Person-Centric Identity Services Deduplication Model

Overview

Development stage

DeployedSource: c) Deployed – The use case is being actively authorized or utilized to support the functions or mission of an agency.

High-impact designation

High-impactSource: a) High-impact

Impact justification

Not reported

Start date

2023-02-01 (day precision)Source: 2023-02-01T00:00:00

Withheld from public reporting

Not reported

Mission

Topic area

Law Enforcement

Operational functions

Claims and case processing Derived by keyword rules; see methodology

Problem the AI is intended to solve

Critical to the success of PCIS is the entity resolution and de-deduplication of individual records from various systems of records to create a complete picture of a person. Using machine learning (ML), the model can identify which case management records belong to the same unique individual with a high degree of confidence. This allows PCIS to compile a full immigration history for an individual without the need for time-consuming research across multiple disparate systems. The de-duplication model plays a critical role in the entity resolution and surfacing of a person and all their associated records. The ML models are more resilient to fuzzy matches and handle varying data fill rates more reliably.

Expected benefits

Using Machine Learning allows us to improve entity resolution as compared to rule based system. PCIS offers the ability to see a person's immigration history organized in one place. Specific benefits do or will include: an organized summary view of the identity with the individual's latest photo from PCIS; full immigration history including receipts associated with the applicant, regardless of case management system; mailing, physical, and safe history of the individual organized in reverse chronological order, allowing users to easily find the most recent address; and all identifiers associated with the applicant, including A-Numbers, FINs, SSNs, SSNs, ELIS account numbers, passport numbers, etc.

System outputs

Numerical likelihood score which is used to determine if the record belongs to the individual. Likelihood scores are subjected to a high threshold (.98, maximum 1) to assess whether the record belongs to the individual.

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)

Person Centric Identity Services

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)

MetroIBR

Vendors (standardized)

None on the standard list

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

www.dhs.gov ↗

Authorization to Operate

Yes

Demographic features

Not reported

Training and evaluation data

USCIS-only data derived from 7 form-processing source systems including C3, ELIS, CPMS, GLOBAL, CIS2, AR-11, CAMINO.

Federal Data Catalog

Not reported

Governance

8 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

There is small risk of false positives or negatives, which are identified and sent to Manual Resolution Queue. The queue is processed by authorized and trained personnel. Human review is still done for the actual benefit or request being sought. AI is used to identify the person seeking the benefit or request.