Abstract
Social comparison widely exists, which involves individuals evaluating themselves in comparison to others. Although this process may provide useful information, there are instances where the comparison holds no relevance to one's outcome. The effect of those irrelevant social comparisons on decision-making under uncertainty is not well understood. The current study examined the effect of those irrelevant social comparisons on uncertainty exploration and the mechanisms behind it. Participants were recruited to engage in a two-player four-armed restless bandit task. The results revealed that presenting irrelevant outcomes of others affected participants' exploration behavior significantly, mainly reflected in directed exploration and perseveration. Social comparison can affect the confidence of the participants but not reward predictions. Additionally, irrelevant social comparisons do not affect exploration strategies by changing subjective expectations. Our work provides insights into modeling complex social interaction and extends valuable implications across social situations with the effect of social comparison on decisions.
Introduction
Humans are fundamentally social creatures, constantly evaluating their own standing relative to others. Traditional social comparison theory posits that we engage in this behavior to obtain information for self-evaluation. However, in modern contexts—such as observing a peer's success on social media—we often compare ourselves even when their outcomes have no causal link to our own. This study investigates how these "irrelevant" social comparisons influence the exploration-exploitation trade-off: the choice between sticking with a known reward (exploitation) or trying an uncertain option (exploration).
Methods and Experimental Design
The researchers conducted three experiments using a four-armed restless bandit task. In this paradigm, participants choose between four options whose reward values drift over time. To maximize points, players must balance exploiting high-value arms with exploring others to detect shifts in reward distribution.
In Experiments 1 and 2, participants played alongside a partner (an algorithm). Crucially, while they saw their partner's rewards, the outcomes were mathematically independent. Experiment 2 included "inquiry sessions" to measure reward predictions and subjective confidence. Experiment 3 served as a control, manipulating "subjective expectations" without a social partner to isolate the specific effects of social comparison.
To analyze the underlying mechanisms, the team employed a Bayesian learning model (Kalman Filter) to track belief updates. They then tested several softmax-based choice models, incorporating factors such as Value (V), Directed Exploration (E), and Perseveration (P).
Key Results
The study found that social comparison fundamentally warps exploration strategies. When participants experienced a "social loss" (earning less than their partner), they were significantly more likely to engage in directed exploration, specifically targeting options with high uncertainty. Conversely, "social gain" (earning more than a partner) led to increased perseveration, where participants were more likely to repeat their previous choice regardless of its objective value.
Quantitatively, the proportion of exploration in social gain trials was 33.6%, compared to 58.0% in social loss trials ($P < 0.001$). Computational modeling confirmed these findings; the "GLM-VEP" model, which allowed social comparison to influence exploration and perseveration parameters, significantly outperformed standard models. Interestingly, while social comparison did not alter the accuracy of reward predictions, it significantly impacted participants' confidence. Social loss reduced confidence, suggesting that the drive to explore is fueled by a subjective sense of uncertainty rather than a change in objective knowledge.
Discussion
These findings suggest that social comparison operates through unique psychological mechanisms, likely emotional or motivational, rather than simply acting as a proxy for reward expectations. While social loss triggers curiosity and a search for better alternatives, social gain breeds a form of behavioral inertia.
Future research is needed to map the specific emotional mediators, such as envy or pride, and to determine if these effects persist with subtle or subconscious social cues. Understanding how irrelevant comparisons influence decision-making has broad implications for social interaction and economic behavior in an increasingly connected world.
Summary updated