Although the daily challenges of reporting and publishing have never been greater, both must continue to evolve. Being a leader of that change is the mission of Julia Angwin and her team at Proof News, which she founded in February 2024 and serves as its CEO. She was previously with ProPublica and founded The Markup in 2018 to bring her version of data-driven investigations of technology into the newsroom environment.
“I named it Proof News because, in the world of AI and misinformation and propaganda circulating online, we must make it very clear why our audiences should trust our piece of information versus another piece,” Angwin said.
The underlying principle of Proof News is that the news industry and the public will be better served when journalists apply scientific methods to their reporting. Often, journalists are “witnesses” and “storytellers” first and “analysts” last, or helping the audience understand what is witnessed and the essence of the stories, she said.
Angwin and her staff are also reapplying the goal of objectivity in journalism from the performance of reporting to the process of reporting, which parallels the scientific method.
“Objectivity was meant to describe the process’ thoroughness: Does your process’ rigor counteract your inherent bias? Because the methods are rigorous, the results will be reliable, and that’s what objectivity should be. We are conducting audience testing during fall 2024 to determine if our meticulous methods are digestible and understood,” Angwin said.
Proof News is testing its model by including “ingredient” labels on all its investigations, which describe the scientific approach utilized in an investigation. On its YouTube channel, Proof dives deeper into discussions of the ingredients of its recent investigations and interviews others about the ingredients of their investigations.
For example, in July 2024, Proof published a video, “How We Investigated the YouTube Videos Swiped to Train AI.” This investigation was inspired by an article by Alex Reisner for The Atlantic, which stated the contents of many books were being used to train AI. He eventually created a searchable tool, so other authors could determine which books were being scraped. When Angwin discovered that YouTube videos were also used for the same purpose, she asked Reisner to create a similar search tool for YouTube videos.
Proof News Reporter Annie Gilbertson dove into the literature to determine which Big Tech companies used the YouTube datasets. She confirmed that companies, including Apple, Anthropic, Nvidia and Bloomberg, had used the datasets for their AI models without the knowledge or permission of the YouTube creators.
The AI Democracy Projects is a collaboration between Proof News and the Science, Technology, and Social Values Lab at the Institute for Advanced Study.
“Late last year, I was concerned about AI and the elections. Not having the means to understand and monitor this new technology was scary. I joined with Dr. Alondra Nelson, the Harold F. Linder Professor at the Institute, to conduct AI testing in January 2024,” Angwin said.
Nelson and Angwin recruited election officials from approximately 20 states to rate the answers to voter queries from five leading AI models. Most of the answers were inaccurate and didn’t deliver on the promise of AI companies to provide high-quality responses to elections and voting inquiries. After this year’s election, Angwin wants to shift the focus to testing AI models on civic participation and information.
Proof News offers its content to any news outlet for republishing at no cost at republish@proofnews.org. For more information, email info@proofnews.org or julia@proofnews org.
Bob Sillick has held many senior positions and served a myriad of clients during his 47 years in marketing and advertising. He has been a freelance/contract content researcher, writer, editor and manager since 2010. He can be reached at bobsillick@gmail.com.
Comments
No comments on this item Please log in to comment by clicking here