United States Citizenship and Immigration Services – AI Use Cases

Use Case Name: Case Processing Improvements in FDNS-DS NexGen (formerly FDNS-DS NexGen) 

Use Case ID: DHS-17 

Use Case Summary: USCIS created the Fraud Detection and National Security (FDNS) Directorate to strengthen the integrity of the nation’s immigration system and to ensure that immigration benefits are not granted to individuals who may pose a threat to national security and/or public safety. In addition, the FDNS Directorate is responsible for detecting, deterring, and combating immigration benefit fraud.  In 2005, USCIS developed a case management system, the Fraud Detection and National Security-Data System (FDNS-DS), to record, track, and manage the screening processes related to immigration applications, petitions, or requests with suspected or confirmed fraud, public safety, or national security concerns, and to identify vulnerabilities that may compromise the integrity of the legal immigration system.  In June 2023, FDNS-DS was replaced with a modernized case management system, FDNS-DS NexGen.  In the future, FDNS-DS NexGen may use artificial intelligence (AI) and machine learning (ML) data from other applications to aid in investigative work, enhance investigative case prioritization, and detect duplicate case work. USCIS may also integrate AI/ML into the predictive modeling for future system enhancements, working side-by-side with the business stakeholders to develop best practices. Fraud occurs in numerous ways; being able to discover and detect persons with multiple identities allows for more comprehensive investigations, reduces investigative cycle time, and improves performance. Future implementation of AI/ML techniques will speed up case and investigative processing by several magnitudes. For more information, please visit: DHS/USCIS/PIA-013-01 Fraud Detection and National Security Directorate. While FDNS-DS NexGen is in production, the AI/ML functionality is still in discovery stages. 

Use Case Topic Area: Law & Justice 

Deployment Status: Pre-deployment (Initiation) 

Safety- and/or rights-impacting? No, use case is too new to fully assess impacts; will be reassessed before end of initiation stage. 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: User Entity and Behavior Analytics (UEBA) for Security Operations (SecOps) Anomaly Identification 

Use Case ID: DHS-372 

Use Case Summary: User Entity and Behavior Analytics (UEBA) assists USCIS Security Operations (SecOps) in identifying behavioral anomalies that most likely indicate malicious intent or heightened risk associated with user identities and endpoint hosts accessing the USCIS network. The analytics provide risk scoring, which helps USCIS SecOps to prioritize highest risk incidents first. The tool identifies credential compromise, risky behavior, violations of the rules of behavior, and other user behavior anomalies that could indicate the presence of an advanced persistent threat. UBEA is enhanced by machine learning models that can adapt to its environment and identify relevant threats, thereby improving the productivity of the USCIS Security Operations Center (SOC). 

UEBA reviews USCIS system logs to determine when an entity, such as a workstation, server, or internal USCIS system account, is performing actions that are anomalous.  The UEBA ingests logs from systems to perform analytics based on manually created and maintained models. These models apply a risk score to the entity, which is then used to create a case (or ticket) for SecOps analyst to review. The AI reviews the actions of the analysts to adjust the risk scoring for future events, assisting in prioritizing cyber events for further manual investigation. 

Use Case Topic Area: Mission-Enabling (internal agency support) 

Deployment Status: Pre-deployment (Acquisition and/or Development) 

Safety- and/or rights-impacting? No 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: I-765 - USCIS Face Capture Mobile App 

Use Case ID: DHS-414 

Use Case Summary: Utilizing facial capture in the mobile application reduces the burden on both Application Support Centers and customers by providing an alternative method for capturing card-quality photographs used for document production. Application Support Center appointments are limited to work hours, often requiring applicants to take time off work. In some regions, applicants may need to travel long distances to reach an application support center and inclement weather can make attending appointments difficult. The mobile application enables photograph collection without the need for a biometrics appointment, enhancing processing efficiency for adjudicators and offering greater convenience for applicants. This will allow the user to complete the biometric verification requirement without having to attend an appointment at an Applicant Support Center. This reduces the burden on the beneficiary and decreases demands on USCIS Applicant Support Center resources. 

Face detection locates human faces in visual media such as digital images or video. When a face is detected, it has an associated position, size, and orientation; it can also be searched for landmarks such as the eyes and nose, which are returned as numerical values. The ELIS Photo Validation service sends a response back to the user based on predefined quality checks to determine if the uploaded photo meets USCIS requirements. The I-765 - USCIS Facial Recognition through IDENT (1:1 Face Recognition/Validation) returns a match or no match response from IDENT. 

Use Case Topic Area: Law & Justice 

Deployment Status: Pre-deployment (Acquisition and/or Development) 

Safety- and/or rights-impacting? Yes. Rights-impacting. Before this AI use case is deployed, it will comply with risk management practices for deployed safety- and/or rights-impacting AI. Read more about compliance with required minimum risk management practices. 

Face Recognition/Face Capture (FR/FC)? Yes. All Face Recognition and Face Capture (FR/FC) technology is tested both prior to operational use and at least every three years during operational use. DHS Science and Technology (S&T) oversees testing and evaluation based on International Organization for Standardization/ International Electrotechnical Commission (ISO/IEC) standards and technical guidance issued by National Institute of Standards and Technology (NIST). DHS S&T applies laboratory, scenario, and operational testing to cost-effectively characterize technology performance and, when feasible, disaggregate performance by user demographics such as gender, age, and skin tone. 

