Courts of Tomorrow

Last registered on November 01, 2024

Pre-Trial

Trial Information

General Information

Title
Courts of Tomorrow
RCT ID
AEARCTR-0012906
Initial registration date
January 28, 2024

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
January 31, 2024, 12:11 PM EST

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Last updated
November 01, 2024, 6:19 AM EDT

Last updated is the most recent time when changes to the trial's registration were published.

Locations

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Primary Investigator

Affiliation
New Economic School Moscow

Other Primary Investigator(s)

PI Affiliation
ETH Zurich

Additional Trial Information

Status
In development
Start date
2024-01-01
End date
2025-12-31
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
We integrate a specialized AI course into the Pakistan Judicial Academy's "Technology and Law” initiative. We will design and build an AI-based judge support tool that empowers judges to search, cite, and summarize the history of Pakistan precedents as well as newly submitted briefs and other documents. The system, especially with the associated training, is protected against plagiarism, hallucination, and providing false citations. The tool and training will be provided in the context of a randomized field experiment, equipping about one-third of Pakistan's trial court judges with generative AI technology and associated training and support. Our research will evaluate the effect of AI technology and training on judge performance, including AI usage, perceived AI usefulness, and quality measures constructed from written rulings. Our study's findings have the possibility to shed light on the potential of generative AI to bolster state capabilities and judicial productivity worldwide.

*Update to Trial - Rerandomization in October 2024
The random assignment of judges was conducted in two distinct waves for registration of judges into JudgeGPT subscriptions.

In February 2024, the first wave of registration saw 979 judges from Pakistan's lower courts sign up to participate in our experiment. We randomly assigned these 979 judges into two groups: 487 judges were allocated to the treatment group (Batch 1) and provided with access to the JudgeGPT subscription and GPT instruction course, while the remaining 492 judges were designated as the control group (Batch 2), scheduled to receive the same access in September 2024. This setup allows for a randomized control trial comparing the outcomes of Batch 1 and Batch 2. Following the initial random assignment, the introduction of the password-protected JudgeGPT, designed specifically to prevent spillovers, sparked considerable interest among judges who had not initially registered for the course but were nonetheless eager to participate but could not access GPT or the course.

An additional 580 judges expressed interest in the course and the JudgeGPT tool. To preserve the study's integrity, we decided against adding these new applicants to our control group (Batch 2), as they were not randomly assigned. Therefore, a second randomization was conducted to maintain the integrity of the study and increase its statistical power, accommodating a total of 1559 judges instead of the initially registered 979. This means more than 50% of the trial court judges (court of first instance) in Pakistan registered to participate in our experiment.

In October, the second wave of randomization, therefore, took place on October 23, 2024, for 580 judges. The 580 judges were randomly assigned to Batch 3 (n = 218), which will take the course in December 2024 and January 2025, and Batch 4 (n = 362). Batch 3 judges would get the same treatment as Batch 1 and 2: JudgeGPT course and JudgeGPT subscription. Batch 4, however, is further randomized into two subgroups: Batch 4a and Batch 4b. Batch 4a is randomly assigned to receive JudgeGPT training and a placebo course on Technology and Law in December 2024 and January 2025, along with a GPT subscription (and an anti-hallucination warning in GPT). Batch 4b will also take the generic Technology and Law course during the same period but will not receive a GPT subscription. The key difference is that Batch 4a will have access to the GPT subscription with a hallucination warning, while Batch 4b will not. Both groups, however, will attend the Generic Law and Technology classes at the same time that Batch 3 is receiving the JudgeGPT course. This will allow us to assess the impact of access to GPT tools on judges' learning and decision-making. Please see Figure 1 and other details in the Pre-analysis plan document for a summary of the experimental design and more details.
External Link(s)

Registration Citation

Citation
Ash, Elliot and Sultan Mehmood. 2024. "Courts of Tomorrow." AEA RCT Registry. November 01. https://doi.org/10.1257/rct.12906-1.2
Sponsors & Partners

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information
Experimental Details

