Posted by: structureofnews | October 25, 2017

Evolving Structured Journalism

GuardianThe Guardian came up with an interesting way to rethink the article the other day: “Smarticles.”

OK, so I don’t love the name, but the idea is a good one – a machine-assisted evolution of the late, lamented Circa’s model of personalized news that incorporates ideas from the also late and lamented Google Living Stories idea, and some notions of structured journalism.  It’ll be interesting to watch.

The Guardian explains it this way:

Here’s how Smarticles work: On a reader’s first visit to a story page, the blocks contain the most basic details and background, as text, embedded video, photos or social media posts.

As the reader returns to the page, the Smarticle will attempt to algorithmically determine which elements of the story are most useful to the reader based on a number of signals, such as the time passed since last visit, the frequency of visits, and the importance of each story development.

In other words, the story figures out what you already know, doesn’t bore you with those facts, and only presents you with the new stuff.  In many ways, it’s a simple and obvious enhancement – and at the same time also a huge, fundamental change in thinking about what we do and how we do it. 

For example, news organizations normally produce several articles about an ongoing story — one that takes several days to unfold, such as the incident earlier this year when a man was violently dragged off a United Airlines flight after he boarded in Chicago. Readers who had followed that story’s every twist and turn would not need to be reminded of the fundamentals (that a man was pulled off a plane at a Chicago airport) that are often included with each article that covers the story’s developments. Instead, they would have been likely to skim through that background to find the newest information.

Other readers, though, who first encountered the United story a few days in, would need the background and context that a later article may have lacked. The underlying Smarticle algorithm hopes to differentiate between these two readers and present them with a story that meets their needs.

And that last sentence is really at the crux of both this idea and a core notion of structured journalism – that readers aren’t alike or all have similar needs, and that we increasingly have the capability to present them information that serves them better rather than the one-size-fits-all notion of story that we’ve lived with for… well, forever.

(That’s not to say we don’t need beautifully crafted narratives, or that breaking news has to be personalized for every reader – but that we can and should do better in understanding what readers (or at least classes of readers) need and want as individuals or discrete audiences.  For example: Hurricane coming?  Write the main story, but then attach a paragraph giving readers personalized information about how it’ll affect their neighborhood, based on each person’s home delivery addresses.)

Smarticles (the name!) aims to fix one part of the personalized story problem – prior knowledge – that both Circa and Google Living Stories also tried to address. Will it be more successful?

One thing going in its favor is – presumably – better technology and hence less dependence on humans to figure out how to break up stories into chunks of content (described as “blocks” in the Guardian post) and which ones to present to which readers.

But there are probably two other key issues that will arise as it rolls out.  The first is simply the underlying structure of information in the articles, and how easily – or not – it will be to embed that into a newsroom’s workflow and technology.  Newsroom content management systems aren’t usually designed to handle stories as structured blocks of information, and newsroom culture – built around crafting great narratives, for good reason – isn’t generally conducive to reporters filing into highly structured formats.  So getting past those two hurdles will be essential.  Circa managed it by being a start-up, and building and hiring from scratch; the Guardian has much more legacy infrastructure and culture. (And Politifact, Homicide Watch and Connected China – my three poster children for structured journalism projects – all lived outside established newsrooms.)

(That said, it’s also possible that language processing and generating technology will advance to the point where we won’t need to make changes in the CMS or ask journalists to do anything differently – but right now that doesn’t seem to be the case.)

Just as important will be how much value readers place on the service.  There was a long debate about the causes of Circa’s demise, to which I added my two cents.  But probably the most critical question is whether this format serves an unmet need; or rather, whether the tradeoffs inherent in creating this format outweigh the benefits of the unmet need, as I mentioned in my piece on Circa’s passing:

So how does or should structured journalism compete?

The same way machine-generated sports stories compete: By finding an unoccupied field to play in. The likes of Narrative Science and Automated Insights don’t try to outdo human sports writers covering the Super Bowl; they generate stories about college and high school games that no one covered before.  Are their stories worse than ones written by a human?  Let’s stipulate yes.  Are they better than no story at all? Absolutely yes.

So, too, Smarticles (ugh!) will succeed if enough people find its functionality more useful than the loss of some narrative cohesion.  However it turns out, it’ll be another step forward in the evolution of structured journalism and its path into the mainstream of news.


  1. Thanks for the thoughtful consideration of the format (I won’t repeat the name, since I get the impression you’re not a fan). The hurricane example is a great one, and I agree that the big question is whether users want and need it (initial survey responses were pretty positive, but suggest a mix of views).

    A few things to note about this project: it’s a product of the Guardian Mobile Innovation Lab, not The Guardian as a whole. We’re part of the Guardian US newsroom in New York but operate fairly independently of the goals the main organization sets. Our work is meant to poke at the edges of what’s possible with storytelling on mobile and offer up examples and possibilities for reaching audiences on mobile, in order to benefit the rest of the industry. We’re going to keep experimenting with S********s — and please do send suggestions of other names if you have them — but the format won’t be rolled out broadly within the Guardian site or apps unless the core teams decide to make it a priority — and at this early date in our experimentation, we wouldn’t advocate that they do so just yet. Also, by sharing the data and insights we glean from our experiments, which we’ll do, we also make it possible for others in the field to learn from it. Lots more on Medium if you’re interested!

    • Sasha, hi, and apologies for not being a fan of the name, even if of the project. (Or at least, thanks for playing along with me on that.)

      It’s great that you have some very clear goals for the project – including not necessarily pushing it into the core newsroom. One of the biggest challenges to structured journalism projects has been newsroom culture and the definition of what it means to be a reporter; although there are also some signs of progress on that front too.

      I’d love to hear what lessons you learn from this project, and what kinds of new formats of news users/readers are willing to try/embrace.


  2. One way I’ve thought about this dichotomy of needs, especially in breaking news, is looking at Twitter and Wikipedia as complementary ways to follow a story. Wikipedia’s job is to tell you the state of the world, as it is right now. Twitter alerts you that something has changed. Wikipedia is the broad overview; Twitter is small-bore, fast and incremental.

    (Obviously there are problems with both as news devices, but I’m mostly thinking about needs here.)

    I’m really curious how this experiment works out, and whether the personalization algorithm holds up.

    I’d also be really interested to see how this works on a story that evolves over months, or longer, instead of days, and whether user needs change as a story ages.

    • Chris,

      That’s a good way to frame it, and it made me think a bit more about the various ways we can bucket how we follow stories/the kind of information we want or need.

      So there’s your analogy of WIkipedia as telling you the state of the world and providing context; Twitter as alerting you to changes in that state (news); but then also adjusting that by current reader activity – standing in line, walking, at rest in front of a screen, running, at a meeting, etc. Then there’s various types of personalization – by location (hurricane will go through your neighborhood); by interests (stock portfolio, family members, children’s school, etc); There can also be the creation of counterfactuals – what were the key variables that led to this outcome? There’s probably a subset of the Wikipedia/Twitter analogy too, which is about personalization in time – are you reading the story in real-time, shortly after an event takes place, or long after the fact?

      Your question about how this works with a story that runs over a longer period is also an interesting one – perhaps this works best not on a running news story, but on something that moves very slowly, such as climate change?


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s


%d bloggers like this: