How data is transforming journalism: obstacles and opportunities
Nearly two decades ago, Christie Aschwanden, currently the lead writer for science at FiveThirtyEight, began her journey to becoming a “stats nerd.” Little did she know that during that journey some would consider her a character on the side of evil
Armed with a reporter’s notebook, Aschwanden set out on her first major assignment: to see what effect mammograms were having on cancer mortality rates for women. What she found was that the data told a strikingly different story than the one accepted by the public: mammograms appeared to make very little difference.
Elated, she brought her data scoop to her editors and told them she wanted to write an everything-you-know-about-mammograms-is-wrong article. They weren’t interested. Aschwanden would later find another publication to run the story. And that’s when the hate mail poured in.
“Once people have a belief it’s really hard to change their mind, and it’s really hard to change their mind with data and facts,” she told a room of statisticians attending Sense About Science USA’s all-female panel on statistics reporting at the Symposium on Statistical Inference.
“We hate uncertainty as human beings,” she told the crowd. “Mammograms offer certainty, or at least that’s the story that’s told.”
The theme of uncertainty in science reporting was something all the panelists spoke about during the session chaired by Rebecca Goldin, Professor at George Mason University and Director of the STATS project, a collaboration between Sense About Science USA and the American Statistical Association.
“Communicating uncertainty does not have to be complicated,” Laura Helmuth, Health, Science and Environment editor of the Washington Post told attendees. She went on to cite an example of a clear explanation of uncertainty from the children’s arts and science magazine, muse.
Aviva Hope Rutkin, Big Data and Applied Mathematics Editor of The Conversation U.S. explained to the audience what she and her colleagues do to ensure they are not creating opportunities for readers to misconstrue data. She also shared the New York Times You Draw It series (an interactive way for people to test their notions of trends and topics against the actual data). But, while the media does its best to prevent misinterpretation, things can go awry. Rutkin offered an example of an article which had explained the limitations of a data set in the text but, because it was accompanied by an appealing, colorful graph on social media, the graph and not the whole story ended up going viral.
And that’s when scientists can step in and help journalists.
“The most effective way to stop viral nonsense from going viral is to, as soon as possible, say no this is wrong and this is why,” said Laura Helmuth. “And if you can do that with authority and clarity it can often stop disasters from happening.” She urged attendees to write op-eds for publications, first-person narratives on their areas of expertise, and to jump into social conversations happening on Twitter. Hosting an Ask Me Anything session on Reddit is also a great way experts can stop the proliferation of misinterpretation.
It is an exciting time in science reporting, as there are more tools now to help ensure data stories are delivered accurately and clearly. SAS USA’s STATS project has an online helpline, which gets reporters answers to their research questions before they publish. These can be questions about percentages, risk, odds ratios, methodology, and more. Once a question is submitted, the writer is paired with a stats expert. And it’s free!
Video produced by Sense About Science USA with the help of the American Statistical Association
Other videos from the American Statistical Association’s Symposium on Statistical Inference produced by Sense About Science USA.
The opening plenary session at the Symposium on Statistical Inference—What Have We (Not) Learnt from Millions of Scientific Papers with P-values?—from John Ioannidis, Stanford University.