From Paperboys to Algorithms

How Metrics Shape Modern Journalism

Before the search engine, we had the news bundle: a cluster of 40 to 50 pages of content thrown at your screen door every morning by some teenage kid on a Huffy bike. Now we have echo chambers and partisan news outlets, and the newspaper comes in bits and pieces on our phones tailored to our interests.

Forget the newspaper boy, forget the newspaper, period.

At this point, we should all be fully aware of the echo chambers we voluntarily place ourselves into (me included) on a daily basis. These can include, but are not limited to, a set of podcasts, publications, and cable news channels, a group of like-minded friends and neighbors that nod their heads to our quick political and cultural takes, and the highly sophisticated, algorithmically structured platforms like TikTok or Twitter that help reinforce a set of beliefs we gravitate towards. These platforms build a more dialectical relationship with us through reading our viewing habits, providing audiences with content they tend to lean towards.

These conditions are not lost on modern-day journalism, where audience engagement metrics (data points that show how your audience responds to and interacts with your content) gently nudge newspapers to publish certain types of content on social media platforms. Some have argued that these metrics have stolen a seat in editorial boardrooms, propelling every journalist in America to adhere to what audiences desire rather than focusing on objective journalistic standards.

But how much do audience metrics impact journalists’ choices and values? Particularly when it comes to AI representation in the news?

Studies in communications are torn between how influential engagement metrics are on journalistic discourse, and where this influence happens. Some studies based on self-reporting journalists support this theory, showing how journalists chase down trending stories that are incentivized on social media, while other studies show a clear resistance to this practice by journalists who do not let metrics fully decide the stories they focus on. A recent, comprehensive review of social media posts published by a variety of news outlets found that while the “overall effect of audience engagement on future news coverage is significant, there is substantial heterogeneity in how individual outlets respond to different kinds of topics,” noting the asymmetrical media structures of both conservative and liberal media outlets as an example.

However, these political divides dissipate when we look at the research focusing on sensationalizing, or the softening of news on social media platforms. Certain scholars notice these trends by focusing on the shifting features from hard to soft news. Hard news focuses on serious matters like politics, economics, and public safety, while soft news covers lighter and entertaining topics such as lifestyle, culture, sports, and human-interest stories. Recent empirical studies have suggested that there’s an intentional softening of hard news thanks to social media platforms, particularly Facebook. There was even a recent study in Germany that showed a 20 percent increase in human-interest topics on Facebook compared to print editions in German news outlets, all at the expense of hard news.

Soft doesn’t automatically equal bad. Soft journalism has been argued as being valuable to a culture, not because it’s necessarily political, but because it can help inform readers on the good life, educate folks on how to form identities, and nurture empathy among its readers. Moreover, while some empirical research shows that platforms shape content production and placement of news, there appears to be a substantial variation in how different news organizations and practitioners use metrics for various news topics and seek to balance their standards alongside social media characteristics.

In other words, the research isn’t completely cut and dry when it comes to audience engagement metrics influencing all newsrooms equally. As a recent study notes:

Our results suggest that audiences on Facebook can, to some extent, collectively affect news agenda for outlets….Journalism’s transition towards a Market Model may further threaten the health of the information environment, and consequently, undermine the democratic process where all voices are heard, represented, and paid attention to. Journalists may prioritize catering to what readers want to consume over providing news content that is needed for preparing citizens for participating in democratic politics. Still, we also see that a number of newsrooms resist audience metrics in news production and show little responsiveness to audience metrics.

But we do know a few things. We do know that certain platforms like Facebook have a News Feed that is constantly being enhanced and is basically a personalized machine learning model, changing and updating its outputs based on users behavior, the behavior of users’ friends on Facebook, and the behavior of the affinity and personality-based sub-group of users the system assumes you to belong to. There’s a set of algorithmic values that drive Facebook’s News Feed, which tends to push users toward news driven by audience metrics. Long term, I would imagine Facebook is fighting to keep audience metrics more–not less relevant–for newspaper editors as users engage more with their flexible News Feed.

And that’s sort of my point: while the research may be a bit unclear in determining truly how much these audience metrics are affecting journalists’ practices, certain tech companies have an incentive to keep the soft news and sensationalizing flowing downstream to the user–to develop a sort a public discourse that both serves their corporate interests while keeping eyeballs on their screens.

Look no further than the recent AI surge that has gripped media outlets in recent years.

As AI has become more popular, we have seen the pressure on journalists, scientists, and their institutions increase, and lead to a mutually beneficial relationship between sensationalism, misrepresentation, and subjectivity, prioritizing newsworthiness over integrity. Before ChatGPT’s release, scholars noted that AI representation was relatively fair, yet shallow, or had even grown more critical in the past decade. However, since the release of ChatGPT, we have seen empirical research finding that AI representation in the media is often sensationalized and tends to focus on warnings for readers. One study noted that ‘Impending Danger’ was the most prevalent headline category representing 37% of the total AI news stories collected. Another recent study discusses an attempt of journalists “freezing out of AI’s controversiality,” whereby the relationships between legacy media, journalists, and cited experts create a cold situation surrounding AI that emphasizes only the benefits of this new technology.

Such reductive approaches to AI representation can lead to mistrust in AI, establishing a barrier between AI and individual users. It can lead to an ineffective use of AI tools by adopters, and a lack of understanding of its impact on society.

When the great “unbundling” of newspapers took shape in the early 21st century thanks to the magic of search engines, each story became a separate product, naked and available for distribution on this new marketplace. Who needed the whole newspaper when you could get chunks of its content tailored to your interests? But, it also relinquished some of the power newsrooms had over their content and placed it into a new monetized structure online, where audience metrics can’t be ignored.

Recent communication research shows that these metrics aren’t the last voice in the editorial room just yet; that certain outlets seek to buck the trend. But AI representation in the media seems to be leaning into this new age of journalism, where sensationalism rules the day and the hype trains, driven by the tech industry, control the narrative.