Â鶹ÒùÔº

Have journalists skipped the ethics conversation when it comes to using AI?

Artificial intelligence (AI) is being used in journalistic work for everything from and to writing and publishing , and stories.

It's even being used to in cases where time-strapped reporters don't have time to do so.

What's lagging behind all this experimentation are the important conversations about the ethics of using these tools. This disconnect was evident when we interviewed journalists in a mix of newsrooms across Canada from July 2022 to July 2023, and it remains a problem today.

We conducted semi-structured interviews with 13 journalists from 11 Canadian newsrooms. Many of the people we spoke to told us that they had worked at multiple media organizations throughout their careers.

The key findings from :

• AI literacy varies within the same newsroom and certainly within the industry as a whole.
• There's agreement that humans play an important role in supervising the use of AI, but there's no agreement on where human journalists must be involved in the process—at the AI tool coding level? Before a piece is published?
• Journalists believe professional practice and industry standards are being followed when using AI in journalism, but there is no agreed-upon "rule book" for how AI should be used.
• There are issues with transparency about how and when AI is being used, both among journalists working in the same newsroom and in terms of what is revealed to audiences about whether the content they are consuming was created using AI tools.

What journalists told us

Some of what we heard was reassuring. One journalist told us: "The one thing that we are very particular about when we use this technology is that our editors always have the ability to override what the machine is doing."

At the same time, however, it became clear that many news organizations are still operating in the ethical equivalent of the Wild West.

In many cases, journalists we spoke to talked about just following their gut when it came to deciding if using that AI tool to do that task was ethical. As one of our interviewees put it: "There's a rule book in my head."

When we asked interviewees how they knew their colleagues at the same publication followed the same ethical code they did when using AI, most could not answer except to imply that their co-workers wouldn't have been hired if they didn't share the same principles. One journalist said, "I've worked there for 14 years now …I can't think of anyone whose ethics I would disagree with."

Getting the ethics of AI right and being seen to be doing so is important because journalism has and needs to do everything possible to reverse the trend.

Multiple studies have shown that Canadian audiences want to know if AI tools are being used in newsrooms, and for journalism created using AI.

AI and news

Audiences, meanwhile, are being fed a steady diet of examples that illustrate how using AI tools to create journalistic work can go very wrong. For instance:

• was forced to disavow its AI audio tool because it was mispronouncing the Manitoba premier's name.

• was accused of "softening the image of the Ku Klux Klan."

• provoked outrage when it quizzed readers on how a woman who was featured in the article had died. The poll was created by a Microsoft news aggregator, but The Guardian stated that it damaged their reputation.

• was caught creating fake bylines for AI-generated stories on their websites.

News organizations might think they're being transparent with audiences about how much content is being created using AI, but our research finds the evidence is mixed at best, especially in circumstances where AI generates the content and an editor approves it in the content management system before it is published.

In one memorable Zoom interview, an editor walked us through the AI-generated content in an article posted online, saying that it was clearly identified as AI on the webpage.

However, upon sharing the page, they were shocked to discover there was no information about the article being AI-generated anywhere. They said it would be fixed immediately, but when we last checked, the article still said nothing about the AI tool used to generate it.

While we gathered data from interviews, newsrooms in Canada started releasing guidance through internal emails and . It is hard to find any language in publicly accessible policies that refers explicitly to how AI is being used or the ethics surrounding such use. It's also unclear who is involved in conversations about ethical AI use in newsrooms, and who is not.

As one journalist we interviewed put it: "I think my frustration personally comes from again the lack of openness to have this conversation about AI, and the urgency of it, because I think … we're so busy trying to survive, we don't realize that having this conversation about AI will help us survive."

Our research suggests journalists and news organizations are still struggling in the midst of rapid technological change to arrive at a shared understanding of AI tools, their usage, the limitations of programming and best practices that build rather than erode trust.