There is growing evidence that much of the material on the internet is entirely fake. This is perhaps well-known. Indeed, there is also equally compelling evidence that a huge number of the people on the internet are fake too. Much of the engagement and virality of content on social media and elsewhere being nothing but automated bot activity and click farms. Much of it is done as part of promoting misinformation for a political agenda and a lot of it is done to scam advertisers into imagining their paid ads are being seen by real people.
However, in the perhaps more mundane world of actual users, searching for information about products and services in which they are interested there is a need to be able detect fake reviews. A review in the International Journal of Intelligent Engineering Informatics, has taken a look at the approaches to detecting deceptive reviews. One recent analysis suggests that two-thirds of customer evalutions, or reviews, of products sold on a major e-commerce site are fake. These fake reviews not only distort the average opinion for a given product, often boosting a low number of “stars” for a shoddy product to make it a more saleable five-star item, but also boost the seller’s overall profile illicitly too.
Rajdavinder Singh Boparai and Rekha Bhatia of the Department of Computer Science and Engineering at Punjabi University in Patiala, Punjab, India, discuss the state-of-the-art in research into this problem. They also survey the various AI, artificial intelligence, machine-learning tools aimed at flagging non-genuine reviews on commercial websites. The team reveals the gaps in the research literature as well as the limitations of the current tools and points to how those gaps might be plugged.
Deceptive reviews can have myriad sources and authors making it difficult to home in on a particular writing style as fake. A significant gap in current research and tools that might be filled by future research would see the development of a more representative model that would be generic, capable, and portable and be able to quickly and accurately flag as fake deceptive reviews based on real-world data. Given the recent public advent of so-called language models and tools, it is likely that we will see more and more fake reviews online. However, the very tools that generate such deceptive content might also be used to detect its presence. We will inevitably see a game of cat and mouse between the e-commerce sites and the fakers and caught in the middle will be the consumers looking for a decent product at a good price.
Boparai, R.S. and Bhatia, R. (2022) ‘Deceptive web-review detection strategies: a survey’, Int. J. Intelligent Engineering Informatics, Vol. 10, No. 5, pp.411–433.