Why journalism expert Charlie Beckett doesn’t fear AI in content creation: Part 1

03.03.2020 · By Farah Khalique

The world of storytelling is changing rapidly, with the advent of new journalism tools powered by machine learning and artificial intelligence (AI). We spoke to Professor Charlie Beckett from the London School of Economics about these tools for newsgathering and how technology is changing the newsroom.

Professor Charlie Beckett is the founding director of Polis, the think-tank for research and debate around international journalism and society at the LSE. He led the influential Polis ‘Journalism and AI’ report, which you can read more about here.

The Polis report paints a picture of how 71 news organisations from 32 countries are using AI in content creation for time-intensive tasks like news gathering, production and distribution to transform journalism.

This interview has been edited for length and clarity.

Loyal AI: When did you first start to explore the use of AI in journalism and was there a Eureka moment? 

Charlie Beckett: The Eureka moment was doing the report! It was clear to me that this is different in lots of ways. One is the hidden invisible complexity of it, the algorithms. The second bit is the way it interrelates with so many different aspects of journalism. This includes people’s consumption through personalisation, content creation, the idea of going through data sets with machine learning. It also touches on newsgathering, finding out what stories you should be doing, what is going on in the world that we should know about.

You can use data on what people are consuming to decide what stories to do in the future. That will help you to do better journalism and sell better. Social media made us think like this as well. The traditional divides between the business side, like the marketing and subscription departments and the journalists has in some sense been blurred. They should at least be talking to each other. But that is also another ethical issue.

If journalists don’t do anything about AI, the news will still change. We will be consuming information differently, getting it through Google’s news aggregators, Alexa and Voice on Demand, so even if newsrooms don’t change anything you as a news consumer will be affected by these technologies. 

When we think about the idea of news changing – if journalists don’t do anything about AI, the news will still change. We will be consuming information differently, getting it through Google’s news aggregators, Alexa and Voice on Demand. So, even if newsrooms don’t change anything you as a news consumer will be affected by these technologies.

The task isn’t just to get access to these tools and learn how to do machine learning; it’s to understand that your whole business is changing. If I make motorcars and I’m told that petrol and diesel will be phased out, you think okay, I had better make a different car!

It’s been three months since you put out the Polis report on artificial intelligence in journalism. What has the feedback been like since? Has it sparked change?

It’s been really good. It’s quite unusual, normally you publish reports like this and they kind of disappear onto people’s shelves. But for this one, the enthusiasm and interest of people was very strong. This is an issue that is very hot and is going to develop. People can see it’s complicated with all sorts of hazards but lots of opportunities.

There was a lot of interest off the back of the report to the extent that we will do another year of work. We will create a network of people around this, with more training, events and experimentation around AI in journalism. It’s been a very positive engagement. It was already a big sample of people but it’s now spreading to other countries. People from other areas of journalism have been in touch as well as people from other areas of media and AI. It is definitely something that is going to grow.

As part of your research, how did you feel that news organisations view AI? Do they get it?

We were mainly talking to digital-savvy people. Even if they weren’t 100% doing AI, they knew they should be. But there is a whole swathe of people that don’t want to know. They say, “My job is to edit the newspaper, not work out how we do it.”

There is some sensible realism there as well. Yes, it can do a lot for you but maybe only 20% or 40% of your budget and content is impacted. It’s not about flicking a switch, like just going online with a website. It’s about thinking through the whole production model and perhaps business model that is quite challenging and shouldn’t be done overnight.

Do they get it? That’s always a bit of a cheeky one. Either you get it or you don’t. It’s not that you should either believe in AI or you don’t, it’s there but it’s very complicated and hyped up by people that want to make money out of it. Google stands to make a lot of money out of it, for example.

Important news organisations need to think about what they were put on earth to do. For example, to create very good political coverage to help people make democratic decisions. Will this help that?

It is important that journalists ask those questions of the technology and don’t just rush into it. I find it difficult to explain how to train data, I can’t do it. If you asked me to analyse a spreadsheet of bank accounts, I couldn’t do it. It’s complicated. But I worked in television for 20 years and don’t know how to work a camera!

