Welcome to The Backstory, a new weekly newsletter that takes you behind the scenes of Times journalism. Today, our Election Analytics team does a dress rehearsal ahead of Tuesday’s big night, a photographer breaks down his drone flyby of surveillance towers along the southern border, and an editor addresses questions over how we use A.I. in the newsroom.
But first: Everyone wants to know who’s going to win the midterm elections. Satisfying that huge reader interest without making predictions can be harder than it sounds.
Why we dialed back the use of ‘blue wave’
News reporters don’t forecast the future in their stories. But the sheer number of polls, articles, headlines and language exploring the momentum of one candidate or party can have a cumulative effect.
When I was running The Times’s politics coverage in 2019, our desk and others around the newsroom published a string of stories about Senator Elizabeth Warren’s presidential bid, her ideas and Democratic excitement around her candidacy. The volume and tenor gave off a vibe that we favored her or thought she’d win. The experience prompted me to strengthen our reporting on other leading candidates and make sure we didn’t again publish several stories in a row that felt similar.
Which is why I felt a little déjà vu last month when we published several stories that explored Democratic energy in the coming midterms and kept repeating the phrase “blue wave,” a term for sweeping Democratic victories in an election. Then, early this month, we published an article about new Times/Siena polls under the headline “Senate Races in Republican States Tilt Toward Democrats, Polls Find.” One showed a six-point lead for the Democrat in Texas, James Talarico, a striking finding in a state that has not elected a Democrat to the Senate since 1988.
Did the volume and thrust of the coverage feel like we were drifting toward predicting “a blue wave,” in effect if not intent? I was thinking about that when I noticed a sentence in an article stating flatly that Democratic Senate candidates were leading or tied in red states like Texas and Alaska. I asked the editor to attribute that sentence to polls to avoid sounding like our reporters were forecasting a Democratic romp.
That small change was made, but it led to a bigger discussion. Michael Cooper, who now runs our politics coverage, made the smart point that while we wanted to avoid predictions, we also didn’t want to give our audience a false picture of reality or underplay our reporting and polling. The key, we agreed, was to prepare people in the homestretch for the unexpected. Recent stories about how Senate Democratic candidates were having a tough time in some blue states and videos about the economy, A.I. and democracy have helped do that.
With all this in mind, we also agreed to dial back the use of “blue wave” in news articles, to avoid a suggestion of inevitability.
As for our Times/Siena polls: We often say that polls aren’t predictions; they are a snapshot in time. Mindful that readers have felt frustrated with polls on past Election Days, I dug into our approach and broader questions of trust this week with Nate Cohn, our chief political analyst, and Will Davis, who oversees our election analytics.
“In the end,” Will said, “I will regard how well we did as journalists by how well our coverage reflected what ended up happening.” In other words, we want you to be prepared, not surprised.
Dress rehearsal
We test every part of our election night process and workflow, including the technology, the communication and the space in the newsroom. Go behind the scenes of a full dress rehearsal before the big day by clicking the play button.
Two ways we’re thinking about A.I.
I’m handing things over to Mike Abrams, the deputy editor for Trust, to talk about how, why and when we use artificial intelligence.
Among the toughest subjects driving a lot of conversation in our newsroom is how we use A.I. in our journalism. When is it OK for our journalists to use A.I., and where are the guardrails? What very human parts of our work should remain that way? How can we evaluate rapidly evolving tools so that they meet the standards we expect of ourselves, and that you expect of us?
We sometimes disagree on the answers. And that’s OK. People have strong feelings on A.I.: I, for one, have wondered if my kids have leaned too quickly and too heavily into A.I. to navigate their schoolwork.
Many Times readers have expressed deep concerns, too, about the potential threats to their job security and the climate. Some are also skeptical of our plans. If you’re part of that group, you can learn more about our approach to A.I. in coverage and in the way we work:
-
In a moving essay, a senior editor and columnist at NYT Cooking writes about her reservations while working on a new A.I.-powered feature for our iOS app. The tool helps home cooks who choose to use it customize recipes to their needs (for example, substituting or scaling ingredients). Some readers, including those who have not used the optional feature, have expressed questions and concerns about A.I., which the piece addresses.
