PUBLIC RELATIONS
Wednesday 5th December 2018
Humans on AI
By Matthew Rock,
After a long career in print and broadcast journalism, Gary Rogers knew that the way audiences consume news – in real time and on mobile devices – had changed forever. But he also knew that the way that news was produced – in a labour-intensive, manual, old-school fashion – hadn’t changed a bit.
That disconnect was the genesis of Urbs Media, the data-journalism company he set up in 2015 with Alan Renwick.
The pair always suspected that open (freely available) data could be a gold mine of stories. As Rogers puts it: “Instead of going out and talking to lots of people, you’d ‘interview’ a data sheet and write a story from it.” It was during a period selling data stories into the PR sector that Rogers observed how well segmented most open data is. “An average national data set might be segmented down to 391 local authority levels,” he explains. He decided to write a crime profile of every borough in London using crime data.
After writing 10, Rogers was losing the will to live. So, he Googled ‘How to write like a robot’. This was his entry into the world of natural language generation (NLG).
After speaking to many of the players in NLG, including Narrative Science and Automated Insights, Rogers and Renwick decided to work with Aberdeen-based company Arria NLG, which was seeking to develop a more accessible user interface for its NLG tools.
“We suddenly realised that there was a way for me to write one story – like the crime profiles of London – in the form
of a template and then feed the data through that template to generate lots of versions,” says Rogers.
This was the breakthrough they’d been looking for. They went to the Press Association (PA) and pitched the idea
of a news agency flipped on its head: “Instead of providing central content that’s the same for everybody, you provide, essentially, local content that’s different for everybody.”
Together, PA and Urbs Media decided to pitch the venture (now called RADAR – Reporters and Data and Robots) to Google, whose News Initiative had been designed to provide ‘innovation funding’ for the media industry.
In the summer of 2017, alongside PA, they did their pitch. By September, they’d secured the funding.
RADAR may still be small-scale, but it’s stunningly productive. Over the summer of 2018, by drawing on open data sets and using written NLG templates, five RADAR reporters (all with brilliant Excel skills, says Rogers) produced 40,000 stories.
There are societal and media industry benefits to RADAR’s work, insists Rogers. “We’re telling people what’s happening in their communities.” And, in an age of mistrust in the media, RADAR is generating stories that are traced “to a verifiable data source”, he says. It’s also providing a lifeline to the decimated local media. “We saw an industry that needs content but doesn’t have the people to generate it,” he adds.
All the significant local publishers now draw on RADAR’s output. Half of RADAR’s stories get printed pretty much verbatim, says Rogers; roughly 40% are developed locally by the media, with additional material and case studies. “They can add a degree of context that we can’t, because we don’t know that area specifically. They can develop the story to get something bigger and better out of it.”
And that, of course, is where the PR professional comes in...
What will be the impact of next- generation media on PR and comms professionals? How will your jobs change? And how do you best position yourself in an automated future?
Here are the main conversation threads from our fascinating Influence/Business Wire roundtable – all written by a human...
1. ADDING THE HUMAN DIMENSION
In a world where machines generate ever more stories, the role of the PR may become more about adding “the human dimension”, says Laura Crimmons, founder of Silverthorn, a digital PR agency. “It’s about our ability to look at the story and think: ‘What’s the human angle? Who are the real people that the data is affecting?’”
Business Wire’s Catriona Gilmour calls this process “humanising the data”. Gary Rogers has seen the impact of this left-brain-right-brain combination with RADAR. In 2017, RADAR looked at marriage and cohabitation figures for the UK and discovered that Belfast had among the highest levels of single parenthood. Rogers and team thought their story would be five paragraphs tucked away on page nine of the Belfast Telegraph but, after imaginative human input, it turned into front-page news, a double- page spread inside with case studies, and an editorial leader about single mothers in society.
Chen-Lee Tsui of Business Wire says open data-driven journalism is a great opportunity for PR professionals to add value, and reduces the pressure on them to generate wholly original ideas: “Here’s a story – go and chase after your clients. These are the solid statistics – now you can provide your case study around them,” she says.
2. GETTING COMFORTABLE WITH OPEN DATA SOURCES
From weather information to crime statistics to company accounts, the UK is a world leader in open data – information from government, local authorities and public bodies that can be used to build products and services. Unfortunately, talk of open data draws blank stares from too many public relations people.
“PR people need to get to grips with these information sources,” says Paul Wilkinson, a PR adviser to the construction industry, itself a very traditional sector. “They need to understand the context – the whole continuum from the closed, highly sensitive and private data sets to fully open data.”
Open data sources will generally publish a calendar of what information they’re going to provide and when, says Crimmons. Public relations people should be building their own – or clients’ – planning calendars around these: ‘We know that data’s going to be published, so how can clients contribute to that conversation? Do we have an angle on it?’ These are some of the questions PR professionals should now be asking.
3. BEING VISIBLE TO THE MACHINES
Some clients might even be encouraged to release more information about themselves than is officially required. “PR people have an opportunity to help by making editorial decisions about what information an organisation shares,” says Wilkinson.
