How Game Theory and AI Stop Fake Reviews on E‑Commerce Platforms

On e‑commerce, group‑buying and review platforms, merchant rankings often depend on user ratings, giving high‑scoring merchants more traffic and incentivizing quality.

However, some dishonest merchants resort to fake transactions (刷单) to boost rankings, harming user experience and fair competition. Alibaba adopts a zero‑tolerance stance, continuously strengthening detection and penalties.

Alibaba’s scientists leverage big data, machine learning, deep learning, large‑scale graph search, and real‑time computing to combat fraud, building a multi‑dimensional monitoring, detection, and enforcement system after years of data accumulation on the world’s largest e‑commerce platform.

Recently, Alibaba’s search team collaborated with Prof. Tang Pingzhong’s group at Tsinghua University’s Institute of Cross‑Media Information to apply game theory and mechanism design to anti‑fraud, publishing their findings at the top recommendation conference RecSys 2016.

Mechanism design, a fast‑growing branch of microeconomics, integrates game theory and social choice to construct games that guide participants toward desired outcomes.

By modeling the cost of fake orders and the traffic gain, the researchers designed a new recommendation mechanism where each seller’s cost from cheating exceeds any benefit, eliminating incentive to cheat—a result validated in simulation experiments.

Read the original paper for details.