Explaining reviewer trustworthiness from reviewer characteristics: a regression model

Online consumer reviews have an essential role in online consumer’s buying decisions and e-commerce sales as consumers trust them more than the information provided by the seller. The reviews supposedly give indications to previous consumer’s buying experiences and satisfaction, giving potential buyers more perspective on the considered product or service. However, the anonymity, the easiness of leaving a review, the increasing amount of reviews and review websites available, and the benefits of good reviews to a seller have opened the online reviews for manipulation and fraud. This fragile trust and uncertainty have hurt the potential buyers, the sellers’ credibility and sales, the review platform’s credibility, and the reviewer’s credibility. One of the ways to address this issue is finding trustworthy reviewers as they write reliable reviews, helping potential consumers reach a credible information source. A previous study in 2017 has proposed a novel model of explaining reviewer trustworthiness from reviewer characteristics. As an attempt to explore the latest research and the novel model explaining reviewer trustworthiness, this thesis aims to verify these reviewer characteristics’ contribution in solving the reliability of a reviewer on a similar dataset. Also, further analysis is conducted on the model and its components to obtain a well-performing and robust model to explain reviewer trustworthiness using reviewer characteristics, including involvement, positivity, experience, competence, reputation, and sociability. All reviewer attributes in the analysis are confirmed to be positively related to reviewer trustworthiness, which is computed by using a more robust and generalizable method than in the previous study. Productivity and recent activeness of a reviewer have been better representatives for the experience than the number of years on an online platform. Sociability, out of the attributes in scope, has been identified as having the least effect on reviewer trustworthiness, implicating that sociability should not be the top or only factor to judge the reliability of a reviewer. A modified inverse Gaussian model has been identified and tested in this paper to perform well and robust in explaining reviewer trustworthiness from reviewer characteristics in a straightforward and computationally efficient manner