What are the use cases and challenges of using automated data analytics in journalism?
Information flows faster every day. As a reader that can easily become exhausting. But for media automated/ algorithmic data analysis (which is basically what happens, at least for now, when journalists talk about Artificial Intelligence or AI) presents huge opportunities.
It can already automate big data analysis, create loads of content without human involvement and personalize newsfeeds down to the individual user. Many outlets have already dabbled with integrating AI into their working processes – and many have run into various challenges.
The Fix examined some use cases of AI in journalism and compiled a list of the most important insights.
What is AI in journalism (at present)
One of the biggest use cases for AI in journalism is content automation – software or robots producing news stories by computer software or robots. These programs use algorithms and natural language generation (NLG) to automatically collect, interpret, and transform huge amounts of data into news-text in real time.
Robot journalism has grown significantly in recent years and has spread globally. According to Charlie Beccket`s report, based on a global survey of journalism and AI, 44% of news organizations have already experienced the impact of AI.
Here are a few examples of using AI solutions in media orgs:
Washington Post: Heliograf robot reporter.
WaPo has a robot reporter called Heliograf. Heliograf doesn’t replace journalists, but helps reporters find important data for news coverage. The publication used the robot to cover multiple large-scale events and processes, like the Olympic Games, sports competitions and elections.
During the presidential election in 2020, WaPo updated the technology with an AI voice assistant. It would automatically find, read and insert the election updates into political coverage.
The New York Times: Editor and Moderator.
NYT started using their AI-based software “Editor” back in 2015. Its goal was to help the editorial team find and analyse information. Editor uses semantic tags to find and compile essential stories. The story then goes through the human edits ensuring proper fact-checking and storytelling.
NYT is also applying AI to its comment section. The “Moderator” tool (created in cooperation with Jigsaw, an Alphabet technology incubator) groups similar comments to help the team of moderators review them faster and alerts them about potential harassment/ insults.
The technology’s main goal is to create a safe space for discussions and free up the team of moderators, who reviews some 11,000+ comments daily. That way, they could engage more with readers (here’s a video with more background).
BBC: AI-powered synthetic voice.
More than half ofthe BBC’s readers (62%) spend between 30 minutes and four hours listening to podcasts every day. In November 2020, BBC Global News launched a synthetic voice toolthat uses artificial intelligence to read articles from its website. The voice automatically updates its output when original content is updated.
Reuters:Automated video reports.
In 2020 Reuters built an automated video report system. The tool generates news scripts on sport matches and transforms a pre-recorded video of a presenter into a new report.
The system uses the technology from British AI start-up Synthesia. The solutions is quite similar to the one used to create deep fake videos.
Advantages of Artificial Intelligence systems for media
Speed and scale of news coverage.Unlike human journalists,robots can generate huge amounts of stories in a very short period of time.
According to Automated Insights, AI allows the Associated Press to automate the production of short earnings and sports reports. This process produces 4,400 stories per quarter, compared to just 300 previously. It also freed up the equivalent of 3 full-time employees.
Cheap cost (kind of). Hiring data scientists, building machine learning models and maintaining them can be prohibitively expensive.
At least two factors make AI solutions competitive compared to hiring journalists (especially in high-wage countries). Firstly, AI used in media is relatively unsophisticated. More importantly, spreading costs across a large number of publishers makes the economics quite attractive.
In 2020, custom-made AI solutions for publishers cost up to $300,000, according to internet marketing and web development company WebFX. However, third-party solutions can cost just a fraction of that.
Efficient research and content management. AI solutions can turn important stories into simple reports much faster than humans. They can provide more context, add supporting data and even links. Moreover, robot systems are able to create stories from raw data and synthesize multimedia files, including audio fragments, images and videos.
Content customization. Automated news can be customized to a particular audience group, transformed into different formats and translated into many languages. For example, Reuterscreated an AI tool that can automatically transcribe and translate archive videos into 11 languages.
Scalable news verification. Correctly programmed tools help fact-check information, uncover fakes and mark suspicious content. Social media platforms such as Facebook and Twitter use AI to catch word patterns that can uncover hate speech, fake stories or misleading information.
3 main risks of Artificial Intelligence for media
Speed isn’t the only that matters in journalism. Quality of media products depends such things as accuracy, balance and separating facts from opinions. In this regard, the quality of automated content is still questionable.
Misuse of AI technology
While most media try to protect readers from fakes, others embrace deep fake AI technology. They produce false images and audio records to manipulate the audience.
British Channel 4‘s “deepfaked” video of the Queen is just one such example. The video features Her Majesty giving a Christmas speech and dancing on a table. The video was planned as a joke, but failed. The UK regulator reported more than 300 complaints about this episode.
AI systems and robots require much less investment and operational expenses than human workers – no lunch breaks or vacations, no sick leaves, automated learning process, no salaries and HR investments.
Newsroom investments in AI and automation are likely to lead to job cuts. Microsoft has already sacked around 50 journalists, replacing them with artificial intelligence since June 2020. To be fair, for now there is more augmentation of editorial capabilities than replacement going on. But this may change.
AI algorithms don’t have critical thinking or transparency, at least yet. If robot journalists collect inaccurate data, this will generate and spread biased stories.
Unlike potential job losses or misuse of technology, ethical misjudgment is already happening. A recent case was associated with the racial conflict coverage and a misunderstanding caused by Microsoft AI software. The story about an artist’s personal opinions about racism ended up being illustrated with a picture of her fellow Little Mix band member. The program, which selected images for MSN (Microsoft Network) website, confused the images causing MSN to be criticized for “ignorance”, the Guardian reported.
All said, one thing remains clear – the use of Artificial Intelligence solutions in media can greatly improve both the quantity and quality of good journalism. This will certainly create new opportunities. But we need to understand AI’s limitations. Moreover, AI solutions can hardly substitute human reporters and editors, increasing the value of human judgement.
This piece was originally published in The Fix and is re-published with permission. The Fix is a solutions-oriented publication focusing on the European media scene. Subscribe to its weekly newsletter here.