Shortchanged: Uncovering and Analyzing Intimate Partner Financial Abuse in Consumer Complaints

TL;DR

This paper tackles IPFA within the realm of digital financial services by mining the CFPB complaint database to understand how intimate partners abuse financial systems and what survivors experience when seeking redress. It develops a novel workflow that fuses pre-trained language models, proximity keyword matching, clustering, and expert review to extract a focused IPFA corpus, yielding 464 narratives and 513 IPFA-flagged complaints from a broader set of nearly 1 million narratives with content. Using Framework Analysis and Critical Discourse Analysis, the study profiles complainants and abusers, catalogs 14 attack types across 24 products, and identifies barriers in policy, evidence gathering, and reporting that hinder recourse, while proposing safety checkups and automated evidence approaches to improve protection and empowerment. The findings offer practical guidance for financial institutions and designers to strengthen digital-safety features, evidence-gathering mechanisms, and staff training, aiming to better support survivors and reduce risk in technology-enabled financial abuse.

Abstract

Digital financial services can introduce new digital-safety risks for users, particularly survivors of intimate partner financial abuse (IPFA). To offer improved support for such users, a comprehensive understanding of their support needs and the barriers they face to redress by financial institutions is essential. Drawing from a dataset of 2.7 million customer complaints, we implement a bespoke workflow that utilizes language-modeling techniques and expert human review to identify complaints describing IPFA. Our mixed-method analysis provides insight into the most common digital financial products involved in these attacks, and the barriers consumers report encountering when doing so. Our contributions are twofold; we offer the first human-labeled dataset for this overlooked harm and provide practical implications for technical practice, research, and design for better supporting and protecting survivors of IPFA.

Figures (9)

• Figure 1: The IPFA complaint collection workflow involves a reference set (ref) of known IPFA examples and unlabelled CFPB complaints containing an intimate partner keyword. Text embeddings are created using a sentence transformer model, and these embeddings are then clustered. The clusters with the most reference set complaints are manually reviewed to identify additional IPFA examples, which are then added to the reference set. This process is iterated with new IP complaints. When the desired iterations are complete, ref becomes the final collected dataset.
• Figure 2: SHAP scores for K-Clusters 1 [Left], and 6 [Right]. The color bar corresponds to the raw values of the variables for each instance. If the variable for a particular word is high, it appears as a red dot, while low variable values appear as blue. \ref{['fig:SHAPs-scores']} in \ref{['sec:app-cfpb-workflow']} shows SHAP scores for all clusters.
• Figure 3: The graphs from our meta-data analysis: the complaint length, distribution over years, and identification of higher ref cases in clusters.
• Figure 4: Example, paraphrased complaint (C5) with indicators for the framework creation. Our analytical framework is built from identifying the , , , , , , and on the complaint.
• Figure 5: Top 20 products to financial attack mapping. Complainant (C), Intimate Partner (IP), [Unspecified] means unspecified product, [Generic] means a collective of consumer banking products.