Generating trust in collaborative environments
Evaluating design parameters in area of semantic annotations
Trust is the key feature for human interaction, including the consumption of information. Due to the increasing distance communication in a digital world and the multiplicity of mostly unknown sources of information on the web, it has become di cult to identify trusted information. Increasing numbers of users consider online communities to be a source of information that can be created by almost any other user or more than one. However, one main challenge in such online communities is how to verify the credibility of this information. Furthermore, the various platforms and the complexity of the subject (trust) make the development of mechanisms to identify trustworthy information more challenging. More precisely, this raises the research question of what are the socio-technical design parameters for building trust in collaborative annotation environments? To this end, this dissertation has examined the collaborative environments Genius and Stackoverflow in light of their real data. The goal is to understand user behavior in order to identify the information characteristics that make such information trustworthy through interaction. This work proposes a trust model that comprises the dimensions stability, credibility, and quality. It calculates a trust degree of short-text based on its characteristics and classifies it into a trust class (very-trusted, trusted, untrusted and very-untrusted). The information characteristics were considered from two perspectives: Metadata and content. The evaluation of the metadata is based on user preferences within a survey, while the content is verified for its text-embedded features using data mining techniques. The proposed trust model supports the identification of trusted information in collaborative environments. It can be used in various online communities that deliver the appropriate metadata of the information provided. The trust model helps to filter the information and thus reduces the information-overload shared on the web. Applications can integrate the trust model into their development in order to increase the likelihood of their use, as users are able to recognize trusted information easily. In contrast to existing works, this thesis proposes a trust model that combines the metadata and short text characteristics to produce a human-readable interpretation of the calculated trust degree.
Berlin, Freie Univ., Diss., 2019