Tags on healthcare information websites: A theatre of the absurd
PhD Thesis by Marit Kristine Ådland
Thesis submitted September 20, 2019
This thesis explores tags and tagging behaviour on health information websites using an empirical, user-oriented, exploratory case study. The purpose is to find out more about tags and tagging behaviour on a health information website.
Method: Data were collected in diverse ways, in order to obtain a view on the research questions from different angels. In a preliminary study, I analysed tags from Blogomkraeft.dk and compared them to the site structure of Cancer.dk. After the launch of the tagging feature on Cancer.dk, a study of user behaviour on Cancer.dk was conducted, with a focus on the role of tags. Participants solved tasks using the newly launched tagging feature, they filled out pre- and post-test questionnaires, and I interviewed them. When the tagging feature had been live for about a year, I also interviewed three editors about their experience and opinions on tags and tagging.
To study the tags themselves, a transaction log containing more than 25,000 tags was analysed, mainly analysed them through categorizing the tags into categories: (1) Internal and external tags, (2) Lay or professional tags, (3) Topical facets and (4) Aboutness.
Results: For taggers, the analysis indicate a connection between computer skills, an understanding of the tagging feature, and a focus on applying tags as topical descriptors.
Topical description was dominant when applying tags at Cancer.dk. Some of the taggers stated that they wanted tags to be exclusively topically descriptive. Participants who did not apply topical descriptive tags all agreed that such tags could be useful. A focus on subject description was often connected to a focus on finding information. To the participants, the topical tags did not have to describe the topic of the article; it was enough that that it described a subsection or an aspect of its topic. Thus, topical tags did not equal subject headings. Their requirements were not as strict.
Other purposes found were tags to explain the content, tags to evaluate articles, and tags to express requests for additional information. These tags represented attempts to communicate with the system, its users, or editors. All the participants agreed that topical tags were good, but they did not agree on whether other types of tags added value to Cancer.dk.
The different purposes users had when they applied tags was a challenge for the editors. From the interviews, it is my impression that the editors in a way did not want tags, but a controlled vocabulary. This would fulfil some of the purposes that both editors and users had when they applied tags. A subject language that includes synonyms and possibly includes relations between terms (e.g. hierarchical) would give lead-in terms that users and editors need. Such a solution is, however, contradicted by the editors’ view that tags are mainly the user’s voice in the system. A controlled vocabulary can never replace this, which the editors were also clear about.
Internal and external taggers behaved differently. The internal taggers were from inside the organization behind Cancer.dk. It was easy to address them as a group and encourage them to apply tags. However, the crowd of external taggers were more stable.
Analysis of the log files showed how difficult it is to apply tags. The aboutness categorization reveals challenges in how tags relate to the topical content of the article. Mixed together, the tags as a whole were difficult to use and difficult to judge.
Tags from internal taggers covered categories that were more diverse and described the article content from various angles. Their tags were more evenly distributed on tag facets compared to external taggers. These results conflicted with the expectation that external users can add new viewpoints to the systems. External taggers, however, applied more tags not related to the content of articles.
The results can inform the design of tagging features; visibility is essential to attract tags, and also influence the characteristics of tags slightly. The information surrounding tagging needs testing. The disagreement between the user groups can also inform tagging features: the tags applied within the system will be influenced by who has permission to apply tags.
The communicative aspects of tags found in Cancer.dk indicate that taggers do not necessarily distinguish between tags in different systems. When looking at systems like Twitter, tags are communicative by intent; they add information to the tweet and do not necessarily cover the topical content of the tweet. However, when moved to an information website, this behaviour is unwelcome.