End of a Journalism Experiment

by | Oct 27, 2023 | Blog, Corrections | 1 comment

We lived by the API, and we were strangled by the API. In the process, we learned a great deal about the possibilities and pitfalls of journalistic corrections – and corporate control-freakery – in our modern, networked information ecosystem. 

Today, we acknowledge a fundamental, yet foreseeable, mistake: giving undue trust to the owners of centralized online platforms. We didn’t anticipate – though history should have taught us differently – that they would pull the rug out from under countless organizations and people who’d created many of the tools that made the big platforms successful.

When it worked, the tool we developed at the ASU Cronkite News Co/Lab, called Correx, offered a distinct improvement in a specific but important facet of trustworthy journalism. Correx helped journalists send corrections and significant story updates down the same social-media pathways that the original stories had traveled. It did this by offering an easy-to-use dashboard that enabled journalists to discover which social media users with large followings had shared the original stories, then  alert major story sharers to the corrections. 

Correx used two key platforms where journalism was (and to some extent still is) widely shared: Twitter and Reddit. It relied on the sites’ APIs — applications programming interfaces. 

What are APIs? Think of them as rules of the road for people who write software that connects to some of the data that big platforms collect from their users and then make available publicly. In its simplest form, an API is like the electrical outlet in your home. Makers of things like lamps, TVs, and other electric-powered appliances know that if they make the power cord and plugs to certain standards they will receive electricity when plugged into the outlet.

Similarly, we were able – for a time – to rely on APIs at Twitter* and Reddit. The Correx software plugged into the APIs and received data from the services. We were able to show journalists tables of who’d shared their work, ranked by the number of followers or upvotes the sharers had earned. From that data,  we offered a dashboard giving journalists a streamlined way to contact and reply to those sharers. 

If you’re interested in the nuts and bolts of how Correx worked early on, please read this. Here’s a screenshot showing the results of a Correx search of Twitter posts:

Over time – including a fallow stretch during several years of the covid pandemic – we built more and more capabilities into the tool. By early this year, we were increasingly pleased with the functionality, and were looking for ways to more broadly test and deploy it inside the journalism craft.

This year, it all came undone. The people who control Twitter and Reddit told third-party developers to either pay for access to the data or get lost. We didn’t have the money to pay what looked like extortionate rates.

So we’ve shut down Correx, with regret.

The code lives on, however. We’ve posted it in a public repository, so others can look at it and use it – as is or modified to their own tastes – in projects of their own. More on that below.

Why corrections mattered (and still do)

Before we dig into the issues that led to the (for now, anyway) end of the Correx experiment, it’s important to explain in a bit more detail why we developed it in the first place. Context matters in journalism, and we think it’s valuable here, too.

The project came out of the News Co/Lab’s purpose: to advance media literacy in the digital-information ecosystem. We need to do it at scale, reaching the largest number of people possible. Among the institutions that can bring scale to this is the news media, which despite its financial woes still commands a significant audience.

One way journalists could help their communities understand how information works, we reasoned, was to be transparent: explaining who they are, how they do what they do, and why they do it. Straightforward, timely corrections of mistakes would necessarily be part of that transparency.

At the start of the project, before the pandemic, we convened a group of journalists in New York City to learn more about how newsrooms were approaching corrections. We learned that there hadn’t been much innovation in journalistic corrections in the digital age, nor was there broad consistency on corrections policies across organizations. One improvement, which has become somewhat standard, is to correct mistakes on web pages so that people coming to the story afterward would not be misled. We hoped to go one better: reach back into the past to inform people who’d already seen bad information that the reality was different. 

To that end, Correx aimed to semi-automate part of the corrections process on social media, which has become a key way people learn about various timely topics. Journalists could help the truth catch up with the mistakes – and the people who’d seen them.

With our tool and other measures, such as standardizing corrections methods and data that corrections generate, we thought journalists could not just improve public understanding but also enhance trust in what they do. (Partly right: A research study conducted as part of this overall project, in collaboration with Dartmouth College’s Brendan Nyhan and his students, showed that journalistic corrections did improve the journalism consumer’s understanding of the truth, but at some cost, at least in the short run, to trust in the media organization.)

When we tested an early version of Correx with several news organizations, the feedback from the journalists was positive. And as noted above, we were moving ahead with a major upgrade when the API debacle did us in.

Trust and control 

Even if we’d had enough money to pay the platforms for the use of their APIs, we probably wouldn’t. Why? Because we don’t trust them.

Not that we should have in the first place. If we’ve learned anything from the technology industry, it is that companies creating platforms will, more often than not, turn on the third-party developers whose work helps the platforms achieve dominance in the first place. The process goes like this:

  1. Company creates a platform and tries to achieve a critical mass of users.
  2. It offers APIs that give third-party developers an easy way to make the platform better for users.
  3. Those third parties become companies in their own right, and sometimes make a lot of money by participating in the platform ecosystem.
  4. Critical mass is achieved, with millions of users and an increasingly robust ecosystem of apps and people.
  5. Platform owner, having achieved dominance in its own area, restricts or charges the third parties for access. Sometimes this is done gently, but often not.
  6. Some third parties pay whatever it takes to keep their business going. Others close down.

