conceptual representations of uncertainty predict risky decision-making

Decisions made under uncertainty often are considered according to their perceived subjective value. We move beyond this traditional framework to explore the hypothesis that conceptual representations of uncertainty influence risky choice. Results reveal that uncertainty concepts are represented along a dimension that jointly captures probabilistic and valenced features of the conceptual space. These uncertainty representations predict the degree to which an individual engages in risky decision-making. Moreover, we find that most individuals have two largely distinct representations: one for uncertainty and another for certainty. In contrast, a minority of individuals exhibit substantial overlap between their representations of uncertainty and certainty. Together, these findings reveal the relationship between the conceptualization of uncertainty and risky decisions.

Investigating attentional scope as a novel indicator of emotional state in animals

In humans, contrasting emotional states can lead to a broadening or narrowing of attentional scope. Whether this is also the case in animals has yet to be investigated. If confirmed, measurement of attentional scope has potential as a novel cognitive method of welfare assessment. In this study, we therefore aimed to investigate a test of attentional scope as a measure of emotional state in animals. We did this by inducing four putatively different emotional states in dogs (N = 10), varying in valence (positive, negative) and arousal (high, low), in two different reward contexts (food rewards in Experiment 1, social rewards in Experiment 2) and then assessing dogs' behavioural responses in a test of attentional scope. We also recorded heart rate variability (HRV) parameters as additional confirmatory affective indicators. In Experiment 1, the dogs showed a narrowing of attentional scope after the induction of both positively valenced emotional states. That dogs were in a positive state was supported by the reduced Standard Deviation of normal-to-normal R-R intervals (SDNN) and the reduced Low Frequency (LF) and Very Low Frequency (VLF) HRV. In Experiment 2, when responses to social rewards were examined, we did not detect any statistically significant differences in attentional scope between the emotional states, but dogs had a slightly narrow attentional scope in the negatively valenced emotional states. The LF tended to be reduced in the high arousal positive treatment. In conclusion, our study provides the first indication that emotional states can also alter attentional scope in animals. The results justify further investigation of this approach for use in animal welfare assessment, although additional studies are needed to refine predictions.