A behavioral signature for quantifying the social value of interpersonal relationships with specific others

The inherent human need for social connection and strong relationships is widely acknowledged, linking a lack thereof to negative mental and physical health outcomes. While existing theories in behavioral science often emphasize the value of individual actions, a robust method for quantifying the value of interpersonal relationships remains elusive. This study addresses this gap by proposing a novel approach grounded in economic principles, specifically the concept of opportunity cost. The central research question revolves around whether a behavioral signature can be developed and validated to effectively measure the social value assigned to specific interpersonal relationships. The study's importance lies in its potential to provide mechanistic specificity to value-based behavioral theories by explicitly incorporating the value of relationships, rather than treating them as a mere context, thereby enriching our understanding of human social behavior. The unique contribution of this research lies in its data-driven, quantitative method which utilizes individual behavior to create a more accurate measure than relying solely on self-reported attitudes or subjective valuations.

Prior research has successfully quantified the value of social phenomena like information sharing and interactions. However, less attention has been paid to formally quantifying the value of specific interpersonal relationships. Existing models often include general contextual factors, failing to isolate the value of individual relationships. Other approaches focus on quantifying altruistic behavior, such as willingness to sacrifice personal monetary gain for another's benefit, but these measures are susceptible to transient affective states and socioeconomic background, limiting their reliability and generalizability as indicators of relationship value. This study aims to overcome these limitations by directly measuring value based on how individuals allocate scarce resources (time and energy) when interacting with specific relationship partners, thereby avoiding subjective biases and external factors influencing previous measurement attempts.

This study comprises two phases: exploratory and confirmatory. The exploratory phase involved three procedures: (i) sourcing activities, (ii) deriving activity weights using Maximum Difference (MaxDiff) scaling (a best-worst scaling approach), and (iii) validating social value scores. Activities were sourced from two independent samples (UCLA's SONA and Amazon's Mechanical Turk, MTurk) resulting in two sets of activities: one with 70 items and another with 56. MaxDiff scaling was used to derive weights reflecting the relative priority individuals assign to each activity under conditions of time scarcity. The weights were obtained using hierarchical Bayesian logistic regression to account for individual differences in preferences. Social value scores were then calculated by taking the dot product of the activity weights and participants’ likelihood ratings for engaging in those activities with specific relationship partners (parents, friends, acquaintances). These likelihood ratings were collected using a free-day scenario prompt, assuming easy access to each partner. The exploratory phase validation included correlating social value scores with relationship quality, social loss aversion, and time spent with each social partner. The confirmatory phase replicated the analyses in a larger, independent sample (N=635) from MTurk and Prolific, adding a multi-trial social decision-making task and measures of affiliative social behaviors to broaden the validation scope. Bayesian statistical methods were used throughout the study to handle individual differences and account for uncertainty in estimates. The study was pre-registered (except for some minor deviations detailed in the supplementary materials). Data were collected using Qualtrics and analyzed using the brms package in R. Due to a high rate of fraudulent responses on MTurk in the confirmatory phase, data collection shifted to Prolific, with the cleaned MTurk data retained for generalizability testing.

The exploratory phase revealed that social value scores were significantly correlated with relationship quality for friends and acquaintances, and showed a less consistent, but generally positive correlation with parent relationship quality. Social value scores were also positively correlated with social loss aversion for all three relationship types. However, the correlations with actual and ideal time spent with partners were less consistent, suggesting that social value isn't simply a function of time investment. Regression analyses revealed that social value scores predicted social decision behaviors, particularly in the dictator game, where higher social value for a partner was associated with greater resource allocation to that partner. In the confirmatory phase, with a substantially larger sample size, the correlations between social value scores and relationship quality as well as social loss aversion were generally stronger and more consistent across all three relationship types (parents, friends, acquaintances). The confirmatory phase also demonstrated that social value scores predicted choices in both one-shot and multi-trial social decision-making tasks, further reinforcing the predictive validity of the measure. Importantly, even when controlling for unit-weighted likelihood ratings, social value scores maintained a significant association with relationship quality and social loss aversion, and largely with dictator game choices. The findings on the forced choice questions concerning time allocation, however, changed upon controlling for unit-weighted likelihood scores, meaning that social value did not predict these behaviors independently of the overall willingness to spend time with the social partners. Post-hoc analyses provided evidence for the reliability of the likelihood ratings, confirmed the importance of the activity weights in constructing the signature (versus randomly permuted weights), and demonstrated good discriminant validity of the social value scores. While social value was associated with social behaviors, post-hoc analysis showed relationship quality and social loss aversion were slightly stronger predictors.

The study's findings successfully demonstrate the validity and reliability of a novel method for quantifying the social value of interpersonal relationships. This behavioral signature approach, rooted in economic principles, provides a more nuanced measure compared to previous reliance on self-reported attitudes or resource allocation tasks that are susceptible to biases. The results support the notion that individuals assign value to specific relationship partners, which influences their social decisions and behaviors. The significant correlations between social value scores and relationship quality highlight the importance of incorporating behavioral choices when studying relationships. The somewhat weaker association compared to relationship quality and social loss aversion could reflect the multi-faceted nature of interpersonal relationships, with social value representing a specific, yet important, dimension influencing social behavior. The relative predictive strength of social loss aversion over relationship quality suggests potentially interesting avenues for future research into asymmetric value processing in social contexts. The method’s simplicity and adaptability make it a valuable tool for broader application in the behavioral sciences.

This study successfully develops and validates a behavioral signature for quantifying the social value of interpersonal relationships. The method's strength lies in its use of readily available behavioral data and straightforward calculation procedure, making it easily replicable and adaptable. Future research could extend this method by incorporating goal-oriented behaviors, modulating likelihood ratings by enjoyment levels, and personalizing activity weights to further enhance its precision and generalizability. Cross-cultural studies are crucial to examine the method's applicability across diverse societal contexts.

The study primarily focuses on Western samples (WEIRD), potentially limiting the generalizability of the findings to other cultures. The current method primarily considers the prioritization of activities, not other potential contributors to social value, such as goal fulfillment. The use of self-reported data, while providing a practical method, is susceptible to biases inherent in self-reporting. While the study addressed potential confounds such as time spent with a partner, additional research is needed to explore more potential confounds or factors that may influence social valuation of relationships and to verify the results using different measurement methods.