Platform owners who do this frame it as “Well, we have a right to make money on what we provide.” This is partly true; they have the right to try to make money. But it’s much more true that they would not have become powerful enough to pull this bait-and-switch had the third parties not done so much to make the platforms better for everyone.

This cycle has happened again and again. The only surprise is that we keep being surprised by such behavior. 

In Twitter’s case, Elon Musk and his much-shrunken corporate team slammed the door last spring. Reddit followed on the heels of Twitter soon thereafter. Our integration with both of the platforms has been broken, and we don’t have the money – or, as noted, the inclination – to pay them.

Again, we confess to insufficient foresight. We knew, abstractly, that the companies might do something of this nature. But we didn’t plan for it.

Open source

From the start, we had a semi-fallback in mind. We planned to eventually make Correx “open source” – designating a category of software in which the source code, or programming instructions, are openly available for re-use and modification by others.

“Eventually” arrived sooner than we expected. As of today, we’ve posted our code on the best-known of the repositories for open source projects, GitHub. If you’re interested, take a look here.

Why would anyone want to look at code that has stopped working due to the platforms’ restrictions? 

First, this is good code. It might well be adaptable for other kinds of uses, by people with ideas that we haven’t had ourselves.

Second, it’s possible that Reddit might see how it’s damaging itself with the new policy. If that happens, the application will be ready to deploy again. (We don’t foresee Musk, who has made a series of astonishingly self-destructive business moves with Twitter, changing his ways.) 

Third, it’s possible that other platforms will have usable – and more durable – APIs to which someone using the Correx code can quickly adapt. Bluesky seems to have that potential, given that it is more open than all but one of the other platforms that are trying to win over Twitter refugees.

Today’s most open social-media technology is what’s known as the “fediverse” – a federated network of networks, all using a set of open “protocols” (rules of the road) and no central entity directing and controlling what users do. In the fediverse, the most popular application is Mastodon, which has many Twitter-like features and is making swift progress. About 15 million accounts now exist in the Mastodon networks, though many are dormant, but there’s great potential.

Mastodon does corrections right

We’ve looked at Mastodon closely, and Correx could indeed fit well into that ecosystem. The irony, however, is that Mastodon – uniquely among social networks – already does corrections the right way. This means Correx isn’t immediately necessary, though the dashboard/reply functions could make things easier for journalists in any case.

The way Mastodon does corrections is simple, ingenious, and elegant. The software has an editing function that enables you to change what you’ve posted. The beauty of how it does this is to insert the edited post into the timeline of anyone who follows you, so they automatically see your correction – if that’s what the edit includes – without further intervention on your part. 

Edited Mastodon posts also contain a flag that indicates they’ve been edited, and you can see every version if you wish. Here’s an example, in the image below:

Recall the scenario that led to the development of Correx in the first place: ensuring that people who’ve shared stories that contain errors or need major updates know about those corrections and/or updates. Mastodon accomplishes part of the process today. The platform has vast potential, and we can’t wait to see what its users do next.

The road forward, and lessons

Working on Correx has been a joy. We’ve had the good fortune to work with two superb software developers along the way, Ted Han and Caige Nichols, who led the technical side during different periods. We’re grateful to both of them.

We’re especially grateful to Craig Newmark Philanthropies, which funded the project in the first place. Craig has been a friend to journalism (and to me) in so many ways. While this was certainly a minor project in his philanthropy’s portfolio, it meant a lot to us – and we believe it has been good for journalism.

We’ve learned a lot from the experience. Not least is reinforcing our abiding belief that journalists need to be vastly more transparent about what they do, how they do it, why they do it, and where they’re coming from. Honest corrections are a key element of transparency, and we strongly believe that journalists still don’t do them well enough. But we hope we’ve had some impact in making the need for them even clearer, and that doing them right in the online environment is better for everyone involved.

Finally, we want to stress – again, and sorry to be a broken record on this – a lesson that journalists (and many others) persist in not learning: the unnecessary risks in entrusting our work to centralized platform companies run by eccentric (or malign) people. 

It’s a lot easier to say this than effectively do something about it, given the enormous power the platforms wield. But if journalists, and people like us, don’t work hard to free ourselves from others’ control, we will always be under their thumbs. The best time to have liberated journalism from platform control was before it happened. The next best time to start is right now.

Dan Gillmor is co-founder of the ASU News Co/Lab. You can find him on social media at  Mastodon.

*Twitter was renamed “X” by its owner after we ended our attempts to work with its API, so we are referring to the site as it was named at the time.

Image credits: 

Dead Twitter bird via pixelant on Flickr

Valley News correction via “Regret the Error”