I really enjoy journalism.
I studied it for four years with the aim of beginning the craft on graduation. Then decided I was so interested in it, I did a masters to study how journalism worked in conflict situations.
I like it for its broad-church cover for reporting on warzones, snarking at political leaders, exposing rancid civil and social situations and telling the world what Cam did on last night’s #LoveIsland (You will not BELIEVE it…).
I’ve watched journalism suffer over the last number of years from a lack of investment, from manipulation in some quarters by the public relations blob (don’t @ me about how it’s #notallPR, I know that, but you’re crazy if you try and say there’s not some parts of PR that don’t do the rest of the industry or the media any favours) and from bizarre but cynically understandable efforts to erode its position of trust in the minds of the people by weak regimes.
So seeing a relatively high profile announcement of investment into a Press Association fund for developing technology which enables computers to write circa. 30,000 stories a month for local media is a mixed bag. Automating reporting, alongside a small team of editors and content guides, will certainly help get stories out there. It’s important that people get information and combining seriously impressive technology, capable of some degree of reasoning and analysis to produce relevant, timely and interesting stories that can then be curated and polished by a human team is, in essence, good from that respect. Combining the PA wire feed with proper reporting from a locally connected team could deliver an interesting business model for local publishers to survive.
However, there remains serious limitations to what an automated reporting system can achieve. Any AI or machine learning system is only as good as the data that feeds it. Streams of press releases, match reports, meeting minutes and other mundane items can be fed in and news nibs created to fill space – no more will a local football game go unreported, no more will a press release about a new chain restaurant coming to town go regurgitated. But where’s the data stream that will tell a story about backhanders in a pub, where’s the feed to describe the web of links to industry a local MP has created when a new car suddenly appears next to their massive house extension?
The regular news items, the basic reports in a local paper are where cub reporters learn what news is. Will an AI actually ever understand what Galtung and Ruge meant? These are the reporters who grow to be the editors, who shape the media. Will a subscription to this service at a local level mean publishers feel they can cut the cost of a junior?
In the hot takes about Google investing good money in journalism, there needs to be recognition of many other factors, not least:
- What’s the long-term impact this production (30,000 stories that will not be written by cub reporters) will have on news media long term
- How can we ensure the data and input into this system is not manipulated unfairly to begin with, so the editors have quality material to work with at the other end
- Is this really helping to produce local journalism that serves the actual aim of local communities?
Dr. Neil Thurman at City, University of London and the University of Munich nails it with an observation to the BBC of "I find it difficult to see how automation is going to help provide additional coverage of local magistrates courts and crown courts."
Other commentators within journalism itself have much better perspectives than I on this (look at the linked tweets) and I'm looking forward to seeing further analysis on this. Through it all though, it feels like this investment figure is a sexy way of glossing over the serious fundamental dangers in modern journalism and needs to be treated very carefully.
And ultimately, it’s worth remembering that The Register also applied for money under this scheme for some investigative work with genuine public interest – unfortunately, unsuccessfully. Read the application and guess why.
"You can't really cover [local government] through automation because it's a lot about investigation, politics, personal relationships, who has said what to whom and so forth - it's difficult to get that information in data feed form," he said.