-
Beyond the newsroom, our reporters and editors confront choices about covering A.I. all the time. To unpack those, we gathered a few colleagues for a conversation about making sense of online images when it’s tough to tell what’s real and what’s slop.
The Times has clear, concise principles for the use of A.I. in our journalism. But here’s an even shorter version: We use it carefully and with humans in charge.
How we got that shot
To understand how a “smart wall” of powerful A.I. cameras is carrying out mass surveillance along the southern border, my colleague Bora Erden documented towers in several states, using a camera-equipped drone. He shares what went into capturing this shot in Amada Acres, Texas:
I spent weeks this spring and summer traveling along the U.S.-Mexico border, from the laid-back surf towns of California to the saguaro cactus landscapes of Arizona. On one hot and humid August afternoon, my colleague Nicholas Nehamas and I found ourselves driving along the Rio Grande in South Texas, interviewing residents for an examination into the expanding network of A.I.-powered cameras in their border towns. These cameras can see for miles, and the number of towers is growing quickly as the Trump administration cracks down on immigration.
We were following a map of these towers, like bread crumbs along the border. Customs and Border Protection does not publish information on tower locations or type. We assessed the scale of the border-tower surveillance program through aerial and satellite imagery, geospatial analysis and reporting.
The previous day, Border Patrol agents had approached our car and requested that I take my drone down, claiming that we were breaking drone flight restrictions (we had not). Later that afternoon, a Laredo police officer asked us to stay put as he ran our credentials by his supervisor, checking if we were breaking any local laws (we were not).
Every time I encountered a border surveillance tower, I got to know them better, but I still felt at arm’s length. This tower from Anduril, an A.I. defense firm with ties to President Trump, finally broke the ice — it looked back at me.
We pulled over to the side of the road, and I flew my drone up until it was level with the top of the tower. Something to the south caught the attention of the tower’s radar, and the camera head rotated. I came face to face with this Wall-E-looking piece of cutting-edge surveillance technology. I knew this shot would show our readers the potential of the visual surveillance program — one of the largest operated by the U.S. government.
Your questions about our midterms coverage
Meg McKeon-Baker of Queens asks:
How do you determine when to declare a candidate the winner on election night?
Lindsey Rogers Cook, the deputy editor on our Election Analytics team, which is responsible for the statistical model that powers the Needle, answers:
We rely on The Associated Press for a majority of our race calls. It employs a team of analysts, researchers and race callers who have a deep understanding of the states where they declare winners. In a small number of high-profile contests, The Times independently evaluates A.P. race calls before making a projection. Both The Times and The A.P. will make a race call only when there is no possible path to victory for the trailing candidate.
Before we make or verify a race call, we want to be certain that the candidate will win that state or race. To determine this, we scrutinize where votes may remain to be counted in a given contest and what the likely makeup of those votes is. We are helped by pre-election estimates, statistical models, analysis of voting results and reporting from the county level on remaining ballots.
Readers sent a flood of questions about our midterm elections coverage. See our answers here.
Pop quiz
Which sentence below is correct, per Times style? Click to see the answer and read the article; check out the reason in the P.S. at the bottom of this newsletter.
-
Pasa and Yoyo, two female chimpanzees, are best friends. When Yoyo was faced with the opportunity to steal from Pasa, she rejected the experiment altogether, turning her back to the ropes and refusing to enter the room at all.
-
Pasa and Yoyo, two female chimpanzees, are best friends. When one was faced with the opportunity to steal from the other, it rejected the experiment altogether, turning its back to the ropes and refusing to enter the room at all.
We welcome your feedback. Email us at [email protected]. And see you next week. — Patrick P.S. Our style dictates that we call an animal it, not he or she, unless its sex has been mentioned or it has been personalized with a name. Which, obviously, Pasa and Yoyo have. Best friends forever.
The post Poll Watch appeared first on New York Times.
