Over the past decade, predatory and fraudulent practices in digital finance and financial technology have increased globally. The use of mobile applications for such purposes is a concern given the increased ubiquity of mobile devices, ongoing concerns about low financial and digital literacy in many population groups, and anecdotal evidence on a proliferation of methods and tools some finance app providers use to exploit vulnerable households and businesses. In addition to the direct harm caused to consumers, this can lead to mistrust of digital finance, which can delay financial inclusion efforts and undermine the benefits of financial technologies.
The objective of this project is to explore whether high-frequency app data and applied machine learning techniques can be leveraged to create a system for flagging and reporting highly suspect apps. As a “proof of concept,” the researchers draw on historical app meta and review data for 63 countries covering from January 2020 to April 2021 to document the prevalence of such problematic apps and test the efficiency and accuracy of such methods. To keep the project tractable, the researchers focus on a targeted subset of personal loan apps. The pilot informs future experimentation with the approach and suggests possible real-world applications, which could help provide targeted shortlists of highly suspect individual apps, country- or global-level monitoring to understand prevalence of problematic apps at a given point in time, or be fed into buyer-beware labelling on the app stores.