How to report on artificial intelligence

Credit: Franki Chamaki on Unsplash

AI is a complex field and there is both hype and hysteria around the subject.

To help you explain it clearly to your audience, two community coordinators from the JournalismAI initiative asked four AI experts and tech reporters from across the world for their best tips.

Build a solid foundation

Karen Hao, senior AI editor at the MIT Technology Review, started covering AI spending lots of time on YouTube and learning the basics like: what is an algorithm? What are neural networks and how are they trained? What are the key milestones in the history of AI? And she also read tons of papers on AI from the open-source database arXiv.

"In the first few months of me starting the beat, I would go there every week and read five to 10 papers just to start familiarising myself with the terminology and the popular research concepts that experts were repeatedly using."

Once she gathered the key information from the papers, she would double down by further researching keywords on YouTube and other online resources to keep building a solid knowledge.

Aside from research papers, it is also important to be well-versed in reading laws and regulations related to AI, says Melissa Heikkilä, AI correspondent for POLITICO Europe.

"When you write about technology and AI, often you have to talk about things like cybersecurity and data protection. When that happens, it’s useful to be able to read regulation like the GDPR, understand what it says and reflect on it in your reporting."

Beat the hype

AI reporters always read new research papers to understand the field and to stay on top of new developments to report on. But how do you translate research findings for your audience to avoid the hype?

Maneesh Agrawala, director of the Brown Institute for Media Innovation at Stanford University and Karen Hao agree that it is important to always run through a set of fundamental questions.

• How was the research conducted?
• How many experiments were done?
• What data was used?
• What initial assumptions did the researchers have when approaching the study?
• What harm might come out of this kind of technology?
• What is the error rate of the algorithm?
• Why does the algorithm fail with certain types of input?

Agrawala adds that reporters should also consider whether they might be able to test new algorithms and systems themselves.

Hao also recommends checking with other researchers, calling them up, asking them to explain the research and then summing it up back to them.

How can we explain citizens' fears without dismissing them as irrational?Genevieve Bell

"When I explain it to them using my own words, that’s when I know what I did or did not understand. At that point, the researcher can correct me and help me clarify any doubts I might still have."

Complicate the narrative

According to Genevieve Bell, director of the 3A Institute at the Australian National University, it is important to reflect on the common narratives and images surrounding AI and what is their historical context.

"We see an over-indexing on fear and anxiety. Think of the persistent notion of killer robots on one hand, and the ubiquitous image of a robotic hand approaching a human hand. What do these images and narratives do? And why do they resonate with people?"

Bell argues that the fear and anxiety we feel about new technologies are a product of centuries of stories about "non-human things coming to life and taking on quasi-human forms or characteristics" in myths, religion, and literature. And this is something journalists must take into account in reporting on AI.

"I would like to see news coverage acknowledge the depths of our imaginations of AI and problematise the ways we talk about both its perils and possibilities today. How can we explain citizens' fears, without dismissing them as irrational, and ask hard questions about the hype that often surrounds the introductions of new technologies?"

Bell adds that AI is only ever part of a solution or a system, hence it is necessary to frame questions at a systems level and to not focus only on the technology. It is also important to make clear that there is no single AI but many – in many different companies, countries and contexts.

Be compassionate but embrace critical thinking

Finally, since the field is still developing, Hao feels there is a lot of thinking that needs to be done about the harmful implications of AI on society. It does not matter how well-intentioned a researcher is, there are going to be blind spots to their work. Critical thinking around reporting AI is an important muscle to develop.

"It’s really important to be compassionate to researchers, just because someone made a mistake doesn’t mean they’re evil. You should recognise when they’re genuinely trying and have good intentions at heart but you shouldn’t cover their work uncritically just because they have good intentions."

Heikkilä agrees.

"It's really, really important to apply a critical lens to technology. And the worst thing we could do is be lulled into the false belief that technology will save everything and that every problem in society can be fixed with technology."

[These recommendations have been edited for brevity. An extended version of this piece, with additional recommendations from Xiaowei Wang, Carl-Gustav Lindén and Andrew Meares, will be published on the POLIS blog]

This article originally appeared in the JournalismAI newsletter and is republished here with permission. Learn more about JournalismAI on its blog.

Free daily newsletter

If you like our news and feature articles, you can sign up to receive our free daily (Mon-Fri) email newsletter (mobile friendly).