Why journalism expert Charlie Beckett doesn’t fear AI in content creation: Part 2

31.03.2020 · By Farah Khalique

Professor Charlie Beckett, founding director of Polis, the think-tank for research and debate around international journalism and society at the LSE, led the influential Polis ‘Journalism and AI’ report.

For further insights from Charlie, read Why journalism expert Charlie Beckett doesn’t fear AI in content creation: Part 1.

He explains the use of artificial intelligence for content discovery, how journalists are using editorial workflow tools to source stories on the coronavirus and why journalists should think twice about the growing media trend of “informative reporting”.

This interview has been edited for length and clarity.

Hi Charlie, great to catch up with you again. Let’s pick up where we left off. How can editorial workflow tools like a social media research tool help writers find stories?

Charlie Beckett: There are different ways. One programme mentioned in the Polis report works whereby if you publish a story say on the Labour leadership election, it will then update you when other information becomes available. Say you publish a story, it’ll tell you if MP Lisa Nandy has done X, Y or Z or when someone else has tweeted a story. That helps you keep track of stories and issues.

Then there’s the audience data scraping – for example, Google trending tells you what people are talking about. There is some interesting work being done looking at where topics are in social media and other data sets. Months ago, people were talking about the “China virus” story that suddenly started to surface first on social media.

An interesting one mentioned in the report is whether it will change the news agenda. One of biggest problems with news is that what journalists thought was news wasn’t always what interested the public. Journalists get really obsessed by certain stories and issues. Research tools for journalists can help us understand public interest better [to bridge] that kind of “news gap”.

On statistics, it will help us learn about trends so, for example, about diseases like the flu. Artificial intelligence is able to tell us, often in real-time, if a story has developed or if more cases are emerging. That will be interesting to see, though it is emergent at the moment.

Right now, coronavirus has taken over the news agenda. How are journalism tools helping writers create content around COVID-19?

This is a huge challenge, because there is so much data and a lot to process in terms of how this virus is impacting the UK. How do you process data from different sources, experts, institutions, universities, and so on? I imagine if you’re a journalist, like everyone else you feel overwhelmed by the story. This is the perfect storm in terms of information overload for journalists, never mind anyone else. Everyone is scrambling to get on top of the story.

We see people deploying technology to try and cope with it. This is where machine learning (ML) and data analysis stories can help. Journalists at the Financial Times (FT) have been doing some brilliant data visualisations of the John Hopkins data on a number of cases and deaths that uses some ML technology. A lot of newsrooms are using news aggregations services where the ML is gathering all the different sources and prioritising them, so journalists don’t have to do all that themselves.

There is some really interesting interactive use of bots to help people ask news organisations about what’s going on. Obviously there’s a lot of personalisation around things like newsletters; your notifications use some of these technologies too.

The rise of coronavirus has fuelled a rise in sensationalist news. How can journalists use journalism tools to report instead on solutions rather than add to the scaremongering?

FullFact – the UK’s independent fact-checking organisation, social media platforms like Facebook and even closed platforms like Whatsapp are trying to bring in new ways of fact checking. It is up to platforms to do a lot of the filtering, but news organisations are working with them. Reuters is working with Facebook on a project to counter visual fakes.

But it’s really difficult. There were articles recently about a perfectly credible Oxford University study that said coronavirus may have infected half the UK population. I said at the time that it’s a modelling exercise, i.e. a hypothesis based on mathematics, not evidence. There is no substitute for journalists being very sceptical; this is deeply complex stuff.

I also tweeted an LSE blog on how to read an academic paper. If you were to read an academic paper, i.e. questioning the methodology, looking at sample sizes, the provenance of people in the study ­– it takes a whole day. You don’t just skip to the conclusion. That’s a real technical expertise, so journalists have to proceed with enormous caution and go to other experts.

Serious media outlets have done that. What we’re witnessing are the kind of tech experts who are good at data going onto Medium, dragging a load of charts together and processing them, but they don’t always have the deep scientific knowledge to put things into context. That is making it very difficult for journalists to be able to say categorically “this bit of information is better than the other one.”

