Help Converts Newcomers, Not Veterans: Generalized Reciprocity and Platform Engagement on Stack Overflow

Abstract

Generalized reciprocity -- the tendency to help others after receiving help oneself -- is widely theorized as a mechanism sustaining cooperation on online knowledge-sharing platforms. Yet robust empirical evidence from field settings remains surprisingly scarce. Prior studies relying on survey self-reports struggle to distinguish reciprocity from other prosocial motives, while observational designs confound reciprocity with baseline user activity, producing upward-biased estimates. We address these empirical challenges by developing a matched difference-in-differences survival analysis that leverages the temporal structure of help-seeking and help-giving on Stack Overflow. Using Cox proportional hazards models on over 21 million questions, we find that receiving an answer significantly increases a user's propensity to help others, but this effect is concentrated among newcomers and declines with platform experience. This pattern suggests that reciprocity functions primarily as a contributor-recruitment mechanism, operating before platform-specific incentives such as reputation and status displace the general moral impulse to reciprocate. Response time moderates the effect, but non-linearly: reciprocity peaks for answers arriving within a re-engagement window of roughly thirty to sixty minutes. These findings contribute to the theory of generalized reciprocity and have implications for platform design.

Figures (9)

• Figure 1: Study Design: Matched Difference-in-Differences Survival Analysis
• Figure 2: Help rate over the $\pm$4-day observation window, pooled across all tenure buckets. The normalized help rate (relative to each group's pre-question baseline) is plotted for treated users (red, who received an answer) and matched control users (blue, who did not).
• Figure 3: Strength of the generalized reciprocity effect across user experience. Bars show the hazard ratio for the treatment effect by tenure bucket. Error bars indicate 95% confidence intervals.
• Figure 4: Treatment effect (hazard ratio) by response time bin, pooled across all tenure buckets. Each point represents the estimated hazard ratio of the treatment indicator within that response-time bin, relative to the matched control. Error bars indicate 95% confidence intervals.
• Figure 5: Help rate in the post-question window by answer-receipt status, pooled across all tenure buckets. Grey dashed: matched control users who received no answer. Blue: treated users who have not yet received an answer (No answer yet). Red: treated users who have already received an answer (Answer received). Green dashed (right axis): cumulative share of treated users who have received an answer by each hour.