Use Case Name: USCIS Translation Service 

Use Case ID: DHS-2305 

Use Case Summary: USCIS is testing AI-powered tools to quickly and accurately translate documents and provide real-time interpretation, enhancing communication and processing for immigration services. The USCIS Translation Service assesses the effectiveness and security of GenAI-powered translation tools for document translation and real-time interpretation. AI and GenAI models trained on relevant subjects (e.g., immigration law, visa applications, family/adoption certificates) provide fast, accurate translation of written or other digital documents in various languages. AI and GenAI-powered speech-to-speech and speech-to-text translation tools enable efficient communication, consultations, and other interactions within DHS. Supporting language translation and transcription is crucial for operations and adjudications across all DHS components. One use case application involves United States Refugee Admissions Program (USRAP) document translation, a time-consuming task for Resettlement Support Center (RSC) caseworkers before USCIS interviews. Documents include passports, national identifications, birth certificates, and more complex documents. The goal is to reduce or eliminate RSC casework dedicated to document collection, translation, data extraction, and summarization by leveraging document translation services. This allows USCIS officers to translate documents ad hoc during interviews.  

USCIS Translation Service use case addresses 4 key problem areas: 1) Limited capacity: Traditional translation methods rely on a finite pool of human translators, causing bottlenecks and delays. 2) Accuracy variability: Consistency and nuance can vary with individual translator expertise, especially with the diversity of languages in the Western Hemisphere and at the Southern Border. AI services aim to enhance consistency and address unique dialects. 3) Cost constraints: Scaling multilingual services through traditional methods can be resource-intensive and expensive. 4) Accessibility limitations: Real-time, on-demand translations are often unavailable, delaying communication and processing. 

Use Case Topic Area: Government Services (includes Benefits and Service Delivery) 

Deployment Status: Pre-deployment (Initiation) 

Safety- and/or rights-impacting? Yes. Rights-impacting. Before this AI use case is deployed, it will comply with risk management practices for deployed safety- and/or rights-impacting AI. Read more about compliance with required minimum risk management practices. 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Verification Match Model 

Use Case ID: DHS-2384 

Use Case Summary: The Verification Match Model utilizes machine learning (ML) to gather information from multiple systems and match it to known records by comparing names, dates of birth, and other identifiers. This supports two key USCIS Verification Division (VER) programs: the Employment Verification Program (E-Verify) and the Systematic Alien Verification for Entitlements (SAVE) Program.  The model streamlines verification data integration services by providing a ranked order of the best matches, expressed as confidence scores, between the identifiers used by the two programs. These improved matches increase the accuracy of E-Verify and SAVE, ensuring they use the most recent and accurate documents to determine benefit status. E-Verify handles approximately 250,000 requests per day and SAVE handles approximately 70,000 requests per day. Multiple data verification models for SAVE and E-Verify exist, complicating efficiency and performance. By consolidating these into a single, unified Verification Match Model within a separate microservice, the use case aims to improve the accuracy of responses and reduce the need for manual review. ML plays a key role in the continuous improvement of these models, ultimately reducing the need for manual case reviews. 

Leveraging AI in the USCIS verification matching process of known records across systems is beneficial because it streamlines existing USCIS reviews by 1) improving system accuracy, 2) reducing human error through automated person-and-record match scoring, and 3) handling a higher volume of matches than traditional tools or manual processes can achieve.  The ouput is a recommendation and score indicating person-and-record match probability, used by verification systems (E-verify and SAVE) to improve accuracy in initial system responses. 

Use Case Topic Area: Mission-Enabling (internal agency support) 

Deployment Status: Pre-deployment (Initiation) 

Safety- and/or rights-impacting? Yes. Rights-impacting.  Before this AI use case is deployed, it will comply with risk management practices for deployed safety- and/or rights-impacting AI. Read more about compliance with required minimum risk management practices. 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Sentiment Analysis - FOD Field Offices Complaints and Reviews 

Use Case ID: DHS-2386 

Use Case Summary: The Sentiment Analysis - FOD Field Offices Complaints and Reviews system analyzes feedback received by USCIS to determine the positive or negative sentiments expressed. This system provides a statistical analysis of quantitative results from complaints, reviews the results, and then uses machine learning (ML) techniques to assign sentiments to categories ranging from strongly positive to strongly negative. This helps survey administrators to identify trends and glean valuable from both quantitative and qualitative data. This capability is currently available and under consideration by the Field Operations Directorate (FOD). FOD receives complaints from the public through the USCIS Office of Investigations (OI) via telephone, mail, email, and a dedicated website. These complaints contain detailed descriptions of the in-office experiences. FOD personnel read, research, and resolve these complaints on a case-by-case basis. FOD uses the Sentiment Analysis - FOD Field Offices Complaints and Reviews system for ML analysis to provide an extra layer of information that helps FOD identify and resolve issues generally (without changing the process of resolving complaints). Currently, FOD reports complaints counts by pending and resolved and general bins determined by FOD HQ. It’s this binning that FOD would like to improve because they may be missing key trends and/or early identification of general issues. FOD were looking for generalization, a summary of aggregated complaints. 

The system categorizes feedback sentiments into positive, negative, or neutral tones, displayed in a dashboard. This graphical representation helps visualize how customer service can be improved but does not provide specific recommendations or decisions. 

