Communications of the ACM

Virtual extension

Why Do People Tag? Motivations For Photo Tagging

By Oded Nov, Chen Ye
Communications of the ACM, July 2010, Vol. 53 No. 7, Pages 128-131
10.1145/1785414.1785450
Comments

Tagging, or using keywords to add metadata to shared content,⁴ is gaining much popularity in recent years.^3,4,8 Tags are used to annotate various types of content, including images, videos, bookmarks, and blogs, through web-based systems such as Flickr, YouTube, del.icio.us, and Technorati, respectively. The popularity of tagging is attributed, at least in part, to the benefits users gain from effective sharing and from organization of very large amounts of information.^2,3

As tagging receives increasing attention in both research and business communities, studies have found that users vary substantially in their tag usage, and suggested several factors that motivate user tagging.^2,8 However, to date no quantitative study has assessed the strength of the effects of each motivation on levels of tag usage. This is surprising, since user participation is critical to the sustainability of content sharing communities, and a collaborative tagging system cannot succeed without higher level of user contribution.⁶ In what follows, we address this gap, by studying the strength of relationships between several motivations and users' tagging levels on Flickr, a prominent Web 2.0 photo sharing community.²

Currently, there are more than 35 million Flickr users, who have so far uploaded more 3 billion photos. Each Flickr user can upload images and make them viewable by self, by designated friends and family, or by all Flickr users. Flickr users can annotate images with tags - unstructured textual labels; and usually images are tagged only by the user who uploaded them.8 These tags make the images searchable by the uploading user, as well as by others.^2,10 In addition, each user can designate other users as "contacts," people whose photos the user follows (contacts are often reciprocal).

To understand what underlies tagging, we need to find out what motivates sharing in online environments,^5,9 and in particular, what motivates tagging. Furthermore, we need to measure the degree to which different motivations affect tagging activity. While some studies^2,11 has identified individual-level motivations for tagging, other studies have looked solely at the social level, focusing on the social presence as a driver of tagging.⁷

Back to Top

Individual-level Motivations

In their study of the motivation for tagging on Flickr, Ames and Naaman² draw the distinction between motivations stemming from three categories of intended audience of the tags: self, family & friends, and the general public of Flickr users. An additional division is based on the function of the tags, or the tagger's intended use. Here, the authors² distinguish between the function of Organization and the function of Communication. The Organization function involves organization and future retrieval of images, while the Communication function involves communication of additional context to viewers of the image.

The Self category involves the organization function, emphasizing organization and order, which are intended to facilitate future search and retrieval, and the communication function, which involves adding context to the image, for example, by tagging with the names of the people that appear in the photo, or the name of the place where it was taken.

The Family & Friends category also involves the organization function, which is intended in this case to facilitate future search and retrieval by friends and family. In addition, the category involves the communication function, which in this case emphasizes adding context to the image not only by including names of people and places, but also by adding details known only to the intended viewers, such as inside jokes and nicknames.

The Public category involves the organization function, which is intended to help the general public of Flickr users find the images. It also involves the communication function, which in this case emphasizes signaling photographic abilities, giving the photographer the satisfaction of knowing that his or her photos are getting attention, and gaining reputation in the general Flickr community.

Based on Ames and Naaman's findings,² we expect to find that the level of the users' motivations will be positively correlated with their tagging level.

Social presence as a driver of tagging. According to research in social psychology, behavior is affected by presence—actual, imagined, or implied—of others.¹ Moreover, perceived social presence was found to have a positive effect on tagging in del.icio.us, an online bookmark management system in which tagging is used extensively.⁷ In Flickr, "contacts" represent social presence because the user is aware that his or her contacts might be viewing the user's images. Therefore, we expect that the more contacts a user has, the higher the user will perceive social presence, which in turn, will translate into increased tagging activity.

Other potential drivers of tagging, which can serve as control variables, are the number of images a user has and the number of years a user has been using Flickr. We hypothesize that the more images a user has, and the more years he or she has been on Flickr, the more tags he or she will have.

Much of the research to date on tagging motivations has been qualitative.^2,11 In the present study, we evaluate quantitatively the effects of taggers' individual and social motivations on actual tagging levels, by drawing on independent sources such as taggers' self-reported motivations, as well as system data of their actual tagging behavior.

Back to Top

Survey and System Data Collection

A web-based survey was administered to Flickr users, using a combination of user-reported data and independent system data of the actual tagging and image uploading behavior of the user.

To measure tagging motivations, we have developed a scale based on Ames and Naaman's² qualitative work. The scale includes three constructs, representing the three categories of intended users of the tags as perceived by the user: Self, Family & Friends, and Public. For each construct, we have included items representing both the communication and the organization functions. All of the motivation items in the questionnaire were presented as statements to which taggers were asked to state how strongly they agree, on a scale of 1 to 7. Table 1 contains examples of such questionnaire items. After the preliminary scale was developed, a pilot study (N = 193) was carried out to validate the scale and some of the items were dropped, while items showing higher than 0.6 factor loading and lower than 0.4 cross-loadings in a principle component analysis (PCA) were retained. The final scale contains 4, 6, and 6 items for Self, Family & Friends, Public, respectively, and was used in the survey.^a

The dependent variable used was the number of distinct tags a user has. The number of distinct tags reflects a users' tag vocabulary size and is used as a measure of tag usage in prior studies on Flickr.⁸ System data, such as number of photos and distinct tags per user, is available via Flickr's API (Application Programming Interface) system upon log in. The Flickr API allows third party websites to communicate with Flickr and exchange information. Respondents were asked, at the end of our web-based survey, to log in via the survey website to their Flickr account, and then to close their browser. This way, Flickr data about the respondents who logged in was automatically extracted using the Flickr API and recorded together with the respondents' responses to the questionnaire.

