Using AI in Small-loan Mediation: A Field Experiment

How to cite

Huang, Zhitao; Liu, Zhuang; Tang, Yingmao. 2024. "Using AI in Small-loan Mediation: A Field Experiment." AEA RCT Registry. AEARCTR-0014962. https://doi.org/10.1257/rct.14962-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Zhitao Huang* , Zhuang Liu , Yingmao Tang

* Primary investigator

Registry ID
AEARCTR-0014962
DOI
https://doi.org/10.1257/rct.14962-1.0
First registered
Last updated
Status
In development
Trial dates
2024-12-15 – 2025-05-31
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/14962

Abstract

Settlement through mediation is a crucial mechanism for dispute resolution. However, the effectiveness of mediation can vary significantly based on case characteristics. This study develops an artificial intelligence (AI) model that predicts mediation success probability in small-loan disputes. We plan to conduct a field experiment in collaboration with a mediation center in China, investigating whether and how the use of AI predictions influences mediation behavior and mediation results. AI predictions of mediation success probability will be introduced in the treatment group.
This research contributes to the growing literature on AI’s applications in the judicial settings and provides empirical evidence on how predictive analytics can affect dispute resolution.

Intervention

The treatment consists of providing mediators with AI-generated predictions of mediation success probability. Before each mediation session, cases in the treatment group will include a report showing the AI-predicted likelihood of reaching a settlement.

Experimental design

Random introduction of AI prediction to half of the mediation cases. These cases will involve loan defaults with amounts typically below $10,000, representing common small-loan disputes.

Design details

Randomization method
Fully Randomized at the case level by a computer.
Randomization unit
Case level
Clusters
5000 cases
Observations
5000 cases
Arms
Treatment arm: 2500 cases Control arm: 2500 cases
Primary outcomes
Whether the use of AI prediction influences: 1. Average time spent on mediation 2. Number of mediation sessions per case 3. Settlement rate 4. Collection of debt

Cited by

No citation data available for this registration yet.

Keywords