Start Date
12-13-2015
Description
The growing interest in social media for legitimate promotion has been accompanied by an increasing number of fraudulent reviews. Beyond fraud detection, little is known about what review portals should do with fraudulent reviews after detecting them. In this paper, we study how consumers respond to potentially fraudulent reviews and how review portals can leverage such knowledge to design better fraud management policies. To do so, we combine randomized experiments with statistical learning using large-scale archival data from Yelp. Our experiments show that consumers tend to expand the variety of their choice set during product search and to increase their trust towards the review portal when it displays fraudulent reviews along with non-fraudulent reviews, rather than censor fraudulent information. Finally, our archival analysis using a Maximum Likelihood Estimation method allows us to design a novel fraud-awareness reputation system that platforms can deploy to better improve consumer trust and decision making.
Recommended Citation
Ananthakrishnan, Uttara; Li, Beibei; and Smith, Michael, "A Tangled Web: Evaluating the Impact of Displaying Fraudulent Reviews" (2015). ICIS 2015 Proceedings. 1.
https://aisel.aisnet.org/icis2015/proceedings/HCI/1
A Tangled Web: Evaluating the Impact of Displaying Fraudulent Reviews
The growing interest in social media for legitimate promotion has been accompanied by an increasing number of fraudulent reviews. Beyond fraud detection, little is known about what review portals should do with fraudulent reviews after detecting them. In this paper, we study how consumers respond to potentially fraudulent reviews and how review portals can leverage such knowledge to design better fraud management policies. To do so, we combine randomized experiments with statistical learning using large-scale archival data from Yelp. Our experiments show that consumers tend to expand the variety of their choice set during product search and to increase their trust towards the review portal when it displays fraudulent reviews along with non-fraudulent reviews, rather than censor fraudulent information. Finally, our archival analysis using a Maximum Likelihood Estimation method allows us to design a novel fraud-awareness reputation system that platforms can deploy to better improve consumer trust and decision making.