Artificial intelligence holds “enormous potential” for local newsrooms (if they know how to use it)

Artificial intelligence is very fashionable in journalism right now. Large Language Models (LLM) like OpenAI’s ChatGPT are the latest iteration of a rapidly developing process. Even though this “generative artificial intelligence” is getting a lot of attention, it’s likely that the biggest impact of artificial intelligence in the news industry will happen behind the scenes.

Huge potential for local newsrooms

So far, machine learning tools have become useful for tasks such as analyzing data leaks, transcribing interviews, fine-tuning paywalls, and giving article recommendations. Publishers like Buzzfeed  and Reach plc are now beginning to experiment with using artificial intelligence to create news articles. Both announcements have led to speculation that this could be the final nail in the coffin for journalists.

Media Voices recently published the report Practical AI for Local Media  which explores what artificial intelligence can do for local news. One of the great advantages is product development.

The US publisher McClatchy uses artificial intelligence to analyze real estate sales and has found that this is a major driver for attracting new audiences, but too resource-intensive to be significantly scalable. So the publisher developed a bot to cover this beat. This new product—regular large-scale coverage of real estate sales—has supported its growth and provided an opportunity to attract new audiences to other parts of its site.

Nordic countries have had news outlets for some time that embrace artificial intelligence to cover resource-intensive sectors. This is significantly helped by the good availability of publicly accessible and well-organized data.

The topics featured in the report, such as property prices, sports reports, and weather updates, all rely on robust and structured data sources, allowing for potentially unlimited granularity in reporting. Artificial intelligence helps local newsrooms filter and identify interesting content from large volumes of data. It can surface stories that might have been missed, such as a bot focused on junior hockey in the US that can detect when a team has won for the first time in 40 games, says Cecilia Campbell, marketing manager at United Robots, quoted in the report.

Ensuring reliable and accurate data sources is crucial. Commonly used data include public data from government and health services, along with data from large charities or commercial data providers. The origin of the data and their potential bias must also be seriously considered.

Misconceptions and misunderstandings

In March 2022, the Associated Press (AP) published a report on artificial intelligence in local news. It examined the understanding and readiness of US local newsrooms and how artificial intelligence could help with newsgathering, distribution, production, and the business side. It found that local newsrooms barely used existing tools, largely due to a lack of staff capacity to learn how to use the technology, as well as issues related to an already complex and fragmented use of technology.

AP has been using artificial intelligence to support news content creation for several years now. A program director had already pointed out in 2021 that the technology “will hopefully provide journalists with opportunities to do deeper and richer stories.”

Artificial intelligence is more than technology: it is a problem solver. In the UK, RADAR by PA Media, created in 2018, uses artificial intelligence to produce national stories with local relevance by digging into datasets and creating versions specific to different areas.

According to the Media Voices report, it uses a combination of artificial intelligence and traditional news. Journalists use models, sometimes different ones, to format a version of the story at the local level, making stories more engaging and relevant to the local audience.

Emilia Díaz-Struck, research editor at the International Consortium of Investigative Journalists (ICIJ), which has used machine learning in investigations for more than five years, told DataJournalism.com in 2021 : “[Machine learning] has a big human component […] it’s not magic, it requires considerable time and resources. Reporters, editors, computer engineers, academics working together: that’s where the magic happens.”

Therefore, “chasing the algorithm” or what might seem like a “quick win” for now is one of the traps publishers should avoid. Professor Charlie Beckett, founding director of JournalismAI, told Journalism.co.uk : “In the short term, publications that prioritize low-quality SEO content might see some advantage, but in the long run others will do it better..

This is also what Jonathan Heawood, executive director of the Public Interest News Foundation (PINF), and his head of impact Joe Mitchell. They told Journalism.co.uk that whether cheaper, “cookie-cutter” content will threaten the work of “hands-on journalists” depends on the “vision and motivation of the publisher.”

In all the case studies in the Media Voices report, knowing the needs of the audience and thinking about the “why” is central, whether to reduce churn or engage readers more with journalism. The goal is to give journalists more time to focus on adding value and storytelling, while AI can handle more mundane tasks. It seems like a good opportunity for all media organizations, not just local ones, to think about their mission and how they serve their audience.

Beware of the technology gap

Many local newsrooms do not use machine learning. Global initiatives like the JournalismAI project could be crucial in supporting local, small, and underfunded newsrooms in developing the AI tools they need.

Heawood e Mitchell of PINF conclude: “Artificial intelligence will not solve the news desert. At worst, it will increase news inequality, and at best, small newsrooms will be able to do more. It depends on the ability of newsrooms to investigate these tools and understand them“.

Source: journalism.co.uk