Minister Fraser clarifies how IRCC uses AI in application processing
At a recent press conference in Vancouver, Canada’s Immigration Minister Sean Fraser told reporters that “by increasing our use of technology, advanced analytics and streamlining our processing we’ve done a couple of very important things in recent months."
Fraser was speaking about temporary residency (TRV) applications for those waiting for permanent residency under family class sponsorship programs. However, the increased use of technology and advanced analytics has had an impact across all lines of business in application processing by Immigration Refugees and Citizenship Canada (IRCC).
In the coming years, Canada is on track to welcome the highest numbers of newcomers ever. In 2022, IRCC made more than five million final decisions on applications across all lines of business, double the number of decisions in 2021.
IRCC has rapidly shifted toward the digitization and “modernization” of Canada’s immigration system, which includes increased use of Artificial Intelligence (AI) and advanced data analytics to speed up application processing time. For example, Fraser said that this streamlining has contributed to processing 98% of spousal TRV applications and new applicants can now expect a processing time of just 30 days.
IRCC says digitizing the system and services will better meet the needs of clients and Canada, bolstered by new technological capabilities. Further, the department says that by leveraging technology, it can process applications more efficiently and free officers to focus more on complex applications.
IRCC uses artificial intelligence (information technology that performs tasks that would ordinarily require a human to accomplish) in some aspects of processing including:
- Automating positive eligibility determinations
- Distributing applications between officers based on the characteristics of the application
- Identifying applications that may require additional verification
- Workload distribution
- Creating “annotations” that summarize basic information on each client to reduce officer searches in our Global Case Management System
- Triaging client emails to enable faster replies, and responding to client enquiries by providing publicly available information
- Assessing biometrics
There are no algorithms in any tools IRCC uses that will accept or reject an applicant. Candidates are not approved or denied a visa or permit based solely on a computer-generated decision.
What are the concerns?
Many worry that the use of AI by IRCC will lead to bias and a general lack of explanation for how decisions are made in processing applications.
For example, the Treasury Board Directive on Automated Decision-Making , a government-wide policy directive, says that basing an algorithm on historic data can amplify race, class, gender, and other inequalities. It cites that some facial recognition software does not work equally well for all skin colours or genders.
Depending too heavily on AI can also lead to a lack of clarity in decision making. The directive says the federal government must be able to explain how administrative decisions are made. Further, individuals denied services or benefits have a right to a reasonable and understandable explanation from the government, which should go beyond indicating that it was a decision made by a computer.
Concerns have also been raised around the use of “Chinook,” which IRCC describes as a Microsoft Excel-based tool to simplify the visual representation of a client’s information. It is used by IRCC officers to assess temporary resident visas, study permits and work permit applications. The department says Chinook does not utilize artificial intelligence (AI), or advanced analytics for decision-making, and there are no built-in decision-making algorithms.
How does IRCC develop algorithms?
In response to concerns, IRCC says it follows a Directive in the Canadian Charter of Human Rights and Freedoms to ensure equality rights and preclude discrimination. It says it follows principles of transparency, accountability, legality, and procedural fairness to define how decisions need to be made and what explanations must be provided to those impacted.
Whenever a new algorithm is proposed, it must pass the Algorithmic Impact Assessment (AIA). The AIA is a mandatory risk assessment tool and part of the Treasury Board’s Directive on Automated Decision-Making. The tool measures risk areas, mitigation, and the impact of the proposed algorithms. IRCC says they were one of the first government departments to use AIAs.
The department also says it requires assessments of the impacts of algorithms, quality assurance measures for the data and the algorithm, and proactive disclosures about how and where algorithms are being used.
IRCC says the rules used to support (but not finalize) the decision-making process are regularly reviewed by experienced officers, legal, policy, data science, privacy experts, and senior decision-makers to ensure they align with the eligibility criteria outlined in the Immigration and Refugee Protection Act.
Also, regular monitoring and quality assurance measures are in place to help ensure that these models continue to perform as intended and that any unforeseen negative impacts can be identified early and mitigated.
Minister Fraser remains optimistic about the use of technology and advanced analytics by IRCC. He said that technology adopted by IRCC in the past few years allows the department to look across a number of factors on a person's application to help determine the likelihood that they are going to be eligible for permanent residence and put them in a category that is more easily approved by IRCC officers. He says this has led to a huge boost in productivity and continues to stress that AI does not make any final decisions.
“At the end of the day our offers still make every determination for eligibility. It not possible for anyone to be denied through this technological solution.”