Feds have used AI in hundreds of initiatives, a new research database shows

Content of the article

OTTAWA — Canada's federal government has used artificial intelligence in nearly 300 projects and initiatives, new research has found, including helping to predict the outcome of tax cases, sorting temporary visa applications and encouraging diversity in hiring.

Content of the article

Joanna Redden, associate professor at Western University, compiled the database using news reports, documents tabled in Parliament and access to information requests.

Content of the article

Of the 303 automated tools in Wednesday's registry, 95 percent were used by federal government agencies.

“There needs to be more public debate about what systems should be used and more information about how those systems are being used,” Redden said in an interview.

It exposes the data problem with the Liberal government's proposed Artificial Intelligence and Data Act, the first federal bill to focus on AI.

“This piece of legislation generally does not apply to government use of AI. So the sheer number of applications we've identified shows what the problem is.”

Content of the article

Bill C-27 introduces new obligations for “high-impact” systems such as the use of AI in employment. That's what the Department of National Defense experimented with when it used AI to reduce bias in hiring decisions in a program that ended in March 2021.

The department used one platform to short-list candidates for interviews and another to assess an individual's “personality, cognitive ability and social ability” and match them to profiles, the spokesperson said. Candidates gave clear consent, and data informed human decision-making.

Immigration, Refugees and Citizenship Canada said two pilot projects to help officers process temporary resident visa applications have been made permanent since 2018. The department uses “artificial intelligence tools to sort applications and determine positive eligibility.”

Content of the article

Content of the article

The register said the department would use AI to review study permit applications from people from other countries, although a spokeswoman said AI would not be used to “make the final decision”.

The agency's automated systems cannot recommend denying or denying an application, the spokeswoman said.

Not all experiments become sustainable initiatives.

The Public Health Agency of Canada said it has suspended a project that analyzes publicly available social media information to look for warning signs of suicide due to factors including cost and “methodology.”

Health Canada, on the other hand, continues to use a social listening tool with a “simple AI component” to scour online news stories for consumer product recalls, a spokesperson said.

Content of the article

Some of the experiments would be familiar to Canadians — for example, the Royal Canadian Navy tried a system similar to Apple's Siri or Amazon's Alexa to send verbal commands to ships.

Efforts to integrate voice-activated technology into warships are ongoing, the spokesman said, but “information security concerns must be considered before deploying such technology.”

AI also works for legal research and predictions.

The Canada Revenue Agency says it uses a system that allows users to input case-specific variables that “produces expected results by using analytics to predict how a court would rule in a specific scenario based on relevance and historical court decisions. »

Content of the article

And the Canadian Institutes of Health uses software to make decisions about employment relations. It compares the actual situation with previous situations and models how different facts might affect the outcome, register contours.

In the Office of the Chief Bankruptcy Officer, AI identifies anomalies in real estate filings.

The system identifies “potential debtor ineligibility based on key attributes found in insolvency filings,” a spokesperson said. The conditions set by the system are evaluated by analysts.

The registry also includes examples of AI used by the RCMP. A spokesperson confirmed that the RCMP used AI to identify child sexual abuse material and help rescue victims.

The legally obtained internal data used “a type of facial recognition technology known as face matching,” the spokeswoman said.

Content of the article

Facial recognition is also used by Canada Border Services Agency. The agency uses the technology on a voluntary basis through kiosks at some airports to “help verify the identity of inbound travelers,” a spokeswoman said.

Redden said there are many reasons to question facial recognition, including examples that have led to illegal arrests in the US.

More broadly, he argued, the government needs to better oversee its use of AI.

The federal government has said that its Automated Decision-Making Directive requires algorithmic impact assessments in cases where the use of AI is “likely to have a significant impact,” such as when assisting administrative decision-making.

Content of the article

These estimates are then published in the public register, the Treasury Board said in an email.

There are currently only 18 entries in the registry.

When asked why that number was so much less than Redden's total, a spokeswoman said the directive and registry “focus on the use of AI that directly affects individuals or businesses.” Many AI applications in the federal government do not fall into this category.”

One such example: technology used to observe nature.

The Canadian Food Inspection Agency uses machine learning to track invasive plants, insects and shellfish, according to the registry.

The agency uses an AI tool to scan social media for plant and animal control, the spokesman said. Fisheries and Oceans Canada says it will use AI to “identify marine mammals from aerial, drone and satellite imagery.”

Content of the article

It took Redden two years to compile the data based on limited information from various sources.

The available information often doesn't show when or why the AI ​​system was introduced, whether it still exists, what data is being used, or whether there have been any problems with the system, he said.

“It's very difficult for outsiders to do this kind of work.”

It is unclear what happened to some of the pilot projects documented by Redden.

A document tabled in Parliament in January 2023 shows that the Canada Border Services Agency is developing an algorithm to automatically detect gun and gun parts in postal X-rays, while Global Affairs Canada is experimenting with AI-generated briefing notes.

Global Affairs did not respond to a request for more information, and the CBSA declined to provide an update on those actions.

“While we can say that the CBSA is currently closely monitoring the development of machine learning algorithms for X-rays to automatically identify items of interest, we are not disclosing specific targeting, enforcement or intelligence information as this may render them ineffective,” the agency said.

As the Register points out, Redden discusses “how pervasive AI is in government across Canada” and how little we know about its use.

Content of the article