star-bg.jpg

CognitiveScale Blog

Consumer Lending and Credit: Tackle Business Risks from AI

By Max Kanaskar Feb 6, 2020 1:08:38 PM

0204_Max_Title (1)

Executive Summary

The key regulatory issues in consumer lending and credit – fair lending, consumer notification and reporting, and unfair, discriminatory or abusive acts or practices – have implications for various AI business risks.  Institutions should get ahead of the issue and tackle AI business risks head-on.

The Potential and Risks of AI in Consumer Lending and Credit

Big data and AI have tremendous potential in democratizing credit access per numerous studies and research by financial institutions, regulators and academics. There is however a tremendous risk potential associated with using big data and AI in consumer finance; for example, the strong opportunity for bias and discrimination through use of alternative data.

Key Regulatory Issues in Consumer Lending and Credit

Consumer finance has a web of regulations enforced by various federal agencies, chief among them the consumer financial protection bureau (CFPB), Housing and Urban Development (HUD), FDIC and Federal Trade Commission (FTC).  A quick summary of key federal laws and corresponding enforcing regulations:

  • Equal Credit Opportunity Act or ECOA / Regulation B governing discrimination and fair lending for consumers and businesses
  • Fair Housing Act / HUD’s 24 CFR Part 100 for residential real estate transactions across the entire lifecycle 
  • Federal Credit Reporting Act (FCRA) / Regulation V (12 CFR part 1022) for credit information reporting to consumers
  • Unfair, Deceptive, or Abusive Acts or Practices (UDAAP) – Section 1036 of Consumer Financial Protection Act and Section 5 of FTC Act enforced by CFPB and FTC 
  • TILA-RESPA Integrated Disclosures (TRID) / Regulation Z governing disclosures in consumer lending and real estate transactions
  • Other related areas: data privacy (Gramm-Leach-Bliley Act), community reinvestment act (CRA) etc.

The significant regulatory issues in consumer lending and credit are as follows: fair lending, consumer notification and reporting, and unfair or discriminatory acts or practices.

  • Fair Lending: Ensuring that protected classes are not discriminated against for credit and lending purposes across two areas
    • Disparate treatment: Overt or inconsistent practices that actively discriminate against a protected class e.g. redlining 
    • Disparate impact: Unintended exclusion or burdens on persons on a prohibited basis
  • Consumer Notification & Reporting: Ensuring appropriate functioning of consumer reporting agencies and full transparency to the consumer 
    • Governing the use of consumer information by consumer reporting agencies such as credit scores 
    • Empowering consumers to request information used in credit decisioning process
    • Consumer notification and credit decision outcomes that need to be explainable to consumers 
  • Unfair, Deceptive, or Abusive Acts or Practices: Ensuring that consumers are protected from deceptive, unfair and abusive practices 
    • Deceptive acts: Representations or omissions with material impact that are likely to mislead consumers 
    • Unfair practices: Likely to cause substantial injury to consumers and that consumers cannot reasonably avoid
    • Abusive practices: Taking advantage of a consumer’s lack of understanding or knowledge

Business Risks from Big Data and AI in Consumer Lending and Credit

Each of the major regulatory issues in consumer lending and credit has implications for AI business risks (see CognitiveScale AI Business Risk framework for more detail on AI business risks).  



  Data Risks

  Fairness and Bias

  Explainability

Fair Lending

  • Lack of nexus to creditworthiness 
  • Proxies for prohibited classes
  • Data privacy and protection concerns
  • Prohibited bias in marketing e.g., “steering”
  • Prohibited bias in credit underwriting e.g., “redlining”
  • Disparate impact of model ecosystem
  • Adverse action notice explanations

Consumer Notification & Reporting

  • Categorization of alt data as “consumer report”
  • Lack of “permissible purpose”
  • Presence of discriminatory factors or proxies
  • Discriminatory or incomplete disclosures of AI in underwriting
  • Adverse credit score explanations 
  • Risk-based pricing notice explanations 

Unfair, Deceptive, or Abusive Acts or Practices

  • Accuracy, reliability and representativeness of data 
  • Discriminatory consumer treatment based on factors lacking nexus and/or consumer control
  • Lack of appropriate and full disclosures 

 

Integrating AI Business Risk Management in Consumer Lending and Credit

Financial institutions have the opportunity to shape the regulatory agenda for AI in consumer lending and credit: witness CFPB’s collaborative regulatory approach with its “no action letters”, regulatory sandboxes and trial disclosures.  Furthermore, incumbents face rising competition from non-bank players (underwrite.ai and Upstart are two examples). Lastly, tools for managing AI business risks in consumer credit are already present in the market. CognitiveScale’s Certifai is a first of a kind AI auditor that enables model and data introspection in a non-intrusive manner.  It enables compliance teams to gain insight into factors that may raise compliance issues, and set thresholds for ongoing monitoring.  Business risk management should be an active consideration with AI in consumer credit, not an afterthought.

 

CTA-Certifai-1-1