One thing that probably is true is that artificial intelligence is going to mean more work for you if you’re a journalist. We know that from previous new technologies like social media. These technologies can help you save time. I can use a search function with a phone and I don’t need to physically visit a cuttings library. It’s more efficient but it also means a lot more labour. I have to produce more content for more platforms. I have to learn about those platforms because each one is different.

It’s the same for AI. Journalists have to learn how to edit the algorithms to see whether those algorithms are giving trustworthy information and content. That inevitably means more work. It’s also exciting, especially with AI, that there is the potential to get rid of a lot of boring labour.

It doesn’t mean we can all go sit on a beach. It will create opportunities that will involve working harder. One of the biggest questions is around quality. Have we got the skill, creativity and enthusiasm to do the extra work that makes the difference?

Journalism AI meeting
Journalism AI meeting, in Google France headquarter, Paris. On 17th of May, 2019.

You preface the report with “No, the robots are not going to take over journalism.” So let’s dive into where AI is already re-shaping newsrooms. Just under half of those surveyed for the report use AI for newsgathering and two-thirds for production. Tell me how AI is changing the way journalists gather news and how it is produced.

I think in some ways it’s very specific and tangible. Bloomberg said 40% of their content now is automated. There is always someone checking but pretty much the robots are creating this content. That is a direct change. If you are an investigative journalist, you can see quite tangibly how AI can go through thousands of accounts and flag up interesting aspects.

The audience data now automatically generated that you get on your dashboard, like Chartbeat, might help you and influence your journalism. You can see how one headline works better than an other one. You can either say “that’s terrible clickbaity headlines” or “yeah, that’s helping me”. So it really varies according to the type of journalism and what you want to achieve with it. For some people, it will be invisible. Some people won’t even notice that AI is helping them. While for others, it will be very much a tool they are consciously deploying.

This idea of journalists being replaced by computers is kind of scary! Journalism professor at Northwestern, Nicholas Diakopoulous, argues in his book Automating the News: How Algorithms Are Rewriting the Media that AI will actually create jobs. What are your thoughts on that?

First of all, you’ve got to put it against context that journalism is under enormous economic pressure. Financially, it just doesn’t make sense. Too many journalists are creating too much news that is similar. 100 people around the world covering the Democrat caucuses story in Iowa are all telling me the same story pretty much. Once you’ve read two articles about that the rest can seem like a lot of knee-jerk reaction pieces.

But the basic facts are there. 1,000 outlets are giving people the same facts. There is too much journalism. Chris Moran at The Guardian said that newsrooms do fewer stories but more on those stories. This is because they know they’ve got to maximise the attention and value they can add to a particular issue. This is so they can really prove to their subscribers they are doing something different to all those other people.

journalists’ labour is changing. There is way less premium on the routine stuff – I used to do that, type out football scores! – and more towards added value. AI will take away a lot of that labour, and create jobs for managing these systems.

I think that means journalists’ labour is changing. There is way less premium on the routine stuff. I used to do that, type out football scores! And more towards added value. AI will take away a lot of that labour, and create jobs for managing these systems. We are already seeing new job titles like Algorithms Editor. This is someone who looks at algorithms to make sure it is not biased and works properly. Someone has to edit this technology. There is no pure AI that can run themselves and come up with brilliant story ideas.

The other more idealistic bit is that all these people are telling me that human creativity and curiosity and inspiration and passion are only thing humans do. Robots can’t do it. My reply is: “Go and do it then.” That will be an added value. In a world where algorithms give us more automated news, there will be a market advantage in creating more interesting journalism.

If you are the person that did the routine stuff like type up weather reports, you’re screwed. Think of an alternative career. You can either re-train or some 25-year old coder will come in, or you’ve got great specialist knowledge so you can add value. It’s a mistake to think of one person sitting there thinking are they gonna get sacked or not, their [entire] world will change. We saw that with social media. It’s a basic competence knowing how social media works but also a cultural attitude and intuitive enthusiasm for using those platforms.

