Navigating the AI Hype: Bias, Misinformation, and Objectivity in the Media

Photo by KOMMERS on Unsplash

You wake up in the morning and ask your AI-assisted clock to stop and initiate your morning routine, as you brush your teeth you check how your sleep was on your smart watch. Now it’s time for breakfast and you know exactly what you’re going to have because a clever app already built your meal plan for the week. Moments later, a GPS guides you to your place of work, and if you can afford high-tech self-driving cars there’s no need to worry about that either. We are already at a point where technology, particularly, artificial intelligence (AI) already has a huge role to play in our daily lives.

“I think we kind of have blinkers on if we don’t admit that we don’t know all of what generative AI is going to impact,” says Jessica Wolfe, professor in the Digital Communications program at Humber College in Toronto.

Unless you were not keeping up with the news, or public conversation around technology, chances are you have heard of artificial intelligence, or AI, particularly now, its generative capabilities. From computational science to education, AI also influences the media as journalists rush to adapt to possible and current changes these technologies bring.  

When GPT-3 was launched to the public in November of 2022, the date marked when the public started to openly recognize the importance of this technology. Until then, OpenAI had been working on the previous versions of ChatGPT, with its first version being trained on 117 million parameters, which can be understood as a set of settings that influence, and guide, how these models respond. Now, ChatGPT-3, its most recent version, has the upgraded number of parameters equal to 175 billion. The impact of this has been seen widespread across various areas in society, with the technology being used to answer factual questions, create full pieces of text, and even translate between languages.

“We must know what these things can do before we try to constrain them. We’ve got to understand what’s at stake,” says Wolfe, who is also a creative writer and journalist. She explains how she likes to encourage an experimental attitude from her post-secondary students with generative-AI technologies, while always emphasizing the importance of transparency. “Because if we limit ourselves to not using them, everyone else, all the competition is gonna go off and get good at using them. And I'm not setting my students up well for success,” the professor adds.  

In the realm of text generation, AI-powered technologies such as Microsoft’s Copilot, which works in a similar way to ChatGPT, as well as Google’s Gemini have also been under the close watch of those who are keeping up with these developments. With the rise of these technologies, so came multiple concerns about their possibilities and how they could impact our lives. A recent poll by Leger surveyed Canadians to find that 63 per cent, an 11-point increase from last year, believe that AI tools could become so smart they could outthink humans. While 80 percent of them also believe this will have an impact on their jobs. These concerns have been widely intensified by AI’s intense period of media coverage. Multiple surveys conducted by both stakeholders and academics have shown that in North America the public has a broad, but shallow understanding of these technologies.

“It's important to focus on your own voice and cultivating your skills, even though this tool exists, because it's your ability to express yourself, it's going to set you apart no matter what field you're in,” says Wolfe.

Justified by the possibility of AI changing jobs, or the aspect of its development being highly motivated by economic interests, many surveys have focused on people’s perceptions towards AI in relation to their workplace. Despite being involved in other aspects of the situation; members of the media are no different when it comes to having their practices impacted by the very technology they are tasked to understand and explain to the public.

“Gosh, the bar gets higher and higher in terms of the basic skills of journalism,” says Brodie Fenlon, Editor-in-Chief of CBC News when asked about the role of a journalist in this landscape, “Because we are going to get tricked easily this stuff is going to be so good, it is really going to be hard to tell what is real and what is not. Fact checking, verification, double and triple sourcing… so our job gets more difficult.”

Multiple newsrooms are already actively integrating AI into different aspects of news reporting. Forbes utilizes an AI-powered content management system for suggesting headlines. Similarly, The Washington Post employs a reporting bot capable of analyzing data patterns in areas such as financial trends and election results. The Associated Press leverages data to prompt an AI to generate preview stories for each NBA game, particularly seen as beneficial given these stories’ repetitive nature.

“We've put out some pretty clear guidelines on how they [journalists] could do that and what they'd have to consider,” says Fenlon about using generative text AI in CBC News’ newsroom, “But we've actually been very, very slow and cautious about these new tools.”

Fenlon also refers to an editor’s note published by him in June of 2023, where he attempts to tackle some concerns and questions readers may have about the technology and its widespread use in the media. In the piece, Fenlon reinforces ethical considerations and the need to stay true to their journalistic standards and practices regardless of the changes brought upon the media by AI. 

