This paper investigates the complex interplay between memory biases, mood disorders, and social performance. Existing models often focus on individual traits or biases in isolation, neglecting the integrated nature of these factors in real-world social interactions. The research aims to bridge this gap by creating a comprehensive agent-based model that simulates decision-making and social interactions within a network of agents exhibiting different psychological profiles. This model addresses the limitations of previous research by considering mood disorders (mania, depression, bipolar disorder) along with memory biases in an interconnected social network, rather than in isolation. The model's significance stems from its potential to provide insights into how these psychological factors influence an individual's success in social settings and to potentially guide future research into therapeutic interventions. The Continuous Prisoner's Dilemma (CPD) game is utilized to represent a scenario where cooperation levels can range from 0 (complete defection) to 1 (complete cooperation), thus providing a more realistic representation of social interactions than the binary choices of standard IPD.

The paper reviews existing literature on cognitive biases, emotions, mood disorders, and their relationships with memory and decision-making. It highlights inconsistencies in the literature regarding the effect of emotion on memory and emphasizes the lack of integrated models considering multiple factors such as memory biases and mood disorders in social interactions. Existing research focuses either on perfectly rational behavior or considers individual biases in isolation. The authors discuss prior models of mood fluctuation in bipolar disorder, including rhythmic, discrete, and those based on the Behavioral Approach System (BAS). Furthermore, they discuss the phenomenon of mood dependence and the modelling of forgetting, drawing upon existing research utilizing exponential decay and power law functions. The researchers point out the need for a model that integrates the effects of emotion, mood, and memory biases on decision-making and social interactions, acknowledging that most prior computational models of mood disorders assume independent agents without social interactions or networks.

The study employs agent-based modeling to simulate an interconnected social network of agents with varying psychological profiles. Four agent types are defined: rational, manic, depressed, and bipolar. The agents interact in a series of iterated Continuous Prisoner's Dilemma (CPD) games. Agent behavior is guided by an integrated model that combines appraisal theory for emotion, a linear encoding function for memory with separate gradients for positive and negative emotions, a probabilistic retrieval model for mood-dependent memory (using a triangular distribution), and an exponential decay model for forgetting. The model includes parameters to control the rate of mood change, forgetting rate, memory strength, and emotional bias. Agent traits such as reward sensitivity, drive to participate in interactions, and network size are linked to their mood disorder type. Simulations track agent payoffs, mood, and memory traces to analyze the effect of different biases and mood disorders on social performance. The model allows for variations in population composition, connection diversity (particularly for depressed agents), rate of forgetting, emotional bias, and mood dependence.

Simulation results reveal several key findings. Manic agents consistently outperform rational agents, while depressed agents show substantially lower payoffs. Bipolar agents achieve payoffs roughly 10% higher than rational agents on average. The average payoff exhibits an almost-linear relationship with the extent of mania or depression. Stronger memory leads to improved performance across all agent types. For rational agents, positive and negative biases show no significant difference in payoffs. However, in manic and bipolar agents, positive bias is associated with higher payoffs, while in depressed agents, negative bias leads to higher payoffs. The effect of mood dependence varies: an intermediate level is optimal for manic agents, while it's detrimental for depressed agents; weak mood dependence is best for bipolar agents. Increasing the proportion of manic agents in the population results in a decrease in their average payoff, while a higher proportion of bipolar agents leads to increased average payoffs. A moderate level of connection diversity among depressed agents leads to optimal social performance. Agents with stronger memory (lower forgetting rate) achieve significantly better payoffs, although the exact difference in payoffs varies among agent types.

The findings provide valuable insights into the influence of mood disorders and memory biases on social interactions and success. The model's ability to reproduce and extend existing empirical observations supports its validity. For instance, the model successfully replicates clinically observed diminished performance in depressed individuals and improved performance in manic individuals. Furthermore, the study highlights the hitherto unexplored role of mood-dependent memory retrieval in social performance, showing that its impact varies across different mood states. The observed relationship between reward sensitivity (as a measure of mania/depression) and payoff underscores the complex relationship between mood and social behavior. The model helps explain how individual psychological states can result in varying levels of success in social interactions. This supports the need to further integrate this framework to understand and potentially treat mood disorders. The implications of this model extend beyond simply explaining existing observations; it offers a robust framework for exploring the effect of various other traits or biases on social outcomes which may be difficult to study empirically.

This research presents a novel agent-based model integrating memory biases and mood disorders to analyze social performance. Simulation results support and extend existing research on the impact of mood disorders, highlighting the previously unexplored effect of mood-dependent memory retrieval. The model reveals nuanced relationships between emotional bias, memory strength, and mood dependence across different agent types. Future work could extend the model by incorporating additional psychological factors or exploring different interaction scenarios to gain further understanding of the multifaceted dynamics of social interactions. This framework can help in designing interventions that account for the complex interplay of mood and cognition to improve social functioning in individuals with mood disorders.

While the model captures key aspects of memory, mood, and social interaction, it simplifies certain psychological complexities. For instance, the model uses simplified representations of mood disorders and might not capture the full range of their complexities. Furthermore, the Continuous Prisoner's Dilemma (CPD) may not fully capture all aspects of human social interaction. The generalization of the results might be limited by the specific parameter ranges and simulation settings used. Further research could explore the robustness of these findings with more varied simulations and broader parameter ranges.