Feng Chia University (TAIWAN)
About this paper:
Appears in: EDULEARN16 Proceedings
Publication year: 2016
Pages: 5113-5114 (abstract only)
ISBN: 978-84-608-8860-4
ISSN: 2340-1117
doi: 10.21125/edulearn.2016.2208
Conference name: 8th International Conference on Education and New Learning Technologies
Dates: 4-6 July, 2016
Location: Barcelona, Spain
Tagging, part of the collaborative nature of Web 2.0, has become popular in recent years. According to scaffold theory, tags generated by domain experts can function as an overarching scaffold providing a conceptual level of abstraction for “hard content” (e.g., technical papers about ergonomics) and novices can use these tags to guide them through the learning process in web environment.

To facilitate nonexperts’ self-learning of hard content, it is crucial to ensure that the tags provided to them are of good quality, or, similar to those tags generated by experts only. Since the tagging process goes through an often-anonymous environment where experts and nonexperts coexist, it is imperative to understand the behaviors of novices under social influence with or without the presence of experts.

Three research questions:
RQ1: Under social influence, how are the tagging behaviors of novices affected by the presence of expert tags, in contrast to those influenced by novice tags only?
RQ2: Under social influence where tags generated by experts and nonexpert coexist, what types of tags generated by experts are more tend to be adopted by nonexperts?
RQ3: Will the size of nonexperts dilute the expert influence and how serious is the effect?

Nine scientific articles regarding human factors and ergonomics, written in Chinese, were collected from the Internet and were modified for the purpose of this experiment. Each article contained an average of 710 Chinese characters and these articles all involved some design concepts related to ergonomics.

The datasets used for analysis in this study include Homogenous Dataset (HD) where tags generated by 15 nonexperts or 15 experts separately, and Influenced Nonexpert Dataset (IND) where tags generated by nonexperts under social influence at the presence of experts’ tags. There two expert-nonexpert ratios when collecting IND, where the expert-nonexpert ratios is 1:1, and where the expert-nonexpert ratios is 1:2. Delphi method was adapted to generate and collect HD and IND.

By observing the similarity among tags generated by nonexperts only, by experts only, and by nonexperts influenced by experts (influenced nonexperts), the similarity analysis (Jaccard similarity, Overlap lst) showed that by the end of Delphi process, tags generated by influenced nonexperts were more similar to those generated by experts, when compared with those generated by nonexperts only.

By categorizing tags into 4 groups according to terminology or non-terminology; explicit or implicit, we can further understand how nonexperts imitate the tagging behaviors of experts. Experiments results suggested that, tags originally generated by only experts categorized as implicit terminology had highest adoption rate (0.862) by influenced nonexperts, compared with explicit terminology (0.222); implicit non-terminology (0.044); and explicit non-terminology (0.282).

When it comes to social influence, it can be reasonably assumed that the ratio between number of experts and number of nonexperts may affect the adoption of expert tags by influenced nonexperts. If there exists more nonexperts in a tagging environment, the influence of expert may be reduced. Experiment results suggest that either expert-nonexpert ratios being 1:1 or 1:2, the quality of tags generated by influenced nonexperts were similar. This finding has an implication in implementing a tag recommender system for facilitating learning hard contents for nonexperts.
Tagging, web 2.0, social influence, experts and nonexperts, similarity.