Ethics, Human Values Critical to Getting AI in Journalism Right

Georgia Tech alum, Nick Diakopoulos, releases his first book "Automating the News: How Algorithms Are Rewriting the Media" this month.

It has been nearly ten years since Nicholas Diakopoulos earned his Ph.D. in computer science from the School of Interactive Computing at Georgia Tech and co-founded Computational Journalism program at Georgia Tech.

Since then, he has been around the world teaching and researching how computer science and journalism can work together efficiently and effectively. He has even written a book about it – Automating The News: How Algorithms Are Rewriting the Media, which debuts this month.

Computation + Journalism

The son of a newspaper and magazine editor, Diakopoulos was exposed early on to the world of news media. One of his first jobs was helping with layout and design, as well as managing subscriptions, for a magazine. Still, he was drawn to computing. When his family brought home a Tandy 1000 personal computer he and his brother would spend hours coding programs and games in Basic.

It wasn’t until his first year of graduate school at Georgia Tech that he started to see the connections between media and computing.

Originally drawn to Georgia Tech by the strength of its computer vision graduate program – currently ranked #2 in the United States – and the campus’ proximity to a buzzing city like Atlanta, it wasn’t until he was on campus that Diakoplous discovered his passion for human-computer interaction (HCI).

Lucky for him, Tech was also strong in this area. He soon began publishing research at HCI conferences like Conference on Human Factors in Computing Systems (CHI), while still studying computer vision with his advisor, Irfan Essa, the director of the Machine Learning Center at Georgia Tech (ML@GT).

It wasn’t until Essa returned to the lab one day from a meeting that a true idea to combine computing and journalism was sparked.

“I remember I was sitting at my desk messing around with some news graphic and interfaces when Irfan walked over and said he was just at a meeting at CNN where they were talking about computational journalism. He wasn’t sure what it was, but he encouraged me to figure it out,” said Diakopolous of the 2006 conversation.

The seemingly random conversation would soon lead to the first-ever seminar in computational journalism co-taught by Essa and Diakopolous in 2007. At CHI 2007, Diakopolous ran into fellow Georgia Tech alum, Brad Stenger, who was running WIRED’s Next Fest in San Francisco. Steiner was intrigued by the seminar and suggested creating a computational journalism event.

In Spring 2008, Georgia Tech hosted the first Computation + Journalism Symposium. That first event brought over 100 students, faculty, and industry employees together to discuss possible partnerships and ways to bring computational thinking together with journalism. The symposium must have struck a chord with attendees because it is still going strong today.

The Future of Artificial Intelligence and Journalism

With nearly twelve years of research in computational journalism, Diakopoulos had an urge to write a book, but was unsure if it was something he should do. After co-authoring a textbook with a senior colleague, Ben Shneiderman, the process of writing a book was demystified and he took the plunge.

Due in June 2019, the book explores the connections between artificial intelligence and journalism. Diakopoulos hopes that readers take away two lessons.

One, that all technology, including AI and algorithmic technology, has the capability to embed human values. Journalism itself is a values-driven institution that holds ideals like independence, verification, and accuracy in high regard. Diakopoulos encourages journalists to step up and collaborate with computer scientists to design AI the right way before someone else steps in and designs agents with other values.

Diakopoulos also emphasizes that AI is not taking away journalism jobs. In fact, his research shows that AI is creating jobs. With the introduction of AI into the workplace, employees are needed to edit and create knowledge bases, maintain quality assurance, manage the agents, and more.

“AI will make jobs change and shift, but it won’t take jobs away. AI is too brittle and bound to the data to completely replace a journalist. The most productive path forward is a collaborative, hybrid relationship between journalists and AI,” said Diakopoulos.

Diakopoulous is excited about this future hybridization because of AI’s ability to sift through large data sets and find original content. This unique content can help to turn readers into subscribers, which affects a news organizations bottom line. Diakopoulous believes that his work helps to make this clear, and that it could serve to inform responsible and strategic adoption of AI in news production.

Little research has been on how people perceive AI-written content, and how this content impacts things like reader trust, but Diakopoulous is looking forward to the longitudinal studies to come.

Besides writing content, Diakopoulous has a team of researchers that is working on a project that is AB testing headlines. He hopes to create a playbook that is data-driven and helps inform writers and editors on how their linguistic choices actually effect search engine optimization or readership.

The Yellow Jacket Effect

Diakopoulous credits Georgia Tech with instilling a mindset that looks at computing broadly.

“Georgia Tech taught me that it’s not about the algorithm or optimizing things. It’s about understanding how computing technology can affect all kinds of stakeholders and the world around us,” said Diakopoulos.

He was also impressed with the GVU Center at Georgia Tech and has tried to bring that ethos with him into every job.

“I loved how the GVU Center was able to draw people from across campus. It was a powerful lesson and eye-opening experience to me on how universities could work. It showed me that we don’t have to be siloed into our individual research, but that some amazing partnerships, friendships, and projects can come out of working with people from around the institute.”

After graduating from Georgia Tech in 2009, Diakopoulos was selected as a Mass Media Fellow by the American Association for the Advancement of Sciences (AAAS), a program that places Ph.D. and graduate students in news rooms across America to be science reporters. He was placed at The Sacramento Bee for a summer.

While he enjoyed his time at the newspaper, he was drawn to academia because of its mission to research and develop new knowledge, while also giving him a platform to explore problems that could have a diverse impact across the world – not to mention the longer deadlines. He left the United States to become an assistant professor position at University of Bergen in Norway.

Diakopolous expected to stay in Norway for a while, but was granted a stimulus fund to do a post-doc at Rutgers University. He took the opportunity and moved to New York where he completed the Rutgers post-doc and fellowships at PUNY and Colombia University before accepting a tenure-track position at the University of Maryland in their School of Journalism. In 2017, Diakopoulous became an Assistant Professor in Communication Studies and Computer Science (by courtesy) at Northwestern University where he is Director of the Computational Journalism Lab (CJL).

For students earning their Ph.D., he encourages them to think about how their research will impact the public.

“If you think about the implications your research could have, it might lead you to ask more impactful research questions. It’s a different way to go about research, but I have found it to usually be quite effective,” said Diakopoulous.

Contact: 

Allie McFadden

Communications Officer

allie.mcfadden@cc.gatech.edu