Sometimes the ‘old dogs’ are great at learning these new tricks, they’ve got residual, traditional knowledge and look at this technology and think, “I can do something really clever with this”. They may even have an advantage over the 21-year that hasn’t covered a beat, but can just code.

We’ve seen a real resurgence in investigative journalism among both large organisations like The New York Times and even smaller outlets like TalkingPointsMemo. You re-tweeted journalist Jeremy Merrill’s tweet that Quartz used AI to read all 715,000 documents in the #LuandaLeaks. If AI is this powerful, does this democratise hardcore investigative journalism. Can anyone with the right piece of kit start investigating?

Theoretically yes, in practice not so much. There are certainly big ones like Panama Papers and Luanda Leaks, that was a question of co-ordinating internationally, which is always difficult. Collaboration with journalists is always tough, it can be logistically tricky.

It’s also not a flick of the switch stuff. You have to spend weeks programming the kit so that it asks the right questions before you press the button and the cogs start to whir. The data training and the algorithmic training can take weeks if not months before you’ve set up something.

Then you do the actual data analysis and say “Well it’s given me this stuff, these are the ten most important documents. Now we have to turn this into journalism.” It is interesting when you can create a data set over tax changes and users can put in information and get a personalised readout, but that all takes designing and oversight.

Even highly technical work like the Bellingcat investigations into plane crashes and Russian missile attacks is always a combination of technology to do the image search and analysis, but you then have to have the skill and expertise to interpret those results.

The 20th annual Edelman Trust Barometer came out. It shows that among the mass population – 83% of the global population – the media is not trusted, whereas it is among an informed public. Do you have any thoughts on how the media industry could use AI to bring back trust among the masses? How important is this? 

I think the Edelaman Trust Barometer can be a bit of a distraction. Trust is a contentious word.  The idea that we get to a place where everyone thinks journalism is great is not likely and not even desirable. I don’t want people to trust journalism implicitly, but to use it and think it’s credible. It’s a relationship you build. The AI can help on some of that, in terms of filtering out disinformation and making us more transparent, and more responsive to the public.

These are all ways of building a better relationship between news media and the public. But in the end, it is the human element; things like making sure if we use this technology that we recognise the inherent biases in some of these algorithms and that we tell the public when we’re using automation, for example, or personalisation.

But the evidence seems to be that people aren’t too bothered whether it’s a machine or a person that tells them. That slightly worries me. I would like journalists to be more emotionally literate and engaged about the news, and the public as well.

Newsrooms are becoming more diverse, more attentive to what’s going on in people’s lives and how they use the news, and why people want to consume our journalism. That’s a human strategy – journalism culture is changing. Technology won’t solve all problems.

Looking ahead over the next few years, what are your predictions for how AI will be used in content creation? Have we fulfilled even 10% of AI’s potential? 

It is incredibly difficult. If you read the report, even these digitally savvy people are quite realistic. They say, “We tried this it didn’t work/wasn’t quick/accurate enough, involved too many people” etc… so one has to be understand the limits of this technology and journalism.

I think it’s also difficult to quantify this, we can see the % figure of online versus newspapers, we could see 60% of the audience is now online and 40% is still buying papers. But with AI it’s much harder to quantify how big the impact will be. Your subscriptions may go up. The Times for example, was using systems that were definitely getting better returns on their subscribers.

But it’s harder to say that was caused by AI. For some work like investigations, it’s more obvious. A long time ago, I asked my former managing editor at Channel 4 news “How do you quantify the efficiencies of digitalization?” and he did come up with a figure: about 40% more efficient. Which meant either they sacked 40% of the people or are doing 40% new work.

I think in three to five years, newsrooms that integrate this in a kind of systematic way will see that order of benefit. I’m not saying 40% extra profit, but 40% of work will influenced by AI. If managed well, it will transfer costs or investment from certain areas into more productive areas.

Find out how LOYAL AI’s editorial research assistants can help you research faster with quick access to insights relevant to your article as you write in real-times, all without leaving your creative space.