NIV Adjudication Review Recommendation Engine (ARRE)

Overview

Development stage

PilotSource: b) Pilot – The use case has been deployed in a limited test or pilot capacity.

High-impact designation

Not high-impactSource: c) Not high-impact

Impact justification

Not reported

Start date

2025-01-01 (day precision)Source: 01/01/2025 00:00:00

Withheld from public reporting

NoSource: a) No

Mission

Topic area

Administrative Functions

Operational functions

No function tag matched Derived by keyword rules; see methodology

Problem the AI is intended to solve

ARRE is used to make post-adjudication managerial quality assurance review time more efficient. The current required managerial review of cases includes a randomly selected portion, which is not optimized to help managers allocate time toward cases where secondary review is most likely to surface documentation gaps, policy-complex adjudications, or process compliance issues. As a result, managers may spend review capacity on routine cases while missing opportunities to identify coaching needs or quality issues within the fixed review window.

Expected benefits

The primary purpose of ARRE is to support the post-adjudication, managerial quality assurance review workflow by helping managers efficiently meet existing review requirements through queue ordering and prioritization. This provides more efficient use of manager review time and more consistent post-adjudication oversight by better targeting the fixed, required managerial review effort toward cases with atypical attributes that warrant a second look for quality assurance purposes. The expected benefits are improved review consistency and improved documentation of decision-making.

System outputs

ARRE produces a unitless anomaly score (0–1) for each already-adjudicated case and uses the score to generate a post-level ranked list that bins cases into priority tiers (e.g., high/medium/low) for managerial review. Outputs are advisory: managers may override or disregard the prioritization and may review any case consistent with existing authorities and review requirements. The output is used to help order the managerial quality assurance workload; it is not a decision output and is not used as an applicant risk determination. ARRE is used only after an adjudication is complete to support internal managerial oversight and does not serve as a basis for any visa eligibility determination or other binding action affecting the applicant. ARRE does not make, recommend, or change visa issuance/refusal decisions; all adjudicative determinations remain the responsibility of U.S. government officials.

Technology

AI classification

Classical / predictive MLSource: Classical/Predictive Machine Learning: Models trained on data to make predictions or classifications based on identified patterns or relatio…

System name(s)

Not reported

Custom-developed code

Yes

Public source code

Not reported

Sourcing

How it was built

Contract and in-houseSource: c) Developed with both contracting and in-house resources

Vendor (as reported)

Guidehouse

Vendors (standardized)

Guidehouse

A vendor is the supplier named by the agency. It does not identify the underlying model or AI technology.

Data and privacy

Involves PII

Yes

Privacy Impact Assessment

Not reported

Authorization to Operate

No

Demographic features

b) Sex; c) Age

Training and evaluation data

ARRE was trained on a random sample of NIV application data from 2021 onward drawn from the Consular Consolidated Database. Performance was evaluated through controlled tests and pilot reviews at multiple posts, where managers compared ARRE-selected cases against randomly selected cases and consistently found that ARRE’s recommendations yielded a higher proportion of cases warranting managerial review.

Federal Data Catalog

Not reported

Governance

0 of 8 minimum-practice questions answered. A blank answer means the agency reported nothing; it does not mean the practice is absent.

Potential impacts and how they were identified

Not reported