Use Case Topic Area: Mission-Enabling (internal agency support) 

Deployment Status: Pre-deployment (Initiation) 

Safety- and/or rights-impacting? No 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Biometrics Enrollment Tool (BET) Fingerprint Maximization 

Use Case ID: DHS-14 

Use Case Summary: The Biometric Enrollment Tool (BET) is custom-built biometric enrollment software for both domestic and overseas USCIS biometric collections. BET is integrated with the National Appointment Scheduling System (NASS) and the Customer Profile Management System (CPMS). This improves data integrity as BET utilizes data already collected in USCIS systems, reducing the amount of data entry by the Biometric Technician and the risk of typographic errors. BET assists in determining if the fingerprint taken is good enough quality to pass the FBI fingerprint check process. It provides immediate feedback when a set of prints is likely to be rejected by the FBI by incorporating machine learning (ML) models into the BET application. The FBI will not disclose their quality grading criteria for fingerprints, leaving BET with the responsibility of determining quality to prevent unnecessary secondary encounters with applicants. Using even the simplest of models would catch 98% of rejected submissions, which could have potentially saved USCIS from scheduling 42,763 additional appointments in 2020. This would come at the cost of forcing recapture during 11% of encounters. This effort aims to maximize the number of successful FBI submissions while minimizing the number of fingerprint recaptures necessary. The output is a Numerical Fingerprint Quality score, which is compared against fingerprint quality thresholds (per finger and per set of fingerprints) to align with FBI specifications. 

Use Case Topic Area: Law & Justice 

Deployment Status: Deployed (Operation and Maintenance) 

Safety- and/or rights-impacting?  No. It was presumed rights-impacting relating to immigration, but the DHS Chief AI Officer determined this use case does not satisfy the definition of rights-impacting AI in M-24-10. This use case simply identifies whether a fingerprint collected is of sufficient quality to pass the FBI fingerprint check process, ultimately maximizing the number of successful FBI submissions while minimizing the number of fingerprint recaptures necessary. This quality assurance step is one task in a series of adjudication activities but is not determinative of the overall adjudication decision. The tool saves personnel time and resources while enhancing customer experience by helping to ensure that only quality fingerprints are passed forward for matching against the FBI Identity History Summary Check. Results of FBI fingerprint checks are subsequently reviewed by a human as part of the immigration adjudicative process. Read more about safety and/or rights-impacting AI.] 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: ELIS Evidence Classifier Machine Learning (ML) Tagging Solution (formerly Evidence Classifier) 

Use Case ID: DHS-16 

Use Case Summary: The Evidence Classifier Service is a machine learning (ML) solution that reduces the time spent by adjudicators and contractors sifting through digital evidence. The solution systematically tags and surfaces critical evidence types for the adjudicators in Electronic Immigration System (ELIS). Until the Evidence Classifier ML solution's introduction, those working on cases and responsible for reviewing evidence documents would often have to sift through dozens, if not hundreds, of unlabeled pages to find one specific artifact. As a result, an ML solution was built to systematically tag individual pages with some of the highest-volume, highest-impact evidence types. 

The service enables end users to navigate directly to the page(s) containing evidence documents of interest, instead of sifting through large PDF documents. Evidence tagging intends to accelerate case processing by identifying specific types of documents (e.g., I-589, passport photo spread, marriage certificate) and applying a metadata tag to that document object in ELIS. When a user opens a case with potentially hundreds of pages of evidence documents they have clickable bookmarks from these tags that will jump directly to the corresponding page. 

The output is tagged evidence. The system inputs an image (scanned document from Lockbox) and outputs either a specific label, such as "Border Crossing Card - Front," or no label if that document is not recognized as one of the classes. 

Use Case Topic Area: Government Services (includes Benefits and Service Delivery) 

Deployment Status: Deployed (Operation and Maintenance) 

Safety- and/or rights-impacting? No 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Person-Centric Identity Services Deduplication Model 

Use Case ID: DHS-55 

Use Case Summary: The vision of Person-Centric Identity Services (PCIS) is to be the authoritative source of trusted biographical and biometric information that provides real-time, two-way visibility between services into an individual's comprehensive immigration history and status. The de-duplication model ingests person-centric datasets from various source systems for model training and evaluation purposes. Our dataset includes biographic information (e.g., name, date of birth, alien number Social Security Number, passport number) as well as biographic information (e.g., fingerprint IDs, eye color, hair color, height, weight) for model training and matching purposes. 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. 

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; all identifiers associated with the applicant, including Alien number, Social Security Number, passport number, etc.; and a numerical likelihood score which is used to determine if the record belongs to the individual, subjected to a high threshold (.98, maximum 1) to assess whether the record belongs to the individual. 

Use Case Topic Area: Law & Justice 

Deployment Status: Deployed (Operation and Maintenance) 

Safety- and/or rights-impacting? Yes. Rights-impacting 

Key Identified Risks & Mitigations: 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.  

Read more about safety and/or rights-impacting AI and compliance with required minimum risk management practices. 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Person-Centric Identity Services A-Number Management Model 

Use Case ID: DHS-56 

Use Case Summary: This use case verifies the ongoing accuracy of the Person-Centric Identity Services (PCIS) information compilation.  PCIS is the authoritative source of trusted biographical and biometric information that provides real-time, two-way visibility between authorized services into an individual's comprehensive immigration history and status, with the purpose of providing DHS employees all needed information to review and process cases in a single system. The aim of this use case is to leverage machine learning (ML) to test the accuracy of PCIS to identify and manage associations between individuals and their assigned alien numbers (A-numbers), which is a unique 7-, 8-, or 9-digit number assigned to a noncitizen by DHS. The A-number plays a critical role in surfacing of a person and all their associated records from across PCIS. The output of the use case is the numerical confidence score which is used to determine the validity of the A-number presented in search results. The confidence score identifies which records from within PCIS best match search criteria for an A-number. 

