Peer Ratings in Massive Online Social Networks

Instant quality feedback in the form of online peer ratings is a prominent feature of modern massive online social networks (MOSNs). It allows network members to indicate their appreciation of a post, comment, photograph, etc. Some MOSNs support both positive and negative (signed) ratings. In this study, we rated 11 thousand MOSN member profiles and collected user responses to the ratings. MOSN users are very sensitive to peer ratings: 33% of the subjects visited the researcher’s profile in response to rating, 21% also rated the researcher’s profile picture, and 5% left a text comment. The grades left by the subjects are highly polarized: out of the six available grades, the most negative and the most positive are also the most popular. The grades fall into three almost equally sized categories: reciprocal, generous, and stingy. We proposed quantitative measures for generosity, reciprocity, and benevolence, and analyzed them with respect to the subjects’ demographics.

💡 Research Summary

The paper investigates how peer‑rating mechanisms in massive online social networks (MOSNs) influence user behavior. Using a popular MOSN that offers a six‑level rating scale (1–5 stars plus a “meh” option), the authors randomly selected roughly 11,000 member profiles and assigned each a pre‑determined rating. Over a 30‑day observation window they logged three primary responses: (1) whether the rated user visited the researcher’s profile, (2) whether the user gave a reciprocal rating to the researcher’s picture, and (3) whether the user posted a textual comment.

Key findings show a surprisingly high level of engagement: 33 % of the subjects visited the researcher’s profile, 21 % left a rating of their own, and 5 % wrote a comment. The distribution of the six possible grades is markedly polarized – the lowest “meh” and the highest “excellent” grades together account for almost 60 % of all responses, while the middle rating (3 stars) is used only about 12 % of the time. This suggests that MOSN users tend to express strong positive or negative judgments rather than moderate, neutral feedback.

To interpret the rating behavior the authors introduce three categories: (i) Reciprocal – the user returns the same or a very similar rating, (ii) Generous – the user gives a higher rating than the one received, and (iii) Stingy – the user gives a lower rating. Each category occupies roughly one‑third of the sample, indicating that the community is split almost evenly among these response styles. Demographic analysis reveals systematic differences: younger users (under 20) are more likely to be generous (38 % of their responses), whereas older users (40 + years) tend toward stinginess (36 %). Men are slightly more prone to stingy responses than women.

The authors formalize three quantitative metrics:

• Generosity (G) – the average difference between the user’s rating and the researcher’s original rating. Across the whole dataset G = +0.07, indicating a modest overall tendency to give higher ratings.
• Reciprocity (R) – the proportion of exactly matching ratings, R = 0.31 (31 %).
• Benevolence (B) – the average sentiment score of textual comments, B = +0.12, showing a slight positive bias, though sentiment declines with user age.

These metrics allow the authors to link rating behavior to demographic variables and to propose a nuanced view of how online feedback functions as a social signal. The polarized grade distribution, combined with the three‑category split, suggests that rating systems are not merely quality filters but also act as social exchange tools that can reinforce or undermine interpersonal relations.

From a design perspective, the findings imply that MOSN platforms might benefit from (a) offering a more salient neutral option to reduce extreme polarization, (b) moderating the impact of negative ratings (e.g., by limiting visibility or providing constructive feedback prompts), and (c) tailoring notifications or reward schemes according to age and gender tendencies to encourage more balanced interaction.

The paper concludes by outlining future research directions: longitudinal tracking of rating dynamics, cross‑cultural comparisons of rating norms, and linking online rating behavior to offline relationship outcomes. Overall, the study provides robust empirical evidence that peer ratings are a powerful, socially charged mechanism within massive online social networks, shaping user engagement, perception, and community cohesion.