Evaluation of an artificial intelligence warning tool to detect irregularities in Paraguay’s public procurement processes

Last registered on November 17, 2025
View Trial History

Pre-Trial

Trial Information

General Information

Title
Evaluation of an artificial intelligence warning tool to detect irregularities in Paraguay’s public procurement processes
RCT ID
AEARCTR-0017223
Initial registration date
November 13, 2025

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
November 17, 2025, 2:31 PM EST

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

Locations

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

Request Information

Primary Investigator

Name
Guillermo Cruces
Affiliation
CEDLAS-FCE-UNLP
Contact Primary Investigator

Other Primary Investigator(s)

PI Name
Gaston Pierri
PI Affiliation
Inter American Development Bank
Contact Investigator

Additional Trial Information

Status
On going
Start date
2025-09-01
End date
2026-03-01
Keywords
Governance, Other
Additional Keywords
Public Procurement; Corruption Detection; Transparency; Algorithmic Monitoring; Government Efficiency; Artificial Intelligence
JEL code(s)
H57, D73, O38, H83
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
This study evaluates the impact of an AI-based analysis tool that raises "red flags" for detecting irregularities in public procurement processes in Paraguay. The intervention consists of granting procurement verifiers access to an AI alert tool integrated into their regular workflow during the adjudication phase. Using a randomized controlled design, the experiment measures whether access to automated risk signals increases the detection of potential irregularities compared to traditional verification procedures. The results will provide causal evidence on the effectiveness of digital monitoring tools to enhance transparency and efficiency in public procurement.
External Link(s)

Registration Citation

Citation
Cruces, Guillermo and Gaston Pierri. 2025. "Evaluation of an artificial intelligence warning tool to detect irregularities in Paraguay’s public procurement processes." AEA RCT Registry. November 17. https://doi.org/10.1257/rct.17223-1.0
Sponsors & Partners
Experimental Details

Interventions

Intervention(s)
Intervention Start Date
2025-09-15
Intervention End Date
2025-12-31

Primary Outcomes

Primary Outcomes (end points)
Detection of irregularities
Primary Outcomes (explanation)
Binary indicator equal to 1 if the verifier identifies any irregularity or anomaly in the procurement process during adjudication. Depends on nature of administrative dataset.

Secondary Outcomes

Secondary Outcomes (end points)
Continuous measure of the number of days (or hours) required to complete the verification process.
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
The study uses a randomized controlled design to evaluate the impact of providing procurement verifiers at Paraguay’s National Directorate of Public Procurement (DNCP) with access to an AI-based algorithmic red flags dashboard during the adjudication phase. Procurement cases will be randomly assigned—half to treatment (verifiers receive alerts indicating potential irregularities) and half to control (standard verification without alerts). The main outcomes are the probability of detecting irregularities and the time required to complete the verification process. The analysis will estimate intention-to-treat effects using linear models with fixed effects and clustered standard errors, providing rigorous causal evidence on whether algorithmic monitoring tools can enhance transparency and efficiency in public procurement.
Experimental Design Details
Not available
Randomization Method
Random assignment of each procurement process to AI tool or not.
Randomization Unit
The randomization unit is the individual process.
Was the treatment clustered?
Yes

Experiment Characteristics

Sample size: planned number of clusters
20 to 50 verifiers
Sample size: planned number of observations
In a first round (September 2025), 473 processes. About the same in the second round (November 2025).
Sample size (or number of clusters) by treatment arms
Equal split of processes into treatment and control.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
MDE of 0.2 standard deviations.
Supporting Documents and Materials
IRB

Institutional Review Boards (IRBs)

IRB Name
CEDLAS-UNLP IRB
IRB Approval Date
2025-09-01
IRB Approval Number
20250901
Analysis Plan