Use Case Topic Area: Law & Justice 

Deployment Status: Deployed (Operation and Maintenance) 

Safety- and/or rights-impacting? No. It was presumed rights-impacting relating to immigration, but the DHS Chief AI Officer determined this use case does not satisfy the definition of rights-impacting AI in M-24-10. The AI outputs of this use case are used to verify the ongoing accuracy of information compilation pertaining to an individual’s identity and their immigration history and status. The use case only identifies which records from within PCIS best match search criteria for an A-Number to support case processing. This use case provides more efficient data compilation and management. Results are reviewed and validated by USCIS adjudicators as part of their overall case review. Read more about safety and/or rights-impacting AI. 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Identity Match Option (IMO) Tool for Record Compilation (formerly Identity Match Option (IMO) Process with DBIS Data Marts) 

Use Case ID: DHS-57 

Use Case Summary: USCIS uses the Identity Match Option (IMO) tool to aid in person-centric research and analytics. More specifically, IMO is used to derive a single identity across multiple systems for each applicant or beneficiary who interacts with USCIS. USCIS maintains a variety of systems to track specific interactions with individuals – benefits case management, appointment scheduling, background check validation, and customer service inquiries. Each system captures its own person-centric data attributes (e.g., Social Security Number (SSN), Alien number (A-number), name, date of birth (DOB), address, etc.) related to individuals interacting with the agency. The system is designed to handle different data formats and potential data quality issues present in the source data to identify duplicate identities in a data set or between two data sets. The derived identities are linked back to the original source records, enabling USCIS personnel to view an individual's comprehensive interaction history with the agency.  The output of this can be visualized through a report or dashboard to assist with case review, ensuring access to helpful and accurate records. 

A user-friendly dashboard is created to display results and shows the data pattern but does not allow for any prediction or decision making. IMO is a Commercial Off-The-Shelf (COTS) product offered by Informatica. The product has a variety of “transformations” that can be used together to build a workflow-based solution. There are several different algorithms (Soundex, Jaro, Hamming distance, etc.) which are available for use in performing identity resolution. 

Use Case Topic Area: Mission-Enabling (internal agency support) 

Deployment Status: Deployed (Operation and Maintenance) 

Safety- and/or rights-impacting? No. It was presumed rights-impacting relating to government benefits, but the DHS Chief AI Officer determined this use case does not satisfy the definition of rights-impacting AI in M-24-10. This use case compiles records from across a variety of USCIS systems to provide a comprehensive history of a person’s interaction with USCIS. The output of this can be visualized through a report or dashboard to assist with case review ensuring access to helpful and accurate records. Adjudicators review the outputs of this use case, alongside other information and insights, to process a case and make a final determination.  The adjudication process can be conducted without this tool, however, doing so would significantly increase the time and effort required to process immigration requests. Read more about safety and/or rights-impacting AI.

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Text Analytics Data Science Sentence Similarity Model 

Use Case ID: DHS-130 

Use Case Summary: USCIS oversees lawful immigration to the United States. Through its Refugee, Asylum & International Operations Directorate (RAIO), USCIS administers programs to provide protection to qualified individuals who have suffered past persecution or have a well-founded fear of future persecution in their country of origin, as outlined in the Immigration and Nationality Act (INA) and Title 8 of the Code of Federal Regulations (C.F.R.). The Text Analytics capability employs machine learning and data graphing techniques to identify patterns that may indicate potential fraud, national security, and/or public safety concerns by scanning the digitized narrative sections of the associated applications and looking for common language patterns. The Text Analytics Data Science Sentence Similarity Model identifies similarity and relevancy in text to determine potential matches in documents submitted into the system. Text Analytics does not make any determinations or decisions but is instead utilized as a research tool by staff in the course of their duties. 

Text Analytics augments the tedious and time-consuming manual process of identifying potential fraud, national security, and/or public safety concerns and enables the identification of such concerns across jurisdictional boundaries. It increases the integrity of immigration programs, strengthens officers’ confidence in their work, and contributes to the reduction in customer wait times. 

Text Analytics does not make predictions, recommendations, or decisions. It is merely a research tool that identifies potential patterns while remaining agnostic as to whether those patterns indicate potential fraud, national security, and/or public safety concerns. Instead, trained staff evaluate the patterns to determine whether they identify potential concerns and then validate and/or invalidate those potential concerns through the course of their investigations or adjudications. 

Use Case Topic Area: Law & Justice 

Deployment Status: Deployed (Operation and Maintenance) 

Safety- and/or rights-impacting? Yes. Rights-impacting.  

Key Identified Risks & Mitigations: There is a small risk of false positives or false negatives due to the model. This risk is mitigated through a manual review of any information produced by the tool. Text Analytics is a decision support tool. It does not make recommendations regarding fraud or benefit/adjudication decisions. Any decisions made based on information stored in the tool are conducted through a manual review by a USCIS employee. 

