Even for June, it was a surprisingly warm afternoon when we met John Keefe, in Lisbon’s Terreiro do Paço. He was wearing a smart black tee, squarish specs, and the kind regard of a friend you haven’t seen for years. We briefly chatted about conversational journalism, chatbots, and, just to remind ourselves that we have beating, bleeding hearts, Keefe’s boyhood dream of forecasting thunderstorms on TV.
Keefe is the technical director of bots and machine learning at Quartz, including his work in the Quartz’s Bot Studio, a project that experiments with conversational interfaces, AI, and other new technologies affecting journalism. He was in Lisbon for GEN Summit, the world’s largest gathering of journalists, editors-in-chief, and media innovators, to talk about the good, the bad, and the ugly of the chatbot world.
Early in 2016, Quartz released Quartz Brief, a news delivering app with an interactive interface that deliberately resembles text messaging and is refreshing in its familiarity. With a series of short messages laden with photos, emojis and GIFS, they don’t claim to be so much revolutionizing news broadcasting as casually tinkering with new ways to consume the news. They’ve since been experimenting with conversational interfaces and machine learning (a particularly interesting bot is the one they created for Westworld, which sadly we couldn’t test since it’s only for US citizens), and how conversational software can help media break news.
Is there a role for news organizations in this new format—which is communicating like you would with a friend, but you’re communicating with a news organization. What does that look like? How does that work? Can we do it in a way that is actually meaningful, and not just annoying?
They could, and did, and Quartz Brief was named Apple editors’ choice and one of the 10 best iPhone apps of 2016, with over 10k downloads and users praising the app for its simple interface, short and sweet blurbs, and for Trump Snooze, a feature that helps mute news about the US president for 24 hours.
Although the interface looks like an automated assistant, all of the text is actually produced by a team of writers and editors. Keefe believes this to be its greatest strength:
The secret that I always say is that the best bots are not written by machines. They’re written by humans, which is kind of amazing and beautiful.
A new wave in journalism
As publications face plummeting circulations, organic traffic, audience shares and look for new ways to engage their audiences, these may just be the bots they’re looking for. But despite the recent proliferation of Medium articles on the subject and the various media attention they’ve gathered, journalists and academics have been discussing the ideas underlying conversational journalism for decades.
It became a particularly salient topic during the 1990s, when the public journalism movement emerged as a response to criticism over how the media was broadcasting the news — specifically, how it was reporting on elections, and ongoing debates over the role of journalism in a democracy.
Its objective was very straight-forward, albeit a bit vague. It was meant to renew the commitment between the press and the citizens it vowed to serve, empowering them to actively participate in the public discourse and civic life, instead of standing as passive spectators at the other side of monologues from governing elites. It essentially proposed a shift from the traditional “journalism as information” to “journalism as conversation”, as James Carey, professor of journalism at Columbia University and an active supporter of the movement, put it. From an all-knowing press that merely reported on the issues of a community to one that framed them in a way that the average joe could understand, confront, and act on them.
While the movement garnered a lot of academic attention and traction in newsrooms all over the world, it wasn’t until the proliferation of the World Wide Web that it really took off. With e-mail blogging, forums, wikipedia, podcasting, chat and the advent of social media platforms, the lines started blurring. Suddenly anyone with a computer, internet access and the slightest inquiring mind could create, and distribute, news.
In a dissertation about journalism-as-a-conversation, Doreen Marie Marchionni writes: ”The conversation no longer takes place in letters to-the-editor, if it ever really did, and the potential for journalists to collaborate with ordinary people seems to grow exponentially in the 21st century.”
And that’s where chatbots come in.
Whatever happened to the bot revolution?
A few years ago, a lot of people thought chatbots armed with Artificial Intelligence were the future of user interfaces. “They were meant to be the new websites. They were meant to kill 99% of the apps,” as Yuriy Oparenko, a product designer at Intercom, wrote in an article about our constant obsession with the next big thing. “There couldn’t have been more hype, and I was fully expecting the hype to bear fruit.” But it didn’t. At least not in the way people were predicting it would. But futurology is a faulty science — not to say a complete exercise in futility — and it’s a lot easier to see those decisive moments in our history in hindsight than it is while they’re happening.
Yuriy Oparenko compares the bot revolution that “never happened” with the Apple Macintosh. It was the first personal computer (or rather, the first commercially viable one) to incorporate the graphical user interface, the result of a decade’s worth of research and experimentation from computer science pioneers Southerland, Engelbart, Kay and his colleagues at Xerox Parc in the 1960s.
Before GUI, there was no desktop, no cursor, no icons. You would interact with computers by writing abstruse instructions in command prompts. But almost 35 years ago, back in 1984, Apple’s Macintosh was introduced to the world and literally spoke for itself in a demo that still echoes through the minds of computer geeks all over. And although Apple sold 72,000 computers in the first 100 days, some still viewed it as an overpriced device, or as Oparenko puts it, “a nice expensive toy.”
It wasn’t obvious at the time that this was a decisive moment that would change the future — that it would define the way we interact with machines. But it did. It opened computers to the masses due to its user friendly interface, and gave birth to a generation that grew with word processing, paint, web surfing, and eventually, chat.
Slowly but surely, we’ve seen something emerge out of this human-machine interaction. As Keefe puts it:
We’re getting more and more comfortable talking to our devices.
