Half of Canadians say they can't tell the difference between real and AI-generated content: survey

(cottonbro studio / pexels.com)

Most Canadians believe they have encountered misleading AI-generated content on social media in the past six months—but half also don’t have confidence in their ability to discern between AI-generated fake news and human-produced content, according to a new study.

Released Monday by the Canadian Journalism Foundation (CJF), the study surveyed more than 1,500 Canadians on their knowledge of and concerns about AI-generated content on social media and online spaces.

There were clear generational differences in the responses, with younger Canadians more likely to state that they had noticed AI-generated content.

"When half the Canadian public is challenged on deciding what's real and not, the credibility of both journalism and media outlets has never mattered more," John Wright, executive vice president of Maru Public Opinion, said in a press release. The survey was conducted for CJR by Maru Public Opinion, a professional services firm which runs public opinion polls.

"Without a reliable anchor for truth in this incredible sea change, online deception will easily distort reality, wreaking confusion and skepticism everywhere."

• The information you need to know, sent directly to you: Download the CTV News App

Nearly three-quarters of Gen Z respondents stated that they had personally encountered misleading AI-generated content on social media in the past six months. This dropped to 63 per cent of millennial respondents, and 53 per cent of both Gen X and boomer respondents.

Those living in British Columbia, Manitoba, Saskatachewan and Ontario were the most likely to report having noticed this content in the past six months, with 61 to 63 per cent of respondents flagging this as something they’d experienced. Those with the highest level of education were also more likely to say they had noticed AI-generated fake news recently.

CANADIANS UNSURE IF THEY CAN PICK OUT AI-GENERATED FAKE NEWS

A total of 48 per cent of Canadians surveyed said they weren’t confident in their ability to discern AI-generated content from human-generated content.

The generational breakdown showed that older Canadians were more likely to admit this, with 54 per cent of boomer respondents selecting this option compared to 37 per cent of Gen Z respondents.

There was a clear gender gap, with more women admitting they doubted their perception, at 54 per cent compared to 42 per cent of men.

The proliferation of AI-generated fake news on social media, where many source their news, has raised flags for journalism groups and those concerned about misinformation.

"These findings emphasize the crucial need for promoting news literacy across all segments of Canadian society," Natalie Turvey, CJF president and executive director, said in the release. "The CJF is committed to serving as a driving force for advancing news literacy skills and critical thinking to enhance public understanding of the media and also to nurture trust between Canadians and their news sources."

Deepfakes—video and photo content which uses the likeness of another person, usually a celebrity or media personality, to trick the viewer—have been on the rise, with experts raising alarm bells about how easily they can be manipulated to spread fake news.

"Canadians are beginning to identify and understand the risks of AI-generated misinformation. With these perilous new threats to the integrity of information, it is imperative that Canadians of all ages comprehend the importance of news and information from credible sources,” Kathy English, CJF board member, said in the release.

Read more: A warning from experts: 'The deepfakes you see now are going to be the worst you're ever going to see'

CAN IT BE REGULATED? CANADIANS DOUBTFUL, BUT ATTEMPTS UNDERWAY

The majority of respondents, at 71 per cent, believe that governments will not be able to regulate AI, the study found.

This was stronger among older Canadians, with 79 per cent of boomers expressing this scepticism compared to 62 per cent of Gen Z—but unlike the other sections of the survey, answers didn’t fall along age lines completely. Millennials were the second most sceptical of the ability of government regulations, at 69 per cent, with Gen X coming third at 67 per cent.

Only 29 per cent of Canadians overall believe governments will be able to effectively regulate AI.

Although it’s a new frontier, there are already attempts underway to tackle regulating AI.

This summer, the European Union announced that AI in EU member states would be regulated by the AI Act, in what it says is the world’s first set of regulations to establish boundaries around the creation and use of AI.

The legislation is still under negotiation, but currently includes generative AI systems such as ChatGPT having to disclose that content is AI-generated, the creation of a database that AI models would have to be registered in before being released onto the market, exemptions for specific research based AI components, as well as a ban on AI being used for biometric surveillance, emotion recognition and predictive policing.

Currently there is no regulatory framework that specifically tackles AI in Canada, but there is proposed legislation that was drawn up as part of Bill C-27, the Digital Charter Implementation Act. Called the Artificial Intelligence and Data Act (AIDA), this framework aims to set up new responsibility requirements for AI systems, based on a risk assessment for which types of AI systems might cause serious harm to Canadians.

Bill C-27 is currently in the House of Commons, but once it receives Royal Assent, the government will open up a consultation period to clarify AIDA.

AIDA is expected to come into force no sooner than 2025.

AI DEVELOPMENT

Despite the wariness that Canadians expressed around the issue of AI-generated content in this new study, a fifth of those surveyed stated that they were personally experimenting with generating their own AI content for work, school or personal use, with a quarter of respondents stating that they knew someone who was doing this.

The two groups most likely to be experimenting with AI were those with the highest amount of education (university and beyond) as well as those with the lowest amount of education (high school or less).

In Canada, Atlantic Canada and Ontario had the highest percentage of respondents who said they were experimenting with developing their own, at 27 and 26 per cent respectively.

METHODOLOGY

Findings are taken from a Maru Public Opinion survey run October 11-12, 2022, among a random selection of 1,525 Canadian adults who are Maru Voice Canada panelists. The results were weighted by education, age, gender, and region (and in Quebec, language) to match the population, according to Census data.

For comparison purposes, a probability sample of this size has an estimated margin of error (which measures sampling variability) of +/- 2.5%, 19 times out of 20. Respondents could respond in either English or French. Discrepancies in or between totals when compared to the data tables are due to rounding.