Read more about safety and/or rights-impacting AI and compliance with required minimum risk management practices. 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Automated Name and Date of Birth (DOB) Harvesting from Existing Records 

Use Case ID: DHS-180 

Use Case Summary: Automated Name and Date of Birth (DOB) Harvesting is being developed to reduce the amount of adjudicative time spent manually harvesting aliases and DOBs from Identity History Responses (IdHS) text. The use case output draws from FBI Background Check System (BCS) source information and is provided to USCIS officers to consider during case processing. For aliases, this solution uses a named-entity recognition (NER) model trained to understand the specific context around where aliases appear—meaning it’s effective at locating true aliases, while at the same time avoiding other places where proper names may exist but are not aliases (e.g., county or street names). For DOBs, this solution uses pattern matching and conditional logic to harvest valid birth dates, while weeding out obvious placeholders and other non-birth dates found within the IdHS text.  This use case has been in production since 2022 and is enabled only for N400. The solution was developed as a Python library, and the encasing Flask application runs on Amazon EKS. Its primary function is to predict where aliases and dates of birth appear within unstructured IdHS text.  It uses a named-entity recognition (NER) model, trained using Spark NLP with a TensorFlow back end on thousands of systematically labeled IdHS records, and word-level embeddings. 

This use case reduces the amount of adjudicative time spent manually harvesting aliases and dates of birth (DOBs) from identity history summary (IdHS) report attached to the ELIS case as part of the Manual Name Harvesting Task during case processing. 

Eliminates need for manual review by extracting unique names and DOBs from IdHS documents and when names/DOBs are already in ELIS, ANH will not suggest any names. This is still a human in the loop process. The ELIS user performing MNH tasks is prompted to decide if the suggested names and DOBs are related to case hence can accept/reject the suggestions. 

Use Case Topic Area: Government Services (includes Benefits and Service Delivery) 

Deployment Status: Deployed (Operation and Maintenance) 

Safety- and/or rights-impacting? No. It was presumed rights-impacting relating to immigration, but the DHS Chief AI Officer determined this use case does not satisfy the definition of rights-impacting AI in M-24-10. AI outputs of this use case are suggested aliases and DOBs related to the individual query, which USCIS staff must review to accept, reject, or ignore the suggested information. The AI outputs reduce the amount of adjudicative time spent manually harvesting aliases and DOBs.  The use case increases efficiency of tasks associated with reviewing existing records for adjudicating requests for immigration benefits. Completing such adjudications are not dependent on the use case, however lack of this tool would significantly increase human processing times and potentially reduce the accuracy of information consulted during the human review process. Read more about safety and/or rights-impacting AI. 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: Automated Realtime Global Organization Specialist (ARGOS) for Company Registration Submissions to E-Verify 

Use Case ID: DHS-181 

Use Case Summary: ARGOS searches publicly available datasets (like a human would) to establish risk and potential fraud associated with a company using E-Verify. The use case output informs USCIS Verification Account Compliance (VAC) analysts' decisions regarding whether a company applying to register on E-Verify is fraudulent and should be referred for investigation. ARGOS searches a multitude of public datasets for information that is then ranked and scored based on text analysis and a risk and fraud model developed in collaboration with the USCIS VAC team. The ARGOS application then displays that information to the end-user, allowing them to view the possibilities of fraud and make an informed decision. The goal of the use case is to streamline the process and accelerate work for the individual who is researching a company that has submitted its information for registration to E-Verify. 

ARGOS sentiment analysis produces a risk score, and keyword extraction identifies the keyword category of interest to the VAC management and program analysts (MPAs) for the aggregated open-source information. This helps quickly identify any pertinent information to aid the MPAs in their open-source investigation of company applications. This process saves potentially thousands of MPA man-hours in open-source investigation and creates a single source of truth for each MPA's investigation of a company application. In turn, this allows for quicker application processing and, if the risk of company fraud exists, much faster referral processing time, expediting the next-step referral to Fraud Detection and National Security (FDNS) for further investigations. 

Responses are sent back to a user dashboard accessible internally only by VAC Management and Program Analyst (MPA) personnel. Keywords related to the MPA's work interests are extracted if present, and risk scores are assigned to the open source collected information. The data is presented to the MPA on the graphical user interface (GUI) dashboard. 

Use Case Topic Area: Mission-Enabling (internal agency support) 

Deployment Status: Deployed (Operation and Maintenance)  

Safety- and/or rights-impacting? Yes. Rights-impacting 

Key Identified Risks & Mitigations: Data Collection Bias: search engine prioritization of certain content may skew results. Lack of Domain-Specific Accuracy: model tested on company data across different industries resulted in inconsistent performance. Limited Generalization to Unseen Data: model’s performance on validation datasets lower than on training data, indicating potential overfitting. Misinterpretation of Sentiment: instances of sarcasm/irony not recognized. All risks identified in testing and evaluation phases. Bias monitoring, fine-tuning data balancing, and statistical precision deviation monitoring have been put in place. 

