There are thousands of them on Facebook Messenger alone, and journalism organizations with plenty of resources and lots of room for innovation have been experimenting with chat bots, voice interfaces and machine learning to engage their audiences.
For the uninitiated, chat bots are dialogue systems that allow purposeful interactions between computers and humans through text or speech. John Keefe is a bot developer and product manager for Quartz, which calls itself a digitally-native news outlet. When Keefe talks about bots, he points to one on Facebook Messenger called Poncho, which messages you every day about the weather.
“What makes it different is that it’s a little bit playful with the weather; it’s very casual; it has a voice,” Keefe said. “When the developers found out people were asking Poncho about other things, they reacted, so now it responds to other questions like, ‘I’m hungry.’ It’s about having playful interaction with the audience.”
The Quartz Bot Studio recently launched its own Facebook chat bot to capitalize on the conversation surrounding the second season of “Stranger Things” on Netflix and another one called Quartzy, which strives to be a “cultural companion and guide.” What’s especially helpful about their work in this area is the bots.qz.com page on which Quartz is sharing its experiments with other journalists and the world at large. One creation is a bot for Slack users that allows you to instantly chat your way to grabbing screenshots of webpages or identifying clichés in your writing.
But that word “experiment” is key to the discussion as most news organizations are still searching for the killer app when it comes to bots and voice assistants. At the Washington Post, Joe Price has the title of system builder and says they’ve been focused on building products for voice interaction, which he says, “decreases the distance for people between thinking and taking action.”
Voice assistants include services like Amazon Echo’s Alexa or Google Home, which are capable of several different interaction types. Price says native interactions include telling a joke or reporting the weather, native integrations allow the assistants to play songs from your Spotify account and custom skills could range from playing “Jeopardy” with you to opening up and reading a section of the Washington Post aloud.
Price says he and his team have already learned a number of lessons in the work they’ve done in this area. For example, in 2016 they built a special election-oriented voice assistant service that looked at how a candidate was doing, how he or she might be polling in different states, etc. He says they saw a huge spike in usage and then a huge drop off.
“What we failed to ask is, ‘How does this compare to just pulling out your phone? Do people wonder about Iowa electoral results in their kitchen?’ We needed to evaluate the friction with other platforms, especially the phone,” Price said.
The Post’s second experiment was launched in conjunction with the summer Olympics and the plan was to provide a service that would report event results along with other relevant information for the avid Olympics fan.
“It turns out that audio lists are the worst; we found out it was like listening to your server tell you about the specials for the day.”
In the end, Price said they determined that Alexa itself is going to handle what he calls tentpole events and general domains, such as Super Bowl or the Olympics, so they moved on to new ideas such as “Dinner in Minutes,” which capitalized on the Post’s vast archive of recipes.
“It was a total bust,” said Price with a laugh. “It turns out cooking is not strictly linear.”
With all of these experiments, Price says users said they liked the content but they didn’t come back to it, so why keep working at this?
“Journalism creation is inseparable from journalism delivery. We need to think about how to craft journalism to meet users where they are.”
One success story has been the Post’s Alexa Flash Briefing. It lasts 3-5 minutes, the average user listens to it seven times per month and 23 percent of listeners are new audience. The final lesson learned?
“We have to create a habit-forming product; we have to create an experience that the audience needs.”