Shutting down your chatbot? OK, but here’s a tip for when you start it back up
If you want to use chat for news in a more organic way, adapt your approach by thinking about how the medium is fundamentally different from the way journalists are used to communicating information.
As someone who’s fairly bullish about the future of chat in journalism, I was bummed to read this week that many news organizations are pulling away from their chatbots, though I understand why. It’s a tough space right now, with a high bar for user satisfaction that oftentimes the technology just isn’t prepared to meet, and apparently it’s currently a bad return on investment.
One of the wonderful things about my job as a journalism professor at USC is that I don’t have as many ROI concerns, so I’m able to experiment with storytelling more freely. My time working on chatbots leads me to believe that it’s not a great idea for news organizations to focus on short-term ROI instead of long-term investments in this aspect of our evolving media landscape, especially as audiences are increasingly getting news on messaging apps. There are benefits to trying out chat that go beyond driving traffic to your site, and the time is now to adapt our thinking to chat.
From the very first time I opened a chatbot CMS and started typing, I knew that writing for chat presented new (and fun) challenges. And since then, I’ve worked with my students at USC Annenberg to launch on Facebook Messenger, WeChat and Alexa, and we’ve developed an app with a chat component. But in all of those cases, we followed a well-worn model: pre-determined prompt → pre-packaged content (a list of stories, a pre-recorded briefing — basically what you’re seeing from lots of other news organizations in the chat space right now). It’s fake two-way conversation.
Enter Amy the Stylebot, the copy editing Slackbot I launched in November with my colleague jenn de la fuente. Amy is a different model — you type your copy editing questions, and she (hopefully) has an answer. Even though I have gone from adapting my writing from wire service, to homepage, to social media, to mobile, I’ve had to adjust the way I think about presenting information the most when working with Amy.
Launching the Slackbot seemed like a straightforward idea: Looking up entries in a stylebook is essentially keyword based, and many chatbots are keyword based. It’s a perfect fit! I am accustomed to using stylebooks by developing keyword-based strategies to uncover the information I want to find, and since my students are being trained to use the AP Stylebook, I thought maybe they’d approach the bot the same way.
It’s not surprising that I was wrong — you’ve probably heard or experienced yourself that user expectations are high when engaging with a chatbot. My idea of what “stylebot” communicates (in my mind, it’s a stylebook, but instead of flipping pages looking for keywords alphabetically, you type those terms into Slack) has been so thoroughly taken down that a student has already asked, “Is this sentence stylized correctly,” followed by a sentence pasted from a story.
Even though I’ve not come across a bot capable of the sophistication required to answer that question, I live by the philosophy that the user is never wrong, only the people behind the product are. So this question, and many others that prompted me to think something along the lines of, “Well, obviously, if you had typed ‘comma,’ you would have gotten what you wanted,” led me to an examination of the way journalists are used to inputting and presenting information and the output audiences expect. I’ve come to this conclusion: A stylebook is like a tweet. And the nightly news. And a social video. And an Instagram story. They are all packages with the message of finality, presented on the producer’s terms, leaving the audience to adapt its approach to the content — just as I strategize about what to look up in the stylebook when I need an answer to a specific question. I fear that journalists have become so accustomed to presenting news this way that the move to real chat will be a bigger disruption than we anticipate.
Chat presents the facade of being infinite, of being smart and capable. Anyone who has been introduced to chatbots, and inevitably has been let down at some point, intuitively knows this, right? I did. But I still wanted to create a chatbot on my terms, based on my experience, workflow and process. I had answers — all I had to do was brainstorm questions that might elicit those answers, and voilà!
Digging into my reaction to questions Amy got that I didn’t anticipate reminded me of something my friend and colleague David Cohn pointed out last year about CMSes: That vehicles for information input (or in a chatbot’s case, output) are so more than the sum of their parts. They communicate something important to users that manifests in the very nature of the product. So thinking about chatbots as extremely conversational ways to present pre-packaged information to people on the other side of the conversation is a good start.
But it benefits us all to think critically about how true conversation is such a departure from decades of journalistic habit and to shift our minds from the notion that chat is just like verbally clicking a link on a homepage. So when it’s time to start on a truly interactive chatbot, start not with a question such as, “What might someone ask when trying to learn about X?” — though that question will also be necessary. Instead, start further back by asking yourself, “What am I communicating to the audience by the format in which I’m presenting information?” This might help you better tailor your approach from the outset, leaving you with fewer surprises and your audience with fewer disappointments.