Read more about safety and/or rights-impacting AI and compliance with required minimum risk management practices. 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: ELIS Card Photo Validation via myUSCIS 

Use Case ID: DHS-189 

Use Case Summary: USCIS uses a system called ELIS to manage immigration requests, and it includes a Photo Validation Service to check if ID photos meet requirements. This helps ensure photos are correct before making ID cards, saving time and avoiding delays. USCIS is the component within DHS that oversees lawful immigration to the United States. USCIS receives immigration requests from individuals seeking immigration and non-immigration benefits. Once a benefit request form is submitted to USCIS, a series of processing and adjudication actions occur. One of the case management systems used to track and adjudicate certain immigration request forms is the Electronic Information System (ELIS). USCIS ELIS is an internal case management system composed of microservices to assist with performing complex adjudicative and processing tasks; one of those microservices is the Photo Validation Service. The card photo validation solution was originally designed for e-filed I-765 (c)(3)s — to be in direct support of the beneficiary as they apply via myUSCIS. In enabling e-filing for the (c)(3)s, one goal was to have everything captured digitally, including these ID photos, all while avoiding any new delays or complications for the applicant in the process. In other words, if we observed an issue with a photo upon upload, we wanted to address it up front — rather than waiting until a failed card production sequence.  Because the photo requirements themselves are fixed, this exact same functionality could also be plugged directly into ELIS, just prior to card production. As of late October 2021, an effort to do exactly this is underway. The primary difference is that an adjudicator would explicitly request that a case’s photo be validated, and if applicable, the service would return a recommendation to edit the photo before submission. 

The photo validation service uses a combination of computer vision techniques and machine learning models to validate photographs and ensure they meet the requirements, so these photographs can be used in card production. The output is Response back to user based on the pre-defined quality checks if the uploaded photo meets USCIS requirements. Users still have the option to ignore the warnings and upload the photo. 

Use Case Topic Area: Government Services (includes Benefits and Service Delivery) 

Deployment Status: Deployed (Operation and Maintenance) 

Safety- and/or rights-impacting? No.  It was presumed rights-impacting relating to immigration, but the DHS Chief AI Officer determined this use case does not satisfy the definition of rights-impacting AI in M-24-10. The AI outputs of this use case are a real-time advisory message to the submitter of the photo that it is of insufficient quality for card production at which point the submitter may choose to resubmit or not. While adjudication of EADs is not dependent on this use case, this tool enhances efficiency and customer service by minimizing the number of cards failing in production and delays related to subsequent requests for information. Read more about safety and/or rights-impacting AI. 

Face Recognition/Face Capture (FR/FC)? Yes. All Face Recognition and Face Capture (FR/FC) technology is tested both prior to operational use and at least every three years during operational use. DHS Science and Technology (S&T) oversees testing and evaluation based on International Organization for Standardization/ International Electrotechnical Commission (ISO/IEC) standards and technical guidance issued by National Institute of Standards and Technology (NIST). DHS S&T applies laboratory, scenario, and operational testing to cost-effectively characterize technology performance and, when feasible, disaggregate performance by user demographics such as gender, age, and skin tone. 

Use Case Name: Large Language Models for an Officer Training Tool 

Use Case ID: DHS-366 

Use Case Summary: The Large Language Model (LLM) for Officer Training tool will use Generative AI (GenAI) to improve the way the agency trains immigration officer personnel. The tool will generate dynamic, personalized training materials that adapt to officers’ specific needs and ensure the best possible knowledge and training on a wide range of current policies and laws relevant to their jobs. The goal is to enhance trainees’ understanding and retention of crucial information, increase the accuracy of their decision-making process, and limit the need for retraining over time. The AI outputs human-like conversation in a text format, like other AI chatbots, but is specially trained and tuned for Refugee, Asylum, and International Operations (RAIO) Officer training.

Use Case Topic Area: Mission-Enabling (internal agency support) 

Deployment Status: Deployed (Implementation and Assessment) 

Safety- and/or rights-impacting? No 

Face Recognition/Face Capture (FR/FC)? No 

Use Case Name: I-765 - USCIS Facial Recognition through IDENT (1:1 Face Recognition/Validation) 

Use Case ID: DHS-413 

Use Case Summary: USCIS intends to leverage facial recognition for several reasons. Multiple court orders require the adjudication of employment authorization applications within 30 days. Currently, applicants must attend an Application Support Center to verify their identity and capture a photograph, which can take up to three weeks of the 30-day period. Using the facial verification service makes this process nearly instant and greatly enhances processing efficiency without compromising the effectiveness of current identity verification methods. Additionally, new rules created by the FBI’s Compact Council require biometric verification to perform fingerprint resubmissions for different filing reasons than the original fingerprint capture. Facial verification brings USCIS into compliance with this rule change. Furthermore, performing facial verification increases the integrity of information and identities by identifying conflicts early on, preventing issues from becoming pervasive across immigration systems. This will allow the user to complete the biometric verification requirement without having to attend an appointment at an Applicant Support Center. This reduces the burden on the beneficiary as well as reducing demands on USCIS Applicant Support Center resources. The output is a match or no match response from IDENT.  

Use Case Topic Area: Law & Justice 

Deployment Status: Deployed (Operation and Maintenance) 

Safety- and/or rights-impacting? Yes. Rights-impacting 

Key Identified Risks & Mitigations:  Potential mismatch of face images and/or bias based on demographic data held by USCIS. USCIS is in the process of developing a reporting mechanism for false negative matches that will highlight any disparate impacts on various demographics. USCIS will produce an internal annual overview of the facial recognition results and any data points USCIS has about demographics. The use case has been found to not lead to or perpetuate unlawful discrimination or bias. 

Read more about safety and/or rights-impacting AI and compliance with required minimum risk management practices. 

