Law Review (Economics): Choi, James J. and Huang, Dong and Yang, Zhishu and Zhang, Qi, How Good is Ai at Twisting Arms? Experiments in Debt Collection (April 2025). NBER Working Paper No. w33669
Law Review (Economics): Choi, James J. and Huang, Dong and Yang, Zhishu and Zhang, Qi, How Good is Ai at Twisting Arms? Experiments in Debt Collection (April 2025). NBER Working Paper No. w33669 Ed Boltz Tue, 10/14/2025 - 17:41 Available at: https://ssrn.com/abstract=5215980 Abstract: How good is AI at persuading humans to perform costly actions? We study calls made to get delinquent consumer borrowers to repay. Regression discontinuity and a randomized experiment reveal that AI is substantially less effective than human callers. Replacing AI with humans six days into delinquency closes much of the gap. But borrowers initially contacted by AI have repaid 1% less of the initial late payment one year later and are more likely to miss subsequent payments than borrowers who were always called by humans. AI’s lesser ability to extract promises that feel binding may contribute to the performance gap. Commentary: A recent study published by the National Bureau of Economic Research examining Chinese consumers’ reactions to debt collection, including the use of AI-driven “collectors,” offers interesting insights—but while its conclusions rest heavily on psychological and behavioral research conducted by U.S. scholars on fairness, compliance, and emotional response in debt collection and originated in the American context leaves substantial questions about how transitive those findings really are between the sharply different regulatory and social environment in China and the United States (let alone elsewhere.) In China (with a full admission that I'm only a North Carolina Bankruptcy Expert!), consumer bankruptcy remains rare, with only limited pilot programs in a handful of provinces and far fewer formal protections for over-indebted individuals. Without leaning too hard on cultural generalizations, collection practices therefore tend to rely on social pressure and moral appeals, often leveraging family networks and reputational risk, rather than the structured statutory regimes familiar to U.S. practitioners. Against this backdrop, the finding that Chinese consumers respond more favorably to “empathetic” or “respectful” AI collectors may reflect local cultural expectations about deference and face-saving, rather than a universal truth about human-machine interaction. By contrast, in the United States, debt collection operates within a robust legal framework of consumer protection—anchored by the Fair Debt Collection Practices Act (FDCPA), the Fair Credit Reporting Act (FCRA), and state Unfair and Deceptive Trade Practices Acts (UDTPA) such as North Carolina’s § 75-1.1. While social stigma and moral hazard still play a substantial role, these laws don’t merely regulate behavior; they define the very boundaries of communication. A U.S. consumer must be told who is collecting the debt, how much is owed, and how to dispute it—and harassment or misrepresentation is strictly prohibited. Moreover, Chapter 7 and Chapter 13 bankruptcy provide predictable, court-supervised debt-relief channels unavailable to most Chinese consumers, fundamentally altering both the power dynamics and the perceived legitimacy of collection efforts. Given that backdrop, an AI debt collector operating in the United States would almost certainly be required not only to disclose that it is a debt collector, but also that it is an artificial intelligence system. Failure to do so could violate the FDCPA’s prohibitions on “false, deceptive, or misleading representations,” particularly if the AI’s design made a consumer believe they were conversing with a human being. The CFPB’s 2024 digital-communication guidance and the FTC’s emerging policies on AI transparency both suggest that accuracy of identity and medium are integral to consumer protection. (The viability of these policies, however, may be in doubt under the current U.S. administration.) The deeper lesson, then, is that technology doesn’t operate in a vacuum. It reflects the social and legal system that deploys it. Where Chinese law emphasizes harmony and moral rehabilitation over statutory procedure, AI may simply mechanize social pressure. In the United States, by contrast, our debt-collection system—grounded in disclosure, due process, and the constitutional promise of a “fresh start” through bankruptcy—would demand that even machines play by the same rules of fairness and honesty that govern human collectors. With proper attribution, please share this post. To read a copy of the transcript, please see: Blog comments Attachment Document ai-debt-collection-20250331.pdf (3.47 MB) Category Law Reviews & Studies
