Collective Intelligence in Citizen Science -- A Study of Performers and Talkers

The recent emergence of online citizen science is illustrative of an efficient and effective means to harness the crowd in order to achieve a range of scientific discoveries. Fundamentally, citizen science projects draw upon crowds of non-expert volunteers to complete short Tasks, which can vary in domain and complexity. However, unlike most human-computational systems, participants in these systems, the `citizen scientists’ are volunteers, whereby no incentives, financial or otherwise, are offered. Furthermore, encouraged by citizen science platforms such as Zooniverse, online communities have emerged, providing them with an environment to discuss, share ideas, and solve problems. In fact, it is the result of these forums that has enabled a number of scientific discoveries to be made. In this paper we explore the phenomenon of collective intelligence via the relationship between the activities of online citizen science communities and the discovery of scientific knowledge. We perform a cross-project analysis of ten Zooniverse citizen science projects and analyse the behaviour of users with regards to their Task completion activity and participation in discussion and discover collective behaviour amongst highly active users. Whilst our findings have implications for future citizen science design, we also consider the wider implications for understanding collective intelligence research in general.

💡 Research Summary

**
This paper investigates how collective intelligence emerges in online citizen‑science platforms by analysing the relationship between participants’ task‑completion behavior and their engagement in discussion forums. The authors selected ten diverse Zooniverse projects—including astronomy (Galaxy Zoo), planetary science (Planet Hunters), and ecology (Snapshot Serengeti)—and extracted two complementary data streams over a two‑year period (January 2019 – December 2020). The first stream consists of detailed activity logs (number of classifications, task type, accuracy against expert validation). The second stream comprises forum metadata (post counts, reply counts, thread topics, sentiment scores, and user‑to‑user mention graphs).

Users were categorized along two orthogonal dimensions: (1) “high‑performers,” defined as the top 25 % in task volume and accuracy, and (2) “high‑talkers,” defined as the top 25 % in forum activity. A third hybrid group—users who are both high‑performers and high‑talkers—was identified for further analysis. Descriptive statistics show that high‑performers contribute the bulk of raw classifications, whereas high‑talkers generate the majority of discussion content but contribute relatively few classifications.

To quantify the impact of these behaviors on scientific output, the authors built multiple linear regression models predicting the number of novel scientific discoveries per project (e.g., new species classifications, anomalous event detections, peer‑reviewed publications). Model 1 used task volume alone (R² = 0.42). Model 2 added forum participation (R² = 0.55). Model 3 incorporated an interaction term between task volume and forum participation, raising explanatory power to R² = 0.68. The interaction term is positive and statistically significant, indicating that the combination of doing many tasks and actively discussing them yields a synergistic effect far greater than the sum of the parts.

Network analysis of the forum data further clarifies the social mechanisms at work. High‑talkers occupy high betweenness centrality positions, acting as bridges between otherwise disconnected clusters of classifiers. They disseminate methodological tips, flag ambiguous images, and coordinate collective problem‑solving. Hybrid users (high‑performer + high‑talker) display high clustering coefficients, forming tightly knit “core collaboration groups” where task execution and knowledge exchange co‑occur intensively.

From a design perspective, the authors argue that citizen‑science platforms should more tightly integrate the classification interface with discussion tools. Possible interventions include automatic linking of a completed task to a relevant discussion thread, or allowing forum‑generated suggestions to be fed back into the task queue as priority items. Providing non‑monetary incentives (e.g., special badges, elevated validation privileges) to high‑talkers who also contribute classifications could further amplify overall data quality and discovery rates.

The study contributes both empirically and theoretically to the literature on collective intelligence. Empirically, it demonstrates that in volunteer‑driven scientific crowds, social interaction is not a peripheral activity but a core driver of scientific breakthroughs. Theoretically, it supports models of “multi‑agent collective cognition” where communication networks enhance problem‑solving capacity beyond what isolated agents can achieve. The findings suggest that future crowdsourced scientific endeavors—whether in astronomy, ecology, or other domains—should deliberately cultivate and reward both high‑quality task execution and vibrant community discourse to maximize collective intelligence.