Face Recognition/Face Capture (FR/FC)? Yes. All Face Recognition and Face Capture (FR/FC) technology is tested both prior to operational use and at least every three years during operational use. DHS Science and Technology (S&T) oversees testing and evaluation based on International Organization for Standardization/ International Electrotechnical Commission (ISO/IEC) standards and technical guidance issued by National Institute of Standards and Technology (NIST). DHS S&T applies laboratory, scenario, and operational testing to cost-effectively characterize technology performance and, when feasible, disaggregate performance by user demographics such as gender, age, and skin tone. 

Use Case Name: Intelligent Document Processing (IDP) for I-539 Form Digitization 

Use Case ID: DHS-2385 

Use Case Summary: Intelligent Document Processing (IDP) for the Application to Extend/Change Nonimmigrant Status Form I-539 uses an Artificial Intelligence (AI)-enhanced tool to identify, categorize, and create separate images for each document type submitted as part of the I-539 benefit application. Before the use case, all pages of a I-539 application were scanned and stored as a single document in the content management system, delaying adjudication and not meeting National Archives and Records Administration (NARA) standards. The tool uses a learning model to identify, classify, and separate individual documents into their component parts for storage. For example, a USCIS form, marriage license, passport, bank statement, etc., will be identified and stored in the content management system (CMS or Electronic Immigration System (ELIS)) as individual images and linked as supporting documents for the I-539 benefit application. This streamlines case processing by organizing supporting information for easy access and brings digitization of the I-539 benefit application into compliance with NARA standards. A human in-the-loop architecture resolves uncertain results from the system. 

IDP for the I-539 reduces case processing time for adjudicators by identifying and classifying supporting documents for ease of use and brings digital images into compliance with NARA standards. 

Use Case Topic Area: Mission-Enabling (internal agency support) 

Deployment Status: Deployed (Implementation and Assessment) 

Safety- and/or rights-impacting? No 

Face Recognition/Face Capture (FR/FC)? No

Use Case Name: Asylum Text Analytics 

Use Case ID: DHS-13 

Use Case Summary: USCIS oversees lawful immigration to the United States. As set forth in Section 451(b) of the Homeland Security Act of 2002, Public Law 107-296, Congress charged USCIS with administering the asylum program. USCIS, through its Asylum Division within the Refugee, Asylum & International Operations Directorate (RAIO), administers the affirmative asylum program to provide protection to qualified individuals in the United States who have suffered past persecution or have a well-founded fear of future persecution in their country of origin, as outlined under Section 208 of the Immigration and Nationality Act (INA), 8 U.S.C. § 1158 and Title 8 of the Code of Federal Regulations (C.F.R.), Part 208. Generally, an individual not in removal proceedings may apply for asylum through the affirmative asylum process regardless of how the individual arrived in the United States or his or her current immigration status by filing Form I-589, Application for Asylum and for Withholding of Removal. The ATA capability employs machine learning and data graphing techniques to identify plagiarism-based fraud in applications for asylum status and for the withholding of removal by scanning the digitized narrative sections of the associated forms and looking for common language patterns. 

Deployment Status: Inactive (no longer used) 

Use Case Name: Timeseries Analysis and Forecasting 

Use Case ID: DHS-20 

Use Case Summary: USCIS is the component within DHS that oversees lawful immigration to the United States. That means USCIS receives, processes, and maintains all applications for admission for Lawful permanent residents (LPRs), or adjustments to LPR status. Also known as “green card” holders, LPRs are non-citizens who are lawfully authorized to live permanently within the United States and are required to fill out Form I-90, Application to Replace Permanent Resident Card (Green Card). Since there has been a considerable influx of green card applications, USCIS used a combination of exploratory data analysis to determine the most used categories for applicants submitting I-90's and machine learning to create predictions of workloads. As a follow-on, USCIS used Autoregressive Integrated Moving Average (ARIMA) models on the I-90 form, which allowed the prediction of the total number of forms for a 2-year period. ARIMA is one of the easiest and effective machine learning algorithms to perform time series forecasting. This capability has been deployed in production for more than a year. This model was eventually enhanced using ML model to have better reusability and performance. 

Deployment Status: Inactive (no longer used) 

Use Case Name: Sentiment Analysis - Employee Satisfaction Surveys 

Use Case ID: DHS-58 

Use Case Summary: The Sentiment Analysis - Surveys system provides a statistical analysis of quantitative results from survey results and then uses Natural Language Processing (NLP) modeling software to assign "sentiments" to categories ranging from strongly positive to strongly negative. This work has not gone beyond proof-of-concept and is not in use by any business unit. The goal of the use case was to determine whether sentiment analysis was technically feasible in the context of employee satisfaction surveys. While the technical capability was proven, potential privacy concerns were also identified. There were/are no plans to operationalize the technical capability proof-of-concept until/unless the privacy blockers are resolved. 

Deployment Status: Inactive (no longer used) 

Use Case Name: Sentiment Analysis – ELIS Case Notes 

Use Case ID: DHS-59 

Use Case Summary: Once a benefit request form is submitted to USCIS, it undergoes a series of processing and adjudication actions. One of the case management systems used for tracking and adjudicating certain immigration request forms is the Electronic Information System (ELIS). ELIS is an internal case management system composed of microservices designed to assist with complex adjudicative and processing tasks. One of these microservices is the Sentiment Analysis machine learning (ML) model. This model classifies the notes taken by adjudicative officers as either positive, negative, or neutral sentiment for a particular case. Although this use case is not yet in production, it has been in development and testing for less than a year. The model was enhanced using ML to improve reusability and performance.  For more information, please visit: DHS/USCIS/PIA-056 USCIS Electronic Immigration System (USCIS ELIS). The sentiment analysis output is provided to officers to consider during subsequent case processing, helping to improve efficiency and streamline workflow. This could serve as a basis for deciding or providing risk assessments related to immigration, asylum, or detention status" and/or "emotion assessment. Survey results coded to specify the sentiments as positive negative or neutral tone in a excel dashboard. 