The rise of AI technology has left many journalists wondering how this will impact their jobs in a time where the journalism industry is still struggling to land on its feet after getting dragged through the dramatic digital revolution sparked by the Internet and its adoption at a global level. Issues like growing costs, ever-decreasing wages, and the constant testing of new business models that could withstand the multiple changes brought by the quick-paced movement of world-wide technological development.

“I mean, journalism is more than just writing and producing content,” says the Editor-in-Chief of CBC News, “It's being physically on the ground, present, to find stories and reports. So that takes a physical presence and being everywhere in a way that, you know, an AI can't do at this point.”

Misinformation and disinformation: Critical thinking now more important than ever

For a while now journalists have been tasked with fact-checking information being spread online, in addition to emphasizing this aspect in their own creation process.

“Certainly, I think one of the big risks of generative AI is the increase in disinformation, misinformation,” warns Professor Wolfe.

The use of computer algorithms to process data and make decisions has been widely discussed years before the broader scope of AI systems came into play. Content coming from social media and search engines are the best example of the use of algorithms in personalizing content for users and it has been well-known for playing a huge role in the disinformation phenomenon we’re currently facing.

“Journalism is just one part of this larger phenomenon, which means that all life in [modern] society is happening within digital platforms,” says Giuliander Carpes, “which are private digital infrastructures.”

Carpes is an executive editor of Farol Journalism, a Brazilian independent newsletter with the purpose of discussing the latest trends in the journalism industry. Carpes has recently completed a PhD with a research focus on what he calls the ‘platformization’ of journalism.   

Carpes explains this change in journalism’s business model has its beginning marked by the turn of the century when the giant tech company Google began targeting the news sector with its search engine and news aggregator.

“These forms of distribution are generally done through the algorithms of these platforms. Each algorithm takes into account several factors, many factors that we don't even know very well, but factors that are defined by the platforms themselves,” says Carpes, “Thus, they also end up influencing the production of content.”

Giuliander Carpes on interview via Zoom: Screen Capture. March 18, 2024. Credit: Thais Grandisoli

Despite being only a small share of the much larger scope AI systems encompass, the use of algorithms is a fundamental aspect of it since they are used to process data and make decisions. 

Recent court investigations have explored the possibility of Facebook’s content personalization and lack of appropriate moderation having a significant impact on the most recent presidential elections in the United States. It is just one example of many, as policy makers take efforts to make companies such as Facebook and Instagram owner, Meta, assume responsibility for the transformational power of these mediums.

“ChatGPT is not going to combust into consciousness and decide to produce disinformation. That's absurd,” says Emily Bender, computational linguist, and professor at Washington University, “On the other hand, people can absolutely use these synthetic media extruding machines to produce misinformation/disinformation, or even just noise that makes it harder to find real information.”

It’s not just about generating false information; it’s about helping create a landscape where no one even knows what’s real anymore. On that note, Bender emphasizes the need to remember that machines are controlled by humans and point to the real actors behind the technology.

“But that doesn't happen on its own,” says Bender, “It happens because somebody decides to take this technology and use it for that purpose, […] And it happens because the companies producing the systems have not taken even the minimal steps that could be done to make it easy for other people to just filter out the synthetic media.”

Bender argues against labeling these chatbots as “intelligent” in the traditional sense. While they are able to process information similar to human cognition, they are fundamentally different in how they operate and lack a conscience or genuine understanding of what it is saying. In a paper published in collaboration with other authors such as computer scientist Timnit Gebru, she compares them to parrots, and how these animals can reproduce human sound, but can’t quite understand the meaning of the words they reproduce with their vocal abilities.

Synthetic media encapsulates every content that is generated and/or curated by AI. This form of media can include images, videos, text, and it can often look indistinguishable from authentic content. Due to the difficulties in this content detection, this has been a highly explored area in research and those looking to understand more about the impacts this can have on multiple areas of society. 