It may mean not publishing at all or publishing content with huge qualifications. Generally, news organisations have been reasonably good at that. Most of the bogus viral stories about curing the virus by drinking tea have come from social media and amateur experts.

Catherine Glydensted, director of constructive journalism at training platform, VersPers in Denmark, described this as a “transformative time that will change the media industry”. She pointed out that Scandinavian newsrooms are now moving towards “informative reporting”, as journalists ask what citizens really need. What are the benefits for our industry to gravitate towards informative reporting, and how can editorial workflow tools assist in that?

Catherine has worked on ideas of solutions-based journalism or constructive journalism and, in a way, it’s common sense. When it comes to coronavirus, the role of the news media is not to say “We’re all gonna die”, rather here are five things you need to do everyday. We’re seeing brilliant stuff: non-stop advice from news organisations on constructive solutions, steps you can take and advice etc…

The data visualisation (DV) tools have been extraordinary, like the New York Times map showing the spread of the virus. Some are almost like GIFs that show how the spread happens. The role of podcasting has been really important in reaching demographics that wouldn’t normally consume news. Automated voice tools for articles means you can click on an article and it’s read out for you. Stuff like that is really important. We are blessed that we can use everything including TikTok to reach people who wouldn’t normally pay attention to the 10pm news or The Guardian.

Where I would question it though, is in China where the media has a completely different role and purpose to us. They are literally employees of the state – their job is to carry out state policy. The media in China was on message, it was completely behind the lockdown strategy which, in some senses, seemed to have worked.

You have to be careful when doing this though – Catherine would probably agree – as there is a difference between advocacy journalism and constructive journalism. The idea of constructive journalism is good – we shouldn’t just be running negative scare stories – but the danger of it shifting to journalists going beyond their competence is real.

Even if productivity tools for journalists can help with stories and take over menial tasks, they won’t bring back advertising revenues. How can newsrooms, particularly smaller ones like local papers, exploit AI to help not just survive but thrive?

Often at that local level it can be most effective, even on the advertising side, to have a proper database of your local economy and much more tailored advertising for smaller companies that can target smaller geographical areas.

Even for adverts, these technologies offer interesting opportunities but a lot of local journalism is quite routine, like school results, football results and planning applications. There’s a lot of grassroots-type stuff that local news organisations should be interested in, but they are badly staffed so there is no time for people to physically go out and get this stuff.

You can hope, if we develop prototypes, that that will improve. Radar (Reporters and Data and Robots) is a news agency that looks at big data sets to create local stories. The Bureau of International Journalism had Bureau Local, which looked at big data sets on homelessness, then worked with local newsrooms to make it relevant to particular towns or areas. I think there is potential there but it is definitely a harder find for those news organisations.

There is a real danger of inequality between small newsrooms and bigger ones, but also newsrooms in smaller markets like Lithuania where they don’t have the economies of scale to invest and scale up. We partly need to see a growth in collaboration and intermediate organisations that provide content discovery tools and services to help enable those organisations to do it.

You can talk in the report about the spread of AI and the Western bias towards richer, more technologically advanced organisations. What are you doing to work on filling those “big gaps”, and what’s the timeline?

We’re hoping with our network that it is already happening, it’s a kind of dating agency. People say I read in your report about that groovy thing Le Monde is doing so we put you in touch with them. You can ask how did they strategise, where get their kit from, how many people worked on it, the best way of taking it forward, what results to expect.

The pleasing thing to come out of this process is that people can see the benefits of knowledge sharing. They see instead of developing something and keeping it secret, there is value in sharing their experiences. We hope that will reduce some of the inequalities, it’s better than paying some fancy consultant to talk to someone at the New York Times or El Pais and see what they did.

We show how people can knowledge share on an artificial intelligence project to create something interesting. The other big collaboration is with universities. There is nothing to stop you making friends with the local artificial intelligence department of your university; they are very interested in deploying knowledge in a very practical way. They have the technological knowledge. There is no reason why they can’t feed that down the chain.

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.