Deployment Status: Inactive (no longer used) 

Safety- and/or rights-impacting? Yes. Rights-impacting. This use case was identified as rights-impacting during pre-deployment but was not deployed and it is retired. 

Use Case Name: Predicted to Naturalize 

Use Case ID: DHS-60 

Use Case Summary: The Predicted to Naturalize model predicts when Legal Permanent Residents would be eligible to naturalize and attempts to provide a current address. This model could potentially be used to send correspondence to USCIS customers of their resident status and notify others of potential USCIS benefits. 

Deployment Status: Inactive (no longer used).  This use case was reported in a previous version of the DHS AI Use Case Inventory but was not deployed and it is retired. 

Use Case Name: I-485 Family Matching 

Use Case ID: DHS-61 

Use Case Summary: I-485 Family Matching is designed to create models to match family members to underlying I-485 petitions. The underlying immigrant petition defines if the I-485 is employment-based or family-based. It also has information about the visa classification and priority date which, when compared against the Department of State’s monthly Visa Bulletin, helps predict visa usage. It is difficult to match an I-485 to its underlying immigrant petition, because the only available field on which to match is the Alien-number (A-number). This number is not always present on the immigrant petition, and name/date of birth matching is not as reliable.  The goal of I-485 Family Matching is to leverage AI to create connections more confidently between petitioners and their families based on limited data.  Additionally, it was intended to help identify and group I485s filed by family members, as well as gather up the many ancillary forms they may have been pending (such as I765, I131). Similar to immigrant petition matching, it can be difficult to match up I485s filed by family members. In these cases, the only similar fields are a common address. Efforts have been made in the past to identify family members by address, but it is effective only to a point. This use case explored the technical feasibility of using an AI model to help make working with this data more reliable (as well as group individual petitioners, their families, and other helpful associated data together for faster and more accurate processing).  Work on this use case did not go beyond ideation and no AI model was built (and no output was rendered) after the proof-of-concept established that the desired end-state was infeasible with the technology available at that time. 

Deployment Status: Inactive (no longer used). This use case was reported in a previous version of the DHS AI Use Case Inventory but was not deployed and it is retired. 

Use Case Name: Topic Modeling on Request For Evidence (RFE) Data Sets 

Use Case ID: DHS-63 

Use Case Summary: Builds models that identify lists of topics and documents that are related to each topic. Topic modeling provides methods for automatically organizing, understanding, searching, and summarizing text data. It can help with the following: discovering hidden themes in the collection. classifying the documents into the discovered themes. 

Deployment Status: Inactive (no longer used) 

Use Case Name: I-539 approval prediction 

Use Case ID: DHS-64 

Use Case Summary: This project attempts to train and build a machine learning throughput analysis model to predict when an I-539 "Application to Extend or Change Nonimmigrant Status" case will be approved through eProcessing. Allows for some potential improvement for the approval process via eProcessing channel.

Deployment Status: Inactive (no longer used) 

Use Case Name: Biometrics Enrollment Tool (BET) Fingerprint Quality Score 

Use Case ID: DHS-182 

Use Case Summary: Duplicate with BET Maximization. The Biometrics Enrollment Tool (BET) team has incorporated the National Institute of Standards and Technology (NIST) Fingerprint Image Quality 2 (NFIQ2) algorithm (a trained machine learning algorithm) for scoring fingerprints (https://www.nist.gov/services-resources/software/nfiq-2) into the BET application. This algorithm takes a fingerprint image and assigns a score between 0 and 100, with 100 indicating the best quality fingerprint image that could be obtained. The higher the score, the more likely it is that the fingerprint will match when captured again. 

Deployment Status: Inactive (no longer used). This use case was reported in a previous version of the DHS AI Use Case Inventory but is a duplicate of Biometrics Enrollment Tool (BET) Fingerprint Quality Check (formerly Biometrics Enrollment Tool (BET) Fingerprint Maximization) (DHS-14).

Use Case Name: Testing Performance of ML Model using H2O 

Use Case ID: DHS-231 

Use Case Summary: USCIS is the component within DHS that oversees lawful immigration to the United States. That means USCIS receives, processes, and maintains all applications for admission for Lawful permanent residents (LPRs), or adjustments to LPR status. Also known as “green card” holders, LPRs are non-citizens who are lawfully authorized to live permanently within the United States and are required to fill out Form I-90, Application to Replace Permanent Resident Card (Green Card). Since there has been a considerable influx of green card applications, USCIS used a combination of exploratory data analysis to determine the most used categories for applicants submitting I-90's, and machine learning to create predictions of workloads. USCIS used the H20 machine learning model to allow USCIS analysts to build and run several machine learning models on big data in an enterprise environment and identify the model that performs the best. It has already been successful in identifying the most accurate model for the I-90 Form Timeseries Analysis and Forecasting use case. 

Deployment Status: Inactive (no longer used)