Developing Metrics to Characterise Flickr Groups

Flickr, the large-scale online photo sharing website, is often viewed as one of the ‘classic’ examples of Web2.0 applications. Flickr is becoming the “eyes on the world” by capturing and recording a visual archive of contemporary life on a diverse and extensive range of subjects. Through sites, such as Flickr, researchers are able to observe the social behaviour of online communities and extract, process and analyse content and metadata for a range of purposes. One of the main features of the online photo storing and sharing web site, Flickr, is groups. They provide a means to organise, share and discuss photos of potential interest to group members. This paper explores the scale of group creation on Flickr and proposes a novel scheme for characterising groups in Flickr. It suggests a set of metrics derived partly from previous studies, looking at aspects of membership, communication activity and communication structure. These metrics could be applied more generally than Flickr to capture characteristics of various online groups or communities. To provide the data to examine these metrics we first gathered a population of as many groups as possible, from which we randomly sampled 1,000 groups. Data about the groups sampled were collected using a variety of approaches and analysed with respect to the proposed set of metrics. The results of our analysis provide new insights into group behaviour in Flickr. We find that, in addition to very large groups in terms of members and photos, the data analysis uncovered a large number of small groups with low activity. Most groups, especially large ones, were not dominated by a few individuals as expected. The contributions of this paper include (1) a novel set of metrics for characterising online groups that extend existing schemes; (2) an approach for sampling Flickr to estimate the population of groups based on a form of dictionary-lookup; (3) new insights into Flickr groups based on results from analysing detailed information of 1,000 randomly selected groups; and (4) reflections on our experiences with using publicly accessible data to characterise groups in Flickr and to measure their “groupness’.