Interventions

Intervention(s)
We recruited about one-third of trial court judges in Pakistan to participate in a judge training course where they will receive access to a judge support chatbot and get training in using this tool. The study will employ a randomized controlled trial design, where about half of the judges will be randomized into a first round of the course, and the other half will be randomized into a second round about 8 months later.
Intervention Start Date
2024-02-19
Intervention End Date
2024-12-31

Primary Outcomes

Primary Outcomes (end points)
1. Course attendance and participation.
2. Use of the AI tool.
3. Judgment Quality using Judgments Texts Data. The Federal Judicial Academy will share judgment texts on each judge that will allow us to construct the following main text-based measures:.
4. Perceptions of AI Utility
5. Productivity. Judges’ expectations of AI's potential to improve their productivity and their support for using generative AI at work.
Primary Outcomes (explanation)
Outcomes 1 and 2 come from our intervention. They are used to estimate a “first stage” for takeup of the intervention.
Outcome 3 is from the Federal Judicial Academy (administrative data) which obtains a random sample of 50 judgements each before and after its training programs.
Outcomes 4-5 are survey perceptions provided by the judges. They allow us to estimate the subjective impact of AI technology and training.
The subsequent surveys include additional follow-up questions on the use of AI tools, attitudes toward AI and tech, perceived work performance, and work-life balance. The follow-up surveys will be short to avoid attrition.
We have data from the course activities and assignments. This includes records on course attendance and participation, recorded in the Zoom logs and lecture software, and assignments recorded via assignment submission software. The assignments will be graded for quality by teaching assistants.
We have the full record of the judge’s interactions with the chatbot. That will include interactions during the lectures, as well as outside the lectures on homework assignments. Most importantly, it includes subsequent interactions, even after the course is done, to see if the judges use the chatbot in their work or outside of work. We can examine the text inputs and text outputs. We can see if the judges uploaded documents and their interactions, including summarization or information retrieval.

Secondary Outcomes

Secondary Outcomes (end points)
1. Case Backlog and Disposal Rate. This administrative data will be procured from the Judicial Academy and the Pakistan Law and Justice Commission of Pakistan if and when it becomes available.
2. Workload Management. Measuring the number of different cases a judge works on at any given time, along with the hours spent on work.
Secondary Outcomes (explanation)
We hope to procure administrative data . That will allow us to track changes in case backlog, case disposal rates, and the shares of various case outcomes. For example, we can look at the share of cases disposed, conviction rates, court congestions, case duration, win rates for plaintiff/defendant, and appeal rate.

Experimental Design

Experimental Design
The study will employ a randomized controlled trial design, where about half of the judges will be randomized into a first round of the course, and the other half will be randomized into a second round in the next course term.
Experimental Design Details
Not available
Randomization Method
Randomization is done by a computer.
Randomization Unit
We recruited about one-third of trial court judges in Pakistan (N=800) to participate. The study will employ a randomized controlled trial design, where about half of the judges (N=400) will be randomized into the first round of the course in February-March 2024, and the other half (N=400) will be randomized into the second round of the course in September--October 2024. Stratification is based on the time of enrollment in the course, the age of judges, and the district court in which the judges serve.
Was the treatment clustered?
Yes

Experiment Characteristics

Sample size: planned number of clusters
Standard errors will be clustered by stratification units.
Sample size: planned number of observations
760 judges
Sample size (or number of clusters) by treatment arms
760 judges
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB

Institutional Review Boards (IRBs)

IRB Name
ETH Zurich Ethics Commission
IRB Approval Date
2024-01-28
IRB Approval Number
2023-N-343
Analysis Plan

Analysis Plan Documents

Pre-Analysis Plan for Courts of Tomorrow Experiment

MD5: c01ecb8f9804cb84a6f3e86cad6a726e

SHA1: b23d04bcc6e46005cd38c1782a768547cb8ed8e8

Uploaded At: January 28, 2024

October 2024 Update Document with re-randomization of additional judges

MD5: 4c13b047c84de2a801dc031ba70c8e43

SHA1: 8a0fa00de39995f8433da23f51972e881fbdec20

Uploaded At: November 01, 2024