AI and sports journalism
They're saying the right things. They always do.
This week’s topic comes from best man and chief of staff
.In Defector, Luis Paez-Pumar writes about ESPN’s plans to increase converge of the National Women’s Soccer League and the Premier Lacrosse League:
Hiring more journalists to cover women's soccer and men's lacrosse is a big win for the sports and for readers, and ESPN should reap the benefits of increased reporting by—oh no oh god it's yet another bullshit AI pivot.
Sure enough, ESPN announced that it will be using AI technology to create roundups of NWSL and PLL games.
The added coverage will augment existing coverage, and the initiative will extend to some other sports in the future. The AI-generated recaps aim to enhance coverage of under-served sports, providing fans with content that was previously unavailable. These sports do not currently have game recaps on ESPN digital platforms, and these AI-generated recaps will be a tool to augment existing coverage – not replace it.
Let’s start with this case and then work outward. To its credit, ESPN has a section in the announcement addressing quality and transparency:
Each AI-generated recap will be reviewed by a human editor to ensure quality and accuracy. Additionally, content created by AI will be clearly indicated for fans, via the byline – “ESPN Generative AI Services”
That’s good. But, as Paez-Pumar pointed out, merely labeling content as AI-generated does not address larger issues about how it was created.
Interesting to note also is that ESPN is touting that the AI has undergone a training process, but on what? By ESPN's own admission, the site does not provide NWSL or PLL recaps, which means that this AI will either be trained on other sports' recaps (which will surely lead to errors in terminology and the general sense that the article's author isn't someone who actually knows the sport) or, more ghoulishly, on NWSL and PLL gamers from other websites.
If it's the latter, then that makes an already cynical pivot, presented in benevolent language, even more insidious. The upshot would be that ESPN is beefing up its coverage of "underserved" sports (and who, pray tell, is responsible for this underserving?) not by hiring people who can and already do write these kinds of stories, but rather by feeding existing soccer and lacrosse journalists' work into a machine aimed at making them obsolete.
But for me, the interesting part of ESPN’s announcement comes under a heading that’s simply labeled “Why?
The aim is to learn, determine how to responsibly leverage new technology, and begin to establish best practices – all while augmenting our existing coverage of select sports and allowing ESPN staff to focus on their more differentiating feature, analysis, investigative, and breaking news coverage. (emphasis added)
Of course. That’s always the case.
That’s always what they say.
If this all sounds vaguely familiar, it’s because AI content has been around sports journalism for the better part of a decade now.
The AP has been using automated stories and other content since 2012. Starting in 2019, the AP was automating some men’s basketball recaps and previews of NCAA Tournament games.
According to the AP story “The automation of data-driven stories frees up journalists to focus on enterprise reporting and reduces the amount of data-processing work.”
Marc Zionts, CEO of Automated Insights, the company that provides the AP with its automated content, said this:
It’s an honor to play a fundamental role in the advancement of the AP’s drive toward innovative, automated news generation for data-driven stories and see the way it’s enabled prized writers to spend more time writing critical, qualitative articles.
See.
It’s what they always say.
It is a compelling idea, isn’t it?
The most basic function of sports journalism is to tell me what happened in last night’s game. There are absolutely cases where some kind of automated computer technology can do this efficiently. And there’s some sound logic in letting automated computer technology do what it can do efficiently.1
And yes, that in theory2 frees us humans up to do the type of human stories that computers can’t do.
There are limits, of course. If you feed a generative AI script this box score, you’ll get a functional recap. But there’s nothing in the box score to tell AI that the story from the game was Tua Tagovailoa concussion, not the Bills winning or James Cook’s three touchdowns.3
And you can point out that that’s an exception, and I’ll agree with you.
Journalism, though, is often writing about the exceptions.
That’s also where the promise of AI lives, right? That’s the core message of what they always say. The AI will do the basic stuff, freeing reporters up to do the types of deeper, longer stories that matter.4
That gets at something Galen Clavio and I wrote about a few years ago, something I’ve been thinking a lot about lately — the Platonic Ideal of sports journalism. It’s this idea of sports journalism the way it oughta be, filled with in-depth stories and features and investigations and all the things writers love to do.
A quick aside about about the journalism pipeline.
This is a bit of a dated concept, given how media has gone in the last 15-20 years, but jobs writing, rewriting, editing those recaps were an entry point for generations of young reporters. It’s how a lot of us got our start in the business. It’s one of the canonical journalism career tropes — a kid starts out writing recaps, gets a chance to cover a few things here and there, proves themselves through hard work and grit and moves up the ladder.
Introducing AI recaps at scale eliminates those jobs from the pipeline. This reduces the opportunities for young people to get into the business, which reduces the number of young people in the business. They’ll never get to write critical, qualitative articles because the jobs that gave them the start in the business are being phased out.
There’s a couple of unspoken assumptions at the heart of this story
Back to the quotes from earlier:
The automation of data-driven stories frees up journalists to focus on enterprise reporting and reduces the amount of data-processing work
The way it’s enabled prized writers to spend more time writing critical, qualitative articles.
Allowing ESPN staff to focus on their more differentiating feature, analysis, investigative, and breaking news coverage.
The first assumption is that the Platonic Ideal of Sports Journalism is, well, the ideal. That these are the stories are what sports journalists should want to do. That’s a much larger discussion and nuanced discussion for another time and, admittedly, much more academic.
The second assumption is much more pragmatic.
The second assumption is that journalists, writers, and staff members will still actually have jobs. Jobs where they are supported, where they are given money and time and resources to actually do the kind of stories talked about in those quotes.
What’s more likely: That news organizations will continue to hire and employ journalists to do the kind of investigative, in-depth stories that AI can’t do? Or will news organizations view AI as the ultimate cost-cutting measure, and since AI can do the basic job of sports journalism, that allows them to reduce staff,
Right now, they are saying all the right things.
They always do.
I know that there are things that make generative AI technologically unique, but part of me can’t help but think that this is just automation dressed up in the dazzling veneer of 2020s tech bro speak.
Why yes, the italics are doing a lot of heavy lifting.
The official stance of Sports Media Guy vis a vis the Bills offense is Let James Cook.
And if this sounds to you like another example of boundary work in journalism, thank you for paying attention all these weeks!
I've been really intrigued to see when AI recaps would break into the mainstream. When I was working as an SID, we used generative AI built into SideArm to start recaps. But it was horrible and would give us 1-2 usable lines. AI's come a long way since then (2022), but it always struck me as an unnecessary extra step that still required human input. I'm worried about ESPN training their AI on other sports, presumably. Why not just have a journalist write the story instead of having to fix everything the AI did?
To your point: https://awfulannouncing.com/espn/alex-morgan-ai-generated-recap-mention.html