In 2019, a group consisting of media and tech companies, including the Canadian Press, Getty Images, Microsoft, The New York Times, BBC, and CBC/Radio-Canada have gathered to kickstart an initiative which proposes the authentication of sources in media. In other words, Content Authenticity Initiative (CAI), as it was called, is purposed with developing and employing a technology that would embed an electronic seal in media content so the public would have another layer for the authentication of reputable sources. A few years later and now it is the Coalition for Content Provenance and Authenticity (C2PA) which draws from CAI efforts to implement the content credentials technology.

“It's now at the point where they're looking at how do we implement it?” says CBC’s Editor-in-Chief Brodie Fenlon, “This is a really interesting piece of the AI puzzle, because everyone kind of agrees: it's really hard to disprove a fake and the technology is always going to get more and more advanced. So can we instead focus our efforts in helping people understand what is real and what has been changed?”  

Many efforts to identify and label synthetic media have arisen and evolved over the years as a response to this conundrum. Content credentials and detection technologies.

So what makes a synthetic media content different from human-created content?

A recent poll conducted by Maru Public Opinion on behalf of the Canadian Journalism Foundation (CJF) found that almost half (48 percent) of Canadians reported not being confident in their ability to tell AI-generated apart from human-generated content. 

Professor Bender believes there  is sort of a characteristic impression that you get from synthetic media, which is reflective of the process that was used to produce it. However, it is intrinsically different from the process of human creativity and cognition. 

“I would say it's the opposite of having, you know, the emotions and curiosity and ideas that underlie human artistic production,” says Bender.

Art and journalism are different in many ways, but both share the same principle of communicating something about the human experience to other humans. So experts point to the fact that AI such as ChatGPT, for example, will never have the human experience necessary to create these works in an authentic way. 

“So yeah, absolutely. ChatGPT can output something that looks like a journalistic writing sample on a specific topic,” says Bender, “But is that a journalist’s job?” 

When it comes to writing factual pieces, there’s plenty of work involved in the process of newsgathering even before it hits publication stages. Journalists need to use their critical thinking and practical skills to look for sources and content that will tell the story in the most transparent and informative way possible. 

“I mean, journalism is more than just writing and producing content.” says Brodie Fenlon, “ It's actually being physically on the ground, present, to find stories and reports. So, that takes a physical presence and being everywhere in a way that, you know, an AI can't do at this point.” 

Bias and Explainability of AI

While some hope that AI will help tackle the issue of human bias, others are worried it will only make it worse. 

“They're going to be designed based on someone's understanding of the world that's being modeled on the algorithm,” says Bender. 

The linguistics professor again warns about the propagation power these tools can have. 

“Instead of saying a biased thing face to face to somebody, you've now made it possible to repeat that over and over and over again. So I don't think automation is any escape from bias,” says Professor Bender.

When analyzing the situation through this perspective, we could reach the conclusion that AI technologies can be understood not as a neutral tool, but as a result of a mutual co-shaping between technologies, societies, politics, and cultures in which they are embedded.

For centuries people have associated machines with having a more impartial aspect as opposed to people and our subjective cognitive thinking. Throughout the recent years, the rise of AI has turned bias into a key issue in public discussions surrounding the technology.

As a chair in the United Nations Educational, Scientific and Cultural Organization (UNESCO), Professor Rosa Maria Vicari, explains that one of the ways they’re trying to make AI systems applied to education act in accordance with their values is by exploring the idea of ethical alignment. 

“The whole idea is whether you can make A.I. make decisions based on ethical principles,” says Vicari, “And that can be applied to every field using A.I.” 

AI applications are being used in various important areas of society, it goes beyond just automation of our daily routines with our smart appliances. AI is also used in larger scopes when taking up decision-making roles from picking out the best fit from a list of job applicants to whether you are eligible for a bank loan or not. 

Despite all concerns surrounding AI tech development being broadcasted by the media, there is still a clear distinction between those who believe AI is powerful in a negative way, and those who are faithfully optimistic about the technology. 

Professor Emily Bender explains that mixing up science fiction and real life when addressing risks of AI diverges attention away from present and possible future factual-based risks. 

“AI tech has always been based on bringing knowledge to machines, and then what happens? When you pass on knowledge to someone [or something], you delegate power. And AI is built to be proactive, so it makes a decision,” says Rosa Vicari. 

