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

Title of the Paper

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

Paper Information

• Research Domains: FinTech, Social Security, Intimate Partner Violence
• Keywords: Financial abuse, Intimate partner violence, Technology-driven abuse, Consumer complaints, Digital financial products, Natural language processing

Research Background and Issues

• Problems or Challenges:

  • Financial abuse is a form of intimate partner violence (IPV) that significantly impacts victims' psychological, physical, and economic well-being.
  • Digital financial products may exacerbate these issues by enabling perpetrators to monitor, control, and exploit victims' financial information.
  • Current financial institutions face policy and technical deficiencies in addressing consumer complaints related to technology-driven abuse.
  • The complexity of digital financial abuse is constrained by the availability of research data and users' self-disclosure behaviors.

• Significance:

  • Intimate Partner Financial Abuse (IPFA) affects victims' quality of life and safety, necessitating action from financial institutions to better protect customer rights.
  • The research contributes to improving the design of digital financial products, reducing security risks stemming from technological or policy gaps.

• Research Motivation and Related Work:

  • The authors aim to explore how publicly available complaint data can help identify patterns of intimate partner financial abuse through natural language processing and manual review.
  • Compared to previous studies, this work focuses on firsthand textual data from consumer complaints, proposing new methods to describe technology-driven abuse and related barriers in digital products.

Proposed Solution

• Methodology or Solution:

  • Developed a five-step workflow that extracts IPFA-related complaints through keyword extraction and matching, text embedding, clustering, and expert manual review.
  • Utilized pre-trained language models (e.g., BERT, SimCSE) to generate semantic text embeddings and applied K-means clustering for complaint classification.

• Innovations:

  • Presented the first manually annotated dataset on intimate partner financial abuse, highlighting underexplored technology-driven abuse in consumer complaints.
  • Combined natural language processing with framework analysis and critical discourse analysis to uncover barriers faced by victims in digital financial products.

• Implementation Steps:

  1. Constructed a keyword list covering intimate partner-related expressions and financial abuse-related terms.
  2. Performed keyword matching (based on semantic similarity and contextual distance) to extract an initial set of relevant complaints.
  3. Generated text embeddings using language models and classified complaints using clustering algorithms.
  4. Manually reviewed clustering results to extract IPFA-related cases and expand the annotated dataset.
  5. Created a high-quality dataset containing 513 cases of intimate partner financial abuse.

Research Outcomes

• Specific Results:

  • Identified 14 distinct forms of technology-driven financial attacks within digital financial products, along with 24 related technological products and services.
  • Highlighted issues such as account takeovers for criminal activities, negligence in asset management, and deceptive behaviors in interactions between customers and financial institutions.

• Advantages and Comparisons:

  • Compared to previous studies, this work provides an in-depth analysis of specific digital financial products and their abuse patterns.
  • Offers a detailed perspective on consumer complaint data, showcasing cases of abuse and barriers to resolution within digital financial products.

• Experiments or Evaluation Results:

  • Extracted 464 complaint texts from the CFPB database, analyzing their linguistic patterns and social dynamics.
  • Found that most consumers did not label their experiences as "abuse," highlighting deficiencies in FinTech design and service policies.
  • Expanded the dataset by 43% through manual and algorithmic methods, covering various abuse types and high-risk behaviors.

• Limitations and Future Directions:

  • The dataset is relatively small, and self-reported consumer complaints may not encompass all abuse scenarios.
  • Future research could incorporate multi-agency data to study IPFA patterns across different time dimensions.
  • Explore improvements in financial product design and support mechanisms, such as user feedback-based protective tools and systematic evidence collection technologies.

Conclusion

This study systematically identifies the technological patterns and social dynamics of intimate partner financial abuse by combining natural language processing techniques with in-depth analysis of consumer complaint data. The research highlights significant barriers victims face when seeking solutions and proposes new methods to enhance the security and functionality design of digital financial products.