Nieman Foundation at Harvard
HOME
          
LATEST STORY
There’s another reason the L.A. Times’ AI-generated opinion ratings are bad (this one doesn’t involve the Klan)
ABOUT                    SUBSCRIBE
May 21, 2014, 10:39 a.m.
Aggregation & Discovery
LINK: medium.com  ➚   |   Posted by: Joshua Benton   |   May 21, 2014

Of the 43,592 response pieces to The New York Times’ innovation report, this one by Joshua Lasky, who works in marketing at Atlantic Media’s branded content/strategy arm.

The Times report complains about the state of article tagging — the assignment of metadata that says, for instance, that a story is about the crisis in Ukraine, has a serious tone, is a profile, ran in the paper with a slideshow, and so on. Much of that metadata at the Times is still based on old standards. The report says this is holding back efforts to create new products for readers or come up with new ways to surface content to audiences. That’s true, but Lasky notes the other loss: Bad metadata makes it harder to know what’s working and what’s not working.

Adam Felder, Associate Director of Analytics at Atlantic Media, uses metadata to help contextualize how different subjects play to audiences. On the back end, using modified baseball statistics such as batting average and slugging percentage, he can compare any number of tags against each other. This could include any of the structured data tags that are mentioned in the NYT report: geographic location, story type, story tone, etc.

There’s a certain amount of content optimization that can be done here. Think if you were to tag articles based on whether they included a video; you could then analyze whether having this element improved article performance. This would help you to decide whether it made sense to embed more videos in the future. I wouldn’t trust an analysis of what sorts of content/elements work for a given audience without the use of metadata to back up the findings.

There’s a dire extrapolation of what Lasky is talking about — everyone, the Times is no longer printing sad stories, turns out they underperform summer cocktail recipes — and as the Times report notes (p. 78), journalists often like to view subjects like this through a worst-case-scenario lens. But it makes a ton of sense for news organizations to know where their investments are paying off — and what they can reasonably stop doing. To do that, you need structured data that describes your work — that’s metadata.

Show tags
 
Join the 60,000 who get the freshest future-of-journalism news in our daily email.
There’s another reason the L.A. Times’ AI-generated opinion ratings are bad (this one doesn’t involve the Klan)
At a time of increasing polarization and rigid ideologies, the L.A. Times has decided it wants to make its opinion pieces less persuasive to readers by increasing the cost of changing your mind.
The NBA’s next big insider may be an outsider
While insiders typically work for established media companies like ESPN, Jake Fischer operates out of his Brooklyn apartment and publishes scoops behind a paywall on Substack. It’s not even his own Substack.
Wired’s un-paywalling of stories built on public data is a reminder of its role in the information ecosystem
Trump’s wholesale destruction of the information-generating sectors of the federal government will have implications that go far beyond .gov domains.