Professor Vicari warns that even more complications arise when we don’t have enough information. Consumers, the general public and even stakeholders still don’t know much about how these tools work. The complex nature of these technologies and the issue of propriety laws protecting the owning companies' right to not disclose their process when developing these systems, among others, create a landscape which many are calling a black box phenomenon. 

Explainable AI, an approach born in opposition to that, has the purpose of doing things differently. Supporters of Explainable AI are primarily concerned with quality and accuracy checks of AI-generated content and the integration of models which are transparent and understandable to users. 

“ It’s an urgent problem. There needs to be more research and more responsibility taken by all parties,” says Vicari.

However, deciding that we can’t do anything in the face of a complex and opaque issue would be to surrender our power as actively-thinking beings

“This is why I have always advocated for education as a way of exercising one’s autonomy,” says Vicari.

Despite generative AI’s popularity and its black box issue, professionals from all over the world are attempting to bridge that gap by promoting discussions within their fields.

Giuliander Carpes executive director and writer for Farol Journalism says last year he asked his team to analyze the content of their own publication by ranking topics in order of popularity and prevalence on their coverage. What he found was that in 2023 alone the journalism publication had over 70 per cent of their articles mention or focus on the topic of AI. That’s when he had the idea of publishing a guide to AI in newsrooms, which was initially aimed at their own audience but ended up being made public. 

If AI algorithms aren’t biased, but humans are passing on our biases to the programming of these systems, how can a journalist be more objective than an AI? 

The answer isn’t as simple as the question seems to imply, nevertheless, it has pondered over the minds of many keeping up with the news of how fast advancing these technologies can be.

Like CBC News’ editor-in-chief Brodie Fenlon, the executive-editor of Farol Journalism Giuliander Carpes, also believes the job of a journalist only becomes more important in this landscape filled with hype and uncertainty about how these technologies will fully impact us.

“AI could help us finish a few tasks quicker during our process, but that doesn't solve everything, right? You need to continue to be rigorous in the matter of fact-checking. You need to be connected with the question of the veracity of the things you say,” says Carpes.

When faced with the ethical dilemmas associated with the mixing of AI and journalism practice, Carpes says journalistic principles such as transparency and a duty to the public become even more important.

“Traditional journalism is concerned with the truth, and that is kind of just a concept, right? There is no absolute truth,” says Carpes, “But a person, the reporter, or editor, had to believe this truth, this was their voice. We’re responsible for making sure it matches the facts.” 

When describing journalistic objectivity in schools, many would jump to the definition that describes a view from nowhere: you are writing to everyone from no one’s point of view, in particular. 

“I think that the push for this idea of there being a single objective view from nowhere that everybody has to admit is the most objective one is basically just a manifestation of white supremacy,” says Emily Bender.

Journalist Candis Callison and professor of journalism at the University of British Columbia also talks about the complexities behind the traditional notion of objectivity in her book “Reckoning: Journalism's Limits and Possibilities.” According to Callison, journalism has long perpetuated dominant power structures in society which guide us to decide what is considered neutral and what isn’t. She advocates for transparency and that includes embracing the fact that every journalist is an individual in society and has a view from somewhere. A similar point is made by journalist Priska Neely in an interview to Poynter, when she outlines important areas on where newsrooms should improve in their quest for more diversity. Neely argues that to do that successfully, they would need to not only hire with more diversity of employees and experiences in mind but also make sure they have agency to be who they are. Neely says this is true diversity: when the complexity of humans and the goals of individuals are respected and considered after the hiring process as well.

“And so to work with integrity is to not acquiesce to that, and to own one's own point of view,” says Bender “And say, I am bringing my ideas and my perspective into it, here's what they are.” 

Being transparent about methodology and criteria used when selecting sources and information to build a story is a way of building trust with audiences and working with integrity in a landscape filled with opacity. To Professor Bender, insisting that it is possible for a human to be completely objective or impartial in that way is not a path towards integrity or transparency in the media, rather only one more reason as to why it becomes so difficult to be able to distinguish between all the information we consume.

The future of the practice seems to rely on adaptation to the inevitability of new technologies while still upholding ethical standards and emphasizing the need for transparency in reporting. While seemingly paradoxical, the protection of journalistic work’s authenticity may depend on an increased support, acceptance and the highlighting of essential human qualities in the profession such as critical thinking, creativity and general life experience.