Crowd-sourced Manipulation and Fraud Detection

Abstract

Prediction markets are a common tool of companies for idea management and evaluation during the innovation process, which enables them to include expectations and opinions of stakeholders across organizational boundaries. However, prediction markets are also known for their susceptibility to manipulation in theory and practice. The irregular and multifaceted occurrence of these phenomena, with sometimes very creative strategies, makes it difficult to detect manipulation and fraud based on algorithms. To ensure robust and reliable forecasts, which are of utmost importance for a focused and successful digital innovation process, there is a need for a monitoring approach capable of dealing with these specific problems. In an Action Design Research project, we address this problem by developing a crowd-sourced manipulation and fraud detection tool. The artifact enables the crowd to successfully decompose the large set of trading data and successfully find even creative strategies without guidance. The artifact is implemented and evaluated in the field in the prediction market [blinded for review]. We conclude, that a crowd-sourced approach can be suggested to monitor ambiguous and rare events with a varying character in our context and presumably other contexts as well.