It sure is great to get out of that bag
Some psychologists believe that, on a neurological level, chatbots trigger the same response as interpersonal conversation. Liraz Margalit, an online behavior psychologist explains that this happens “as bots create a false mental perception of the interaction, encouraging the user to ascribe to the bot other human-like features they do not possess.” This tendency to anthropomorphize non-human entities, such as animals or objects, is widely documented, and bots are no exception to the rule, even when they’re not pretending to be anything other than bits of code.
For a long time, we’ve been designing machines in the hopes that they’ll eventually win the Turing Test, a thought experiment proposed by computer pioneer Alan Turing wherein a human would evaluate the machine’s ability to simulate intelligent behavior indistinguishable from that of a human. Or, put more simply, the machine’s ability to “fool” humans into thinking it’s human too.
Although AI has come a long way since the 60s, chatbot programmers are taking things in the opposite direction, which perhaps says more about the difficulty of passing the test than about their ethical stances or artistic preferences. Language is extremely hard to grasp. It’s full of nuance and metaphor, cultural idioms and idiosyncrasies, figures of speech and other literary devices that to this day boggle the smartest and statest-of-the-artest machines.
So, at least for now, virtual assistants like Siri or Alexa are openly artificial, with no objections from our side. Not only are we not objecting, we’re fully on board, and brands have been quick to catch up. Marketing and sales managers are having a field day, adopting bots in their lead generation funnels, automating processes, optimizing costs, and combining them with agents for 24/7 customer support. According to State of Chatbots report by Drift, Salesforce, SurveyMonkey Audience, and myclever, throughout 2018, 15% of consumers used chatbots to communicate with businesses, and that number is expected to rise through 2019.
And if we look at it from a psychological perspective, it’s not that surprising. Margalit explains that “we’re essentially designing technologies that will give us the illusion of companionship without the demands of friendship.” When we communicate with bots, we can get the information we need, no emotional investment necessary. No need for pleasantries, for reciprocating kindness or availability. As Margalit points out, that sort of detachment comes with a sense of gratification. Maybe the lack of judgement or familiarity gives us the same freedom to reveal details about our life than that of talking to a therapist or strangers in bathroom lines.
But I digress, and the question remains. Can chatbots help journalists engage with their audiences?
Texting the news
The public will begin to reawaken when they are addressed as a conversation partner and are encouraged to join the talk rather than sit passively as spectators before a discussion conducted by journalists and experts.James Carey in The Press and Public Discourse — The Kettering Review, 1992
Chatbots are probably the last thing on anyone’s mind when it comes to helping newsrooms and journalists. In the last couple of years, millions of AI-armed bots have been revealed to be agents of disinformation — creating noise, pushing political agendas, trolling, harassing, manipulating, and distracting us from healthy debate. But the technology isn’t inherently evil, and it can certainly be used for good. Bots can help journalists access data, identify bot-like accounts and behaviours, as well as fact-check. Quartz’s own @probabot searches Twitter for accounts tweeting about politics, using machine learning to determine how likely it is that a given account is a bot.
At GEN Summit, a lot of people were wondering whether chatbots could bring people closer to the newsroom, and if messaging really was the future of journalism. Keefe is leery about such definitive statements, but he does believe chatbots and conversational interfaces will be a part of it. “It is incredibly common to just be able to chat with your friends over text. There are people who I don’t speak to on the phone anymore, I just text them, and a lot of people have that experience.” He continues: “If that’s happening in people’s lives, then I think we need to figure out how we might provide a service along those lines.”
And that’s conversational interfaces’ best feature. Your audience invites you in to one of their most intimate digital environments, one they constantly use to reach out to their loved ones, and one where they spend considerably more time than browsing news online.
Although he doesn’t think publications and news organizations have fully figured out how to inhabit that space, he finds Quartz a great place to experiment with such interfaces. “We like seeing how people interact with things. We’re very interested in user experience. And so that’s why we’ve been playing with chats and voice bots.”
Who dat bot
The Quartz brief app belongs to the kind of chatbot that relies on a content-first experience, where human editors write scripts based on the news that will be distributed via chat. But there are other kinds. Some are essentially glorified search bars — bots which, given a specific keyword or set of words, send you a number of related articles. Others can be used to drive traffic or engagement, sending you articles, but also soliciting your feedback, opinions on certain issues, or even your participation in polls.
The NBC Politics bot provides customized breaking news via Facebook based on their user’s preferences. WeChat launched a bot that works as a bullshit detector for news you find online, telling you if a certain link is real news or fake (although it doesn’t say much about politics). Purple is a subscription-based messaging platform that allows you to reach your audience through text messages. Popular bots include Mueller Time, to follow Robert Mueller’s Russia investigation, or AI Boe News, for updates on the industry. Politibot is a bot for Telegram and Facebook Messenger, designed to cover the 2016 Spanish elections, providing articles, polls, charts and any useful information that could serve its electorate.
Maybe it is, as Keefe said, a matter of experimentation. As 21st century media moves from mass communication to a personalized and almost artisanal model based on social and cultural understanding, it’s up to news and media organizations to see what works best, to get journalists to understand these new technologies and explore them in their newsrooms, and perhaps more importantly, in light of recent events, to question their applications and whether governments and social media platforms are using them in fair and ethical ways.
These are some of the things that keep him awake at night, though most days he just hopes the bots are behaving. As for the future, it may be too soon to tell if conversational technology can save journalism, or whether journalism really needs saving, but we can take some comfort knowing that for every self-assured (or ill-advised) Nostradamus preaching the death of journalism on Facebook, there are plenty more people exploring new ways to do it.