CIMM – Advanced Data Analytics to Sort and Help Process Temporary Resident Visa Applications – May 12, 2022

Key Messages

• Immigration, Refugees and Citizenship Canada (IRCC) is using advanced data analytics to help IRCC officers sort and process temporary resident visa (TRV) applications.
• Advanced data analytics helps simplify IRCC’s work and shorten processing times by identifying non-complex applications for streamlined processing, such as those from certain clients who have been previously approved to visit Canada in the past 10 years. It also helps to sort and triage non-routine applications to officers based on their level of complexity.
• The system of advanced data analytics assumes a significant portion of clerical or repetitive tasks related to sorting applications, leaving officers with more time to focus on assessing applications and making decisions.
• The use of advanced data analytics is part of IRCC’s commitment to improving the experience of its clients, and is helping to assist in managing our increasing volume of TRV applications.
• Officers always make the final decision on all temporary resident visa applications. The system never refuses or recommends refusing applications.

Supplementary Messages

Use of advanced data analytics for TRV applications

• At IRCC, advanced data analytics has been used since 2018 to help sort and triage, and make some positive eligibility determinations on TRV applications from China and India due to their high volume of applications. While the system does not make the final decision on any application, it has been shown that non-complex files can be assessed for eligibility 87% faster using the system, resulting in some applicants receiving decisions more quickly.
• When the advanced data analytics system finds a non-complex application, it can positively determine that an applicant is eligible. The application then goes to an officer to review for admissibility, which includes security and criminality checks.
• Rules used by the system to positively determine eligibility go through an extensive review process for potential discriminatory impacts. Rules are vetted by subject matter experts to determine that they reflect the legislative and regulatory requirements for the program, and reflect officer best practices. The system is also subject to a host of due diligence activities, including legal review, privacy impact analysis and Gender-Based Analysis Plus, in order to identify and mitigate risks related to bias, procedural fairness, privacy and accountability.
• A certain percentage of applications selected by the system for positive eligibility determination are redirected for full human review to provide a feedback loop that ensures that the system is fair and functioning as intended.

Background

• IRCC is working to responsibly develop and deploy data-driven technologies, in line with all privacy requirements.
• As part of our compliance with the Treasury Board Directive on Automated Decision-Making, an algorithmic impact assessment has been completed to assess the analytical models used for TRV applications. The assessment is available on the Open Government website, as required by the Treasury Board Secretariat.
• This is the second algorithmic impact assessment that the Department has published. The first was posted in summer 2021 for the Spouse or Common-Law Partner in Canada advanced analytics pilot.
• Some previous media and academic articles have incorrectly reported on how IRCC is using advanced analytics, including claims that the technology is replacing officers and automating decisions for humanitarian and compassionate as well as pre-removal risk assessment applications. Some of this attention has centered on IRCC’s use of the Chinook tool. Chinook is an Excel-based tool that streamlines the administrative steps that would otherwise be required to process applications. No advanced data analytics or artificial intelligence technology was used to build Chinook.
• IRCC has spent about $15 million spread over five years (2017-2018 to 2021-2022) for activities related to advanced data analytics. These expenses consisted primarily of salaries, hardware and software costs in IRCC’s data warehouse to maintain databases used for analytics. Some of the expenses consisted of salaries of data scientists who build the models. Only a very small portion of the expenses pertained to software used to build models, and these expenses will decline in the coming years as IRCC shifts to open source software.