Few Britons think the positives would outweigh the negatives – although there are some tasks they tend to think it would be ok to allow AI to do
Artificial intelligence is coming for the newsroom. German newspaper Bild – the highest selling tabloid in Europe – has said that it expects to make cuts to its editorial staff “due to the opportunities of artificial intelligence”.
Here in the UK, the publisher of the Mirror and the Express has said it is exploring ways in which ChatGPT can be used to help human reporters compiling coverage of topics such as local weather and traffic.
Such a scheme is already in operation in Australia, with News Corp’s Michael Miller telling the World News Media Congress last year that the organisation is producing 3,000 local news stories a week on weather, fuel prices and traffic conditions. Tech news website CNET has come under fire implementing the same model, with simple financial articles producing multiple errors.
To date, few Britons – just 7% – believe they have at some point read a news article that was written by an AI. But the expectation among the public is that the introduction of AI to journalism is going to do more harm than good.
Approximately half of Britons (48%) think the impact of AI will be more negative than positive, including 12% who think no good will come from AI in journalism whatsoever.
By contrast, only 6% think it will do more good than harm, and a statistical 0% say the impact of AI would be entirely positive with no drawbacks. A quarter (23%) think AI in journalism will do bad and good equally.
It is unsurprising, therefore, that the vast majority of Britons (79%) say that media organisations should be required to display on a news article any ways in which that article has been created using the assistance of AI – something which Buzzfeed has implemented for its AI-assisted travel content.
Younger Britons are, however, slightly less likely to be pessimistic about the prospects for AI in journalism than their elders. While 54% of the over-65s think AI in newsrooms will have more drawbacks than benefits, this falls to 32% among 18-24 year olds. And while a mere 3% of older Britons expect more positives than negatives, this stands at 17% among the youngest adults.
Similarly, while 86% of the over-65s want news outlets to have to signpost ways in which AI has been used to contribute to an article, this falls to 63% of 18-24 year olds.
To what extent does AI journalism impact trust in the credibility of a news article?
With strong scepticism towards artificial intelligence in journalism, we sought to quantify quite how much the technology would affect trust in the news.
To test this, we created a randomized test, based on the following question:
To what extent would you trust an article on [randomly-selected news outlet’s] website that had been written by [random selection of: a human journalist / an AI journalist], and had been overseen by [random selection of: a human editor / an AI editor]?
Across all combinations respondents saw in which both the journalist and editor in question were human, 48% said they would have a great deal or fair amount of trust in the article. This is compared to 39% who would have little to no trust in the article.
However, for those combinations where one of these two roles was conducted by an AI, trust levels fall to just a quarter – 25% where the AI is the journalist, and 23% where it is the editor. In both cases, 60% of Britons say they would distrust an article created in such a way.
And where artificial intelligence is fulfilling the roles of both the journalist and the editor, trust slumps to just 13%. Fully 72% of Britons say they would distrust an article created by this AI duet, including 39% who would not trust it “at all”.
Breaking down the results by the category of randomly-selected news outlet* that was included in the question shows the same trend, with a human journalist-editor combination being the most trusted, a single human-AI combination being less trusted (by about the same rate whichever job is given over to the AI), and the double-AI combination being least trusted.
The main difference is simply the greater level of trust in news on broadcast and ‘up-market’ newspaper websites compared to mid-market and tabloid newspaper websites.
What journalism tasks is it acceptable for AI to do?
While the public tends to have negative expectations for artificial intelligence in journalism, that is not to say that they don’t see any ways in which it can be appropriately implemented.
From a list of 15 potential uses for artificial intelligence in the newsroom, two are considered acceptable by large majorities of the public: performing spelling and grammar edits (73%), and translating articles to other languages (67%).
A clump of minor journalistic and editorial tasks are supported by around half of Britons while being opposed by close to a third. These include: generating a summary of an article for readers; choosing an image to use for an article from stock photography websites; performing analysis of datasets for journalists to use; suggesting article topics for journalists; fact-checking claims made in a draft article written by a human journalist; suggesting interview questions for journalists to ask someone.
By contrast, Britons are divided about 40-40 on AI conducting research for a journalist on a topic, as well as writing article headlines.
When it comes to fully AI-generated content, resistance grows with the complexity of the article. For instance, while the public tend to be ok with artificial intelligence writing and publishing “short routine data-driven stories such as financial reports, sports scores, or weather updates” (by 48% to 35%), they are opposed to a scenario where it would be writing and publishing “short articles reporting the content of press releases put out by a company / organisation / government and drawing in simple additional context relevant to the story” (by 52% to 26%).
And support falls to just 16% for AI writing and publishing “long articles that investigate an issue in depth that would typically include expert or eyewitness interviews”, with 67% opposing giving such creative control over to artificial intelligence.
Most are likewise to giving an AI program editorial sign-off over whether an article written by a human journalist conforms enough to journalistic standards/best practice that it can be published (55% to 22%).
The public are most vehemently opposed, at 70%, to AI conducting interviews on behalf of journalists. Only 14% say this would be acceptable.
Across the board, young Britons are more willing to see AI taking on newsroom functions than their elders.
This gap is largest when it comes to headline writing and article summaries (both of which 18-24 year olds are 23 points more likely to say are acceptable than over-65s).
But they are also noticeably more accepting of AI writing and publishing wholesale articles, even longer, more in depth ones: 30% of 18-24 year olds say this would be ok, compared to 13% of over-65s.
* The list of news outlets was comprised of 13 major UK news publications, which for the purposes of our analysis have been categorised as follows: Broadcast (BBC News, ITV News, Channel 4 News, Sky News); ‘Up-market’ newspapers (The Times, The Telegraph, The Guardian, The i, The Independent); Mid-market and tabloid newspapers (Daily Mail, Daily Express, The Mirror, The Sun).
See the full results here and here
What do you think about artificial intelligence, the state of journalism, and everything else? Have your say, join the YouGov panel, and get paid to share your thoughts. Sign up here.
Photo: Getty
