Computer-Generated News

By | March 18, 2014

It seems that writing articles is on it's way out. The replacement is computer-generated articles, and they are surprisingly readable. Much more so than Google Translate. The one difference, it seems, is storytelling. Software isn't good at telling stories yet – though this may just be a question of time.

I'd concur with +Thomas Baekdal on his thoughts on the upcoming revolution that is already transforming journalism today. But I think it will also transform more than just journalism. It will transform sensemaking. Human ability to analyse and make sense out of data will need to be done on an increasingly abstract level: today, analysis and synthesis is still heavily human-operated. But this does not have to be the case, and as algorithm accuracy starts to cross the 80% and 90% thresholds for identifying and handling literary devices and emotions correctly, we should start to see some close-to-sentient analysis and story production.

And sure, when we move to that state, we will still need humans to do some of that work. But many things can be accomplished without human intervention at all. Indeed, superfluous workers would likely be the status quo. This breaks the economic model that we currently live in – and that will be another society-rocker change. How will society cope? What socioeconomic models will develop?

On a side note, I will bet that there is no large difference on computer-generated general news articles and computer-generated technical articles, for those journalists who think that specializing in a particular technical vocabulary brings a selling point.

Reshared post from +Thomas Baekdal

Robots are taking over the role of reporting news, especially when it comes to the type of news that is based on merely being the bringers of news.

One example from yesterday was an article about the earthquake in LA, which was covered in this story, written entirely by a computer: … and compared to so many other stories, I actually find this to be one of the best. Especially if you compare it to articles like this one from the Associated Press, which is filled with irrelevant fluff:

What possible value does it have to send out a journalist to get eyewitness reports like this: "My dog got out of bed, and she came looking for me," Smith said. "She was shivering terribly."

But how good is automatic software-based reporting really? The answer is very, very good… at least according to a new study from Sweden.

Christer Clerwall, from Karlstad University in Sweden decided to do a 'blind test', asking students to compare news reports done by both journalists and software algorithms. The result was that the software algorithms are now just as good at writing news reports as people.

As you can see from the graphs below, when it came to how the story was perceived, journalists and software algorithms were nearly identical. The only slight difference being that journalistic-written articles appeared slightly better written, although less useful, trustworthy, and informative. 

In other words, journalists are better storytellers, but their stories are just not that good.

Even more interesting is when you he asked them to guess who had written the article. 56% responded that they thought the articles written by journalists were actually written by a computer, while 37% thought journalists had written those made by a computer.

In other words, in terms of simply being the bringer of news, journalists have now been outcompeted by software. People can't tell the difference, and, as we all know, software algorithms can work much faster, more accurate (if the data is accurate of course), and 24-7 at almost zero cost. 

The old business model of being a reporter and to 'bring the news' is over. This will in the future be done completely automatically, and it's only a matter of time before the big tech companies move in. The next time there is an earthquake, Google Now, Apple Siri and Microsoft Cortana will simply read aloud the story like the one done automatically by LA Times. Several startups will try to capitalize on this by creating data-news content services. And while many of these will not succeed, the best ones will eventually be incorporated into other services.

In many ways, what is happening to journalists is the same that once happened to the fashion industry. As you may know I started my career in fashion, and the first company I worked for employed about 60 people. There were 2 sales people, 2 designers, 4 administrative staff, 3 technicians, a janitor, and about 50 seamstresses. During the time I was there, those 50 seamstresses lost their job, because the production was outsourced to Poland (and later China).

The people in Poland could do the same job, at a higher quality, with fewer resources, and much, much cheaper. A company with 60 employees could now be run using only eight – and still make the same product.

This is what's happening to journalism. 

So the question all journalists should ask themselves is "are you a seamstress? …or are you a designer?" If you are merely a seamstress, i.e. a journalist merely reporting stories based on available information, your future looks bleak. 

But if you are a designer, i.e. a journalist who goes beyond just reporting the news, but instead focus on providing insight, perspective, analysis, then your future is looking better than ever. In a world of data, we need people to give us perspective and to help us translate it into something that is relevant for us as individuals.

And no, this doesn't mean interviewing eyewitnesses that have nothing insightful to tell, nor does it evolve interviewing experts or pundits who merely conjecture without having any real insight. It's about being better than the news itself.

Sadly though, many newspapers are currently moving in the wrong direction with their pageview-optimized news reporting. They are trying to build traffic by bringing people even more news, not realizing that that's exactly the what automatic algorithms do best.

9 thoughts on “Computer-Generated News

  1. Frank Nestel

    +Sophie Wrobel German newspapers very much missed to earn money online. Thus online has molten away the market and marketing income for them. Newspapers needed to shrink work force to stay affordable.

    Of course there are great sources of online journalism, but creating news consists of various stages before the writing and the infrastructur for obtaining information also suffers from theses shrinkages. It is much more likely today, that you get the very same information from different newspapers.

    In summary, I think quality has degraded, quantity of course is higher than ever.

  2. Sophie Wrobel

    +Frank Nestel Do you think just the quality has dropped because of staffing issues, or because of the education / demand of the mainstream market? I mean, Bild still sells so well, but FAZ is not exactly the first place to turn to for news anymore… perhaps societal expectations have turned to different media, and the expectations for newspapers (as in the role they play in bringing "news" and what sort of "news" they bring) has changed instead? It would mean that while the level of quality in newspaper journalism drops, the level of quality in other journalism (online, or even technical writing) has maintained the 'old' standard?

  3. Tihomir Stoev

    +Frank Nestel for me personally content matters, a machine have difficulties doing headlines('a man cuts his head off and walks away','aliens kidnapped Elvis Presly'…bla bla). 
    and I find this extremely good. Today newspapers are not bringing the news like a couple of decades ago, just headlines…

  4. Frank Nestel

    Well, in Germany newspapers are now operates with less than half the staff than 15 years ago. It is observable, that the quality of human "generated" articles in newspapers has dropped a lot in that time. Now that we have degraded journalists to piece-work text production machines, it is not so surprising, that other machines can compete.

  5. Robert Pratt

    I'm wondering if the robots will take the time to verify facts.

    Automatic stock trading has crashed the market a few times….so

  6. Daniel Sprouse

    there are a whole lot of people writing articles who should be doing almost anything else; that is going to skew the numbers a bit.


Leave a Reply

Your email address will not be published.