Algorithmic Accountability
A news frame is the central idea or primary storyline that organizes the information you include in your news story and gives meaning to it. This section provides guidelines for framing stories that will help your audience appreciate the complexity of new technologies driven by algorithms, without misleading them about how great—or bad—those developments may be.
Even though they’re often characterized as such, algorithms aren’t truly black boxes. They’re “fundamentally understandable pieces of technology,” according to UC Berkeley computer scientist Joshua Kroll.
When journalists describe algorithms as confusing or inscrutable, they inadvertently draw attention away from the companies and people that produce them.
Instead, journalists should focus on the need for transparency and accountability from these companies and developers.
Seeking truth and acting transparently and accountably are human activities.
By contrast, AI systems are “not designed to seek the truth or engineered to ensure their representations of the world are factual and accurate,” as Bronwyn Jones, Ewa Luger, and Rhia Jones explained in a 2023 risk-based review published by the University of Edinburgh. In their stories, journalists should avoid framing AI outputs as inherently accurate or objective.
Likewise, as journalist Hamilton Nolan has argued, “AI lacks the accountability of a human journalist.” Journalists should make clear the limitations of AI, such as lack of transparency, as MandiCai and Sachita Nishal explain in their 2023 article on AI literacy for journalism.
Audiences need to understand both the technology behind AI and the context in which it was developed.
Automated decision systems do not operate in isolation; they are embedded in society, including specific institutions and organizations.
There’s no AI as we know it today without Big Tech, creating a power imbalance in which only a few companies control the vital infrastructure of AI, as Felix M. Simon explained in a December 2023 report on artificial intelligence in journalism.
This concentration of corporate control ripples out to other areas of society, potentially undermining public trust in information, devaluing human labor, reproducing inequalities, and degrading the environment, as Jones, Jones, and Luger noted in their 2023 review of generative AI, journalism, and societal risks.
News reports often frame technological developments in artificial intelligence (AI) as “races” between competing companies or nations in which the speed of innovation determines the victor.
Scholars who study “horse race” coverage of election campaigns have found that such coverage can increase public distrust in both politicians and news outlets while leading to inaccurate reporting of polling data and, ultimately, an uninformed public. New York University journalism professor Jay Rosen summarized his guiding principle for improved election coverage as “Not the odds, but the stakes.”
The same applies to AI journalism. Emphasizing who’s “winning the race” leads to flawed assumptions in reporting on AI technology and distracts audiences’ attention from other vital questions.
Racing metaphors often facilitate unjustified claims about future progress, which Sayash Kapoor and Arvind Narayanan, the authors of AI Snake Oil, have identified as one pitfall of AI journalism.
“Horse races are run on tracks, where progress happens in straight lines. This is not true for AI progress,” Kapoor explains. “Extrapolating AI advances based on recent progress is common, but could give a false sense of how quickly AI progress will continue.”
Framing tech development primarily as a contest between rival developers marginalizes questions about how new technology is likely to impact the general public, including what one international team of researchers has identified as a host of “sociotechnical harms.”
As AI competition escalates, “research about keeping these tools safe is taking a back seat,” according to Time. Without proper accountability and transparency, this technology could upend workers’ rights, accelerate the existing climate crisis, and exacerbate economic disparities.
Rapid developments in AI make it difficult for regulators to keep up: “By the time a congressional committee can hold hearings and draft legislation, AI will likely be several generations ahead.” But regulatory and policy recommendations —such as the 2023 Paris Charter on AI and Journalism—deserve more frequent and better coverage.
Investigate how bias in AI algorithms disproportionately affects marginalized groups and provides the public with misleading or incomplete information. Look into the data sets used to train AI systems and question whether they perpetuate existing inequalities.
For instance: ChatGPT is prone to producing biased and incorrect answers. “Some A.I. researchers have accused OpenAI of recklessness. And school districts around the country, including New York City’s, have banned ChatGPT to try to prevent a flood of A.I.-generated homework.” stated Kevin Rose in the New York Times.
Reporting that highlights interdisciplinary efforts, ethical AI initiatives, and international collaboration will help ensure that technology serves the public good.
For example, as Alaa Abdulaal, head of digital foresight for the Riyadh-based multilateral foundation Digital Cooperation Organization, told Yahoo, “It cannot be only done by government itself … It needs to take a cooperative approach, where we have at the same table the private sector, public sector, civil society — all of them sitting together to come up with the right set of frameworks for AI.”
There is nothing about AI that is inevitable; its development and trajectory are the result of human choices.
Be cautious when comparing AI tools to major historical transformations, such as the invention of electricity or the Industrial Revolution. As Sayash Kapoor and Arvind Narayanan explain, dramatic claims are “a great marketing tactic,” but they “often convey a false sense of potential and progress,” especially when they’re not supported by relevant evidence.
For instance, ProPublica’s Machine Bias series illustrates how crime news, a staple for many newsrooms, presents opportunities to address the impacts of AI, including how use of generative AI can reproduce and reinforce existing social inequalities.
AI systems powered by algorithms can potentially cause job losses and transform working conditions, leading to the devaluation of human creativity. Similarly, AI systems comes at a cost to the planet.
Not every story about algorithms needs to be a labor story or an environmental story, but journalists and newsrooms committed to algorithmic accountability reporting will always look for labor and environmental angles.
Include labor and environmental angles in your reporting on algorithms and AI, by raising questions about the labor and environmental impacts of technology, with both industry and independent sources.
For example, despite its significance in pioneering ten principles to “uphold the right to information, champion independent journalism, and commit to trustworthy news and media outlets in the era of AI,” the November 2023 Paris Charter on AI and Journalism received nearly no news coverage beyond reports by signatory organizations, such as the European Journalism Centre (EJC), the International Consortium of Investigative Journalists (ICIJ), and Reporters Without Borders (RSF).
Among English-language outlets, coverage of the Charter appears to have been limited to three reports: the Voice of America (VOA), which published an original article; Editor & Publisher, which republished RSF’s report; and Nieman Lab, which linked to RSF’s report. No national US news organizations appear to have even mentioned the Paris Charter.
Better reporting of developments such as the Paris Charter is one basic way that journalists can help the public understand how AI is impacting their profession—and, ultimately, the public’s right to freedom of information.