To measure social presence we used the number of contacts each user has. As for control variables, the number of photos a user has is extractable via the Flickr API and is therefore an independent, system generated measure. The number of years on Flickr, on the other hand, is not available via the API and therefore the respondents were asked to report it as part of the questionnaire.

Since we are interested in tagging behavior, only users who tag were approached. This information is available in the users' Flickr page. In addition, we avoided approaching users who tagged in languages other than English, to ensure that all respondents have a clear understanding of the survey items.

A randomly chosen sample of 1467 Flickr users who had at least one publicly viewable photo were emailed an invitation to participate in our web-based survey. A total of 251 valid responses were received. This represents a 17.1% response rate, much in line with similar studies.12 54.3% of the respondents were male, and the respondents' median age was 33.

Back to Top

Results

All the motivations measured exhibited high reliability (Cronbach's alpha was well above the .7 threshold). Potential multicollinearity was checked using Variance Inflation Factors (VIFs). The maximum VIF obtained was 1.84, well below the traditional cutoff of 10 for regression models, indicating the multicollinearity is not a concern.

Tagging for public Flickr users was ranked as the most important motivation among the respondents (see Table 2), and the Family & Friends motivation was ranked lowest, though only slightly lower than the Self motivation.

The mean number of distinct tags per respondent was 387 (SD = 633). The numbers, however, varied greatly across taggers. As expected, it was found that the levels of the Self and Public motivations, as well as the social presence factor (for example, the number of contacts), and the number of photos and years on Flickr, were positively associated with tagging level (see Table 3). However, contrary to our expectations, the Family & Friends motivation was not correlated significantly with tagging level. How can this lack of correlation between the Family & Friends motivation and tagging be explained? One possible explanation, drawing on Ames and Naaman's interviews,² might be that for the Family & Friends audience, the Communication function was more prominent than the Organization function (in other words, tagging is used to describe images to family and friends, not to help them find images). The Communication function on Flickr, however, can be facilitated by means other than tagging (such as, titles and captions attached to the photos, and sets that group photos together according to a theme chosen by the photo uploader). In addition, users can communicate details about the photos to their friends via other, external means (such as, email). Thus, overall, it seems plausible that the Family & Friends motivation would not be correlated with the level of tagging.

Back to Top

Conclusion

Given the growing popularity of tags as means of sharing and organizing large amounts of information,⁴ and since user participation is critical to the sustainability of content sharing communities,⁶ developers and managers of collaborative content sharing systems such as Flickr, del.icio.us, and YouTube may benefit from understanding what motivates users to tag. To understand why users tag, we looked at individual level motivations, using a newly developed scale, based on the work of Ames and Naaman,² as well as the social presence driver (the number of contacts), and the number of photos users have. We used a Web-based survey, drawing on both user-reported responses, and independent, system-generated performance data. The findings of the survey suggest that, as expected, both social presence and individual level motivations affect users' tagging level, with the exception of the Family & Friends motivation. The social presence drivers were found to be stronger predictors of users' tagging, which comes as no surprise given the collaborative, public nature of Web sites such as Flickr. The number of photos a user has is also a predictor of tagging level, as expected.

Assuming that the correlations found also involve causality, it is advised that managers of collaborative content systems seeking to increase tagging activity focus their communication and marketing efforts on those factors that have a strong impact on the level of tagging. For example, the motivation of tagging photos for public users who are not friends or family has a positive effect on tagging level. Therefore, it would make sense for organizers of collaborative content systems to focus their cultivation efforts in this area, by highlighting to such users the possibility of being exposed to other, unknown users.

In line with findings from previous research on social presence in other collaborative systems,⁷ social presence proved to have a positive effect on tagging in the present study. It would therefore make sense for developers and organizers of content-sharing systems to focus efforts in this area as well, by exposing users to the benefits of having contacts and encouraging users to add contacts. In addition, the social presence effect on tagging lends support to the recommendation for designing content sharing systems in such ways that they provide opportunities for social presence in order to boost tagging.

The present study was conducted on one system, Flickr, and therefore the findings may be influenced by Web site specific features. Studying other tagging systems, such as del.icio.us and YouTube.com, can verify the findings' generalizability. Further research may be helpful in understanding how different motivations influence contribution in different content sharing systems. This article, addressing one of the most prominent collaborative content sharing systems, is hopefully a useful step in this direction.

Back to Top

Back to Top

Back to Top

Back to Top

Back to top

The Digital Library is published by the Association for Computing Machinery. Copyright © 2010 ACM, Inc.

---

No entries found