Alma Media > Releases > Releases > Living information > 2016 > News produced by algorithms

News produced by algorithms

News articles made by robots are just around the corner in Finland, too. Imagining it, a publisher may see euro symbols, a journalist a horror scenario. It will be easier to form an opinion on the matter if you know what it is about.

“At its purest, automated news production means a process that creates news articles on the basis of algorithms without human intervention. And then there are all sorts of variations,” says Carl-Gustav Lindén who has studied the topic. A lecturer at the Swedish School of Social Science (University of Helsinki), Lindén received a grant from C. V. Åkerlund Foundation in 2014 for the research of the robot journalism phenomenon and its impact in the United States. 


Carl-Gustav Lindén

“There is the search filter type of algorithm, of which Google Search is the most advanced example. Then there are news alerts: a search engine can go through material such as records of police investigations and give an alert if a celebrity has been caught for something – but it’s still a human who actually writes the article. This isn’t done in Finland yet, as far as I know, but the Swedish tabloid Aftonbladet is using news alerts. And there are various ways to combine automatically filtered data.”

First steps in Finland, the USA far ahead

In order to fully employ algorithms, the data used has to be extremely structured. This will easily create a bottleneck as data is rarely available in the required format.

Sports and financial data typically meet the requirements. In the USA, interim reports have to follow a certain format with standard layout of figures. Therefore they are created automatically, and the data they include can also be used with algorithms.

This is not the case in Finland or elsewhere in the EU, however, where decisions on the standardisation of financial data are still underway. Finnish newspapers such as business paper Kauppalehti cannot really automate their reporting period news even if they would be motivated to do so. In Denmark, Ritzau Finans has had to develop their own solutions to be able to produce automated business news.

Sports results data is more or less the only form of data that meets the standard requirements in Finland, but according to Lindén it has not so far been used for journalism.

“Weather news are produced automatically, which anyone can see on their mobile phones. Finland is genuinely a model country of open data; you just have to find the interesting data and examine if it can be combined with some other data.”

In the USA, automated news articles are already produced on a large scale. Associated Press (AP) publishes automatically financial news in particular.

“There are two companies that have created a new market, in a way. Formerly owned and then sold for big bucks by AP, Automated Insights now produces corporate sales summaries for AP and others. And Narrative Science specialises in sports results,” Lindén says.

Similar companies exist in Sweden, Germany, France and China, he says – and language is no barrier when texts are simple.

Letting go of routines or jobs?

Lindén himself is an experienced journalist. Future in the world of automated news doesn’t sound scary to him?

“Isn’t it good if the machine does the boring, time-consuming routine stuff, and we can focus on something sensible? If jobs are lost, they will be lost anyway. But automation can in fact save jobs as it can help publishers maintain the necessary news volumes with lower costs. In the best-case scenario, this will create more resources to concentrate on in-depth news coverage.”

This has happened at AP that had already cuts lots of jobs. The remaining staff wrote 300 articles per reporting period, most of them routine news about financial results. Now robots create 3,000 articles per reporting period, while journalists can focus on a few major corporations that have to be analysed manually. But they are happy.

“I’ve also heard about an American company, though, that buys local papers, automates everything and sacks most of the staff. But I don’t think this one will be a long-term player as one gets bored quite soon when reading lots of robo-journalism.”

Demand for meta journalists

Lindén believes that journalists will surely be needed in future as well.  It is the journalist who chooses the angle, decides what is interesting and relevant and what people should know. This responsibility cannot be left to other professionals such as coders.

But the journalists’ work is changing.

“Certain journalists will have to know how to code at least a little, and they’ll have to be able to think of structures: how to build an information infrastructure. They are called meta journalists. And they already exist in the USA if not elsewhere.”

“If a journalist decides that the world of algorithms is outside their area of competence, they will be in trouble. But if they instead decide to seize the day and learn something new, it will open a whole lot of new opportunities workwise.”

Lindén emphasises, however, that transparency is vital in journalism. The reader will always have to know how the article was produced and the principles and data behind it. 

The power duo: man and machine

Lindén calls the combination of man and machine invincible. For a human being, it’s nearly impossible and at least frustrating to write dozens of variations of the same article, but for a machine it’s no problem.

“A machine can learn the reader’s way to use content and then edit the article accordingly. It can change pictures, rewrite the headings, try different colours – and learn to choose the best options for various targets. This can be really decisive when focusing at young readers.”

It’s already happening in advertising. In journalism, newspapers like the Guardian and the New York Times are trying it out in their online versions. But Lindén points out that the model is a big whole and the produced text is just a part of it. User analysis is at least equally important and also done automatically.

Development is incredibly fast and algorithms are becoming self-learning. Even now it’s often impossible to say if a text has been created by man or machine. And this applies not only journalism but also literature and poetry. But Lindén is fearless.

“Computers can also compose music and make art. But we still have composers and artists.”


Date: 24 February 2016
Theme: Journalism
Text: Diana Törnroos