When eight news companies joined forces under the Journalism AI collaboration to look at whether artificial intelligence could help fight bias in the media, questions abounded. Which biases? Were they all bad? Could a machine do better than experienced editors? How did each group’s experience differ given we had publishers from Argentina, Japan and Europe?
Bias in the media has become something of a lightning rod in the last years. At its worst, failure to represent and report on different people, points of view and lived realities feeds the social divisions and ructions we see around us all too often. It leaves great swathes of people without a voice and perpetuates long-standing imbalances and injustices. If the moral argument isn’t enough, money also talks. If news publishers under-represent or misrepresent parts of our population, they will likely fail to find new subscribers beyond their established base or, in the public service world, be called to account with potential penalties.
At flashpoint moments like the death of George Floyd in May, news organisations are forced to turn the mirror on themselves and ask how well they are truly representing communities beyond powerful men in suits who typically dominate the front pages. To do that, this group decided to break the task into three: to understand, identify and mitigate bias. To work out where machines could help, we focused on several areas where bias can slip in. Here are three key ones.
First, what do publishers consider to be news? How is “news judgment” formed? How much of it comes from traditions, from clicks or TV ratings, and how much comes from being openly curious about the world? Of course, there’s a growing number of niche publications that serve specific groups and so have conscious biases in their approach. But for those media that serve a broader audience, it is key to ask how the news agenda is determined – and by whom. Without giving positions of influence to people from different backgrounds with varied perspectives, the media will continue to see - and project - the world in the same light.
Beyond what news media cover, the group turned to how stories are reported.
Who are the go-to sources? Who else is profiled or quoted? What are they asked about and how are they represented? Despite more women being in powerful positions now than a generation ago, there is still a tendency to describe a woman in terms of her appearance or relationship status. Relatively few people of colour are given column inches or airtime and even fewer are quoted as experts.
The third area was how media illustrate the news visually. Who is shown doing what? Are young people shown being more outrageous than they are in reality? Do people of colour feature more prominently in the entertainment and sports pages rather than the business and technology sections? How many women feature in news photos vs men? What people see has an impact. A poll of 8-12 year-olds last year found that almost a third of children in the United States and the UK wanted to be a vlogger. Another 40% wanted to be a musician or sports star - the people they saw as successful. That can change. As U.S. Vice President-elect Kamala Harris said “While I may be the first woman in this office, I won’t be the last. Because every little girl watching tonight sees that this is a country of possibilities”.
So how might news publishers fight some of these biases?
Various other projects have tackled gender imbalance in the news, and as part of the JournalismAI Collab the team behind the AIJO Project decided to add our machine learning experiment to that work. Using facial recognition technology, each publisher provided a week of photographs to see how many men and how many women were featured. You can read more about how the experiment worked here. Depending on how the technology develops, potential next steps would be to add racial diversity and other measurable forms of bias across text, audio and video.
The media cannot change who is in charge of countries, institutions and companies and we will continue to cover them regardless of race, gender or age. But we can decide how we report on the impact those leaders have on real people of all backgrounds and the way we represent diverse communities.
As the late journalist Gwen Ifill said:
“Diversity is essential to the success of the news industry… We have stories to tell, but many in our audience have stopped listening because they can tell that we're not talking about them.”
* After this blogpost was written, we were given an opportunity to also analyse our texts through the Gender Gap Tracker. You can learn all about it here.