“It doesn’t just have to be public bodies that make data open,” he adds. “Private organisations could release more data, too.” One straightforward example might be for a private company to release comprehensive financial results, rather than just abbreviated accounts; when those ‘Top 100’ company rankings get generated, they’ll then stand a much better chance of appearing in them.
4. INFLUENCING DATA EARLY (OR NOT AT ALL)
Influencing journalists is one thing; influencing an algorithm is quite another. If the future really is ‘machine journalism’, PR people will have to learn new techniques to shape the story.
In most data journalism, the early programming stage is crucial. This is when the data parameters are set and decisions are taken about which data to include or exclude. The ‘training set’ or ‘ground set’ determines the long-term outputs, and it’s at this stage, says Andrew Bruce Smith of analytics business Escherman, that PR people should be aiming to exercise influence. “In order to develop the templates, a human being still has to make judgements at the start of the process,” he suggests.
5. GETTING NEW SKILLS
The experts at our roundtable believe that PR and communications people can flourish in an automated era,
but that they must acquire new skills, in areas such as digital and data literacy, statistical analysis, building dashboards, and Excel, in order to stay relevant.
Stephen Waddington has led the open #AIinPR project (with Canadian academic Jean Valin) to examine the effect of automation on PR skills, and he outlined their main findings so far. He explained that around 12% of a current PR practitioner’s tasks – forecasting, analysis and future-gazing, in particular – could be done by technology. In future, however, “around 40% of what we do in PR could be completely stripped out by technology”. The core PR skills that will remain relevant are mainly around empathy, persona judgement, ethics and legal compliance, Waddington believes.
6. FAMILIARISING YOURSELF WITH NEW TOOLS
When asked about the impact of AI on PR, most professionals sit on a spectrum between grudging acceptance and “blind panic”, says Waddington. The market for tools in public relations is exploding, he notes. “AI is about to massively change our lives. The public relations profession needs to keep up. We need more experience with these tools and more critical reviews to learn how best to use them, and their limitations.”
Crimmons agrees. “There’s no ethical considerations or potential damage with many of these tools and tasks
– they just save time!”
7. TAILORING MESSAGING TO EMOTIONAL NEEDS
We’re approaching a time when AI will influence what, as well as how, you write. John Savage is chief commercial officer at Symanto, which uses – scary word alert! – ‘psychographic segmentation’ to identify audience segments and how they want to be communicated to.
Symanto recently did some work around the diesel emissions scandal. It found that consumers fall into two camps: a more emotional segment that wants the automotive companies involved to own up to, and explain, what they’ve done; and a more rational group that just wants ‘next steps’ explanations.
There’s a wider phenomenon at play here: a growing divide between the kinds of information that consumers want to receive in a cold-blooded, factual format – about their finances, for example – and those where they want more emotional, human coverage. (An interesting aside: Savage explained that consumers of sportswear are often ego-driven, so content in this area often benefits from a dash of ego-tailoring.)
8. LOSING CONTROL
Chatbots are increasingly at the front line of communications, as we discovered in the Q3 edition of Influence. Wikipedia, for example, uses a number of bots to create content and detect and prevent ‘cyber-vandalism’ on its site. As long as such bots remain “harmless and useful”, this kind of bot deployment is bound to grow, says Escherman’s Smith.
But a word of caution. Microsoft’s humiliating experiment with Tay – the automated Twitter bot was corrupted and became a racist within 24 hours of launching – showed that automated communications can have often memorable limitations. Managing these – especially as ‘deep learning’ and impenetrable ‘black boxes’ become pervasive – may yet become the PR professional’s greatest ever challenge.
9. SOME OLD RULES STILL APPLY
One final thing. As Gary Rogers says with a smile: “With data, PR people and journalists will still say: ‘Is it new? Is this the latest? Has anyone else got it?’”
Whether you call it AI, machine journalism or augmented media, some old rules still apply.
IN SHORT: ARTIFICIAL INTELLIGENCE AND THE MEDIA
AI is turning PR and the media on their heads. Every day thousands of stories
are generated by machines. Human journalism is routinely augmented by computers, enabling huge, data-intensive investigations (the Panama Papers being just one example). Tasks previously undertaken by human interns are now performed by algorithms – which don’t complain and always get better at their job.
“We are at the very, very early stages of AI,” observes Scott Jamieson of Business Wire. He says that many news organisations are adopting AI, applying it particularly to data that comes in highly structured and tabulated formats. This enables them to conduct rapid, rock-solid analysis of, say, company performance.
The next chapter, says Jamieson, is when AI starts looking at more unstructured data, such as text, video and email. “That’s when things will start getting really interesting.”
In his important report The Future of Augmented Journalism, Francesco Marconi, The Wall Street Journal ’s R&D chief, suggests we’re only just scratching the surface of AI: “AI can do much more than churn out straightforward sports briefs and corporate earnings stories. It can enable journalists to analyse data; identify patterns, trends and actionable insights from multiple sources; see things that the naked eye can’t see; turn data and spoken words into text; text into audio and video; understand sentiment; analyse scenes for objects, faces, text or colours – and more.”
This article was originally published in Influence magazine, Q4 2018.