What is it about?

Most of us use e-commerce sites like Amazon or eBay, where algorithms suggest products we might like based on our past behavior. These "recommender systems" are designed to make shopping easier, but they often come with hidden problems that can harm shoppers and businesses. This research paper identifies these "hazards" and explores ways to address them to ensure online shopping is better for society. The paper categorizes the main problems into four groups: Bias: Sometimes systems only suggest popular items, leaving out great new products or favoring items that make the company more money rather than what the customer actually needs. Malicious Activities: This includes activities such as fake reviews or "shilling attacks," where people create fake accounts to trick the system into recommending their products. Privacy Risks: Systems often collect vast amounts of personal data, which can lead to privacy leaks or even price discrimination. System Inefficiency: Sometimes the recommender isn't smart enough to understand what a customer really wants, leading to information overload and frustration. While many researchers look into these problems one by one, this paper is unique in bringing all the solutions together in one place. It reviews three main ways to fight these hazards: creating better, fairer technology; helping customers understand how these systems work; and enacting stronger laws and regulations to hold companies accountable. By looking at the problem from all these angles—technical, human, and legal—the paper provides a complete roadmap for making online shopping safer and more ethical for everyone.

Featured Image

Why is it important?

This research is timely because AI-based recommendation technologies now influence almost every purchase we make online. When these systems are biased or insecure, they don't just affect our wallets; they erode our trust in technology and can even lead to unfair treatment of different groups of people. What makes this paper stand out is its "hazard-based" approach. Instead of just talking about abstract risks, it identifies specific events—like a fake review or a biased algorithm—and shows how they can be measured and stopped. By bridging the gap between computer science, law, and consumer behavior, this work helps ensure that the future of e-commerce isn't just about maximizing profits, but about protecting the well-being of the people who use it.

Perspectives

From my perspective, the core of this research is about "Social Good". For too long, the development of recommendation technology has focused almost entirely on accuracy and sales. I believe it is time to shift our focus toward "Safety by Design". We shouldn't just wait for something to go wrong; we should build systems that are inherently fair, transparent, and protective of user privacy from the very start. My goal with this paper is to empower both the people who build these systems and the people who use them. By making the "black box" of AI-based recommendations more transparent, we can move toward a digital marketplace where everyone has a fair shot, and where technology serves the interests of society as a whole, not just the biggest corporations.

Eranjana Kathriarachchi
Massey University

Read the Original

This page is a summary of: Overcoming Hazards of E-commerce Recommender Systems for Social Good, ACM Transactions on Recommender Systems, December 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3785355.
You can read the full text:

Read

Resources

Contributors

The following have contributed to this page