Warning: Parameter 2 to SyndicationDataQueries::posts_search() expected to be a reference, value given in /Library/webserver/documents/seriously/media/sub-topics/media/wp-includes/class-wp-hook.php on line 298
Warning: Parameter 2 to SyndicationDataQueries::posts_where() expected to be a reference, value given in /Library/webserver/documents/seriously/media/sub-topics/media/wp-includes/class-wp-hook.php on line 298
Warning: Parameter 2 to SyndicationDataQueries::posts_fields() expected to be a reference, value given in /Library/webserver/documents/seriously/media/sub-topics/media/wp-includes/class-wp-hook.php on line 298
Warning: Parameter 2 to SyndicationDataQueries::posts_request() expected to be a reference, value given in /Library/webserver/documents/seriously/media/sub-topics/media/wp-includes/class-wp-hook.php on line 298 How the Coloradoan Is Experimenting with Bots and What They’re Learning – Media
How the Coloradoan Is Experimenting with Bots and What They’re Learning
This piece was originally published by the Donald W. Reynolds Journalism Institute at the Missouri School of Journalism as part of a new RJI series. It is used here with permission.This Q&A has been edited for space and clarity.
A planning editor with the Fort Collins Coloradoan is experimenting with bot technology and it’s taught her that it doesn’t require a team of technologists. Jennifer Hefty says it doesn’t even require HTML coding experience.
Relying on research, a free bot interface and resources already in the newsroom, Hefty built and launched her first bot, dubbed Elexi, prior to the November 2016 elections to provide audiences with the necessary information to become more informed voters.
Hefty was inspired to bring the technology to her newsroom after learning more about bots at the Online News Association annual conference in September 2016. With positive feedback from the first experiment, she’s getting ready
launch her next bot experiment.
Senior Information Specialist Jennifer Nelson spoke recently with Hefty about Elexi and its use during the fall elections.
Jennifer Nelson: What is the Elexi project and what prompted you to launch it?
Jennifer Hefty: We really wanted to find a way to present our election information in a way that was easy for our audience to use and interact with, and to try something new.
We had been talking before ONA about text alerts or something similar for the election. What could we do and what would that look like?
I decided to build a bot with the technology I had learned at ONA. There is a program called Chatfuel. It’s pretty intuitive and it connects with Facebook. They say you can build a bot in seven minutes or less, and you probably can, but this one took me a little bit longer.
My goal for the bot was to create something that wasn’t just routing users through a pre-determined set of options. I wanted Elexi to be smart and conversational. I wanted people to be able to ask questions, just as they would if they were in a chat with a human being. We had all these preview stories on our site on local issues. If a user asked a question about a candidate or a race, Elexi would come back and say, “Here’s the preview on this and what you need to know.”
I also wanted to give her a little bit of a personality, too. If people asked about Donald Trump or Hillary Clinton, I wanted her to be a little cheeky in her responses, since we weren’t covering the presidential race on a local level. For example, if a user asked, “Who should I vote for for president?” Elexi would encourage the user to instead ask her about a local race and say, “I’m staying out of the heated presidential debate.”
Her purpose was two-fold: I wanted her to provide a guide so that readers could ask the bot questions about ballot initiatives and learn more about them. And I also wanted her to be a tool to provide election results in real-time on election night.
What kinds of questions did your audience ask Elexi?
Hefty: During the election, our readers asked a multitude of things about specific candidates or ballot measures.
If they asked, “Who is running for Senate in Colorado?” or “Tell me about the senate race,” Elexi would send them a link to the story “What you need to know: U.S. Senate race: Michael Bennet versus Darryl Glenn.”
Readers could also ask how they could register to vote or ask how to check their registration status, and Elexi would give them appropriate links on the county or secretary of state websites. Elexi could also provide ballot drop-off locations and hours, as Colorado is a mail-only ballot state.
If Elexi didn’t know the answer to a question specifically — for example, we got a lot of questions about whether judges should be retained — she would send them to the Blue Book online to learn more.
Readers could also ask Elexi’s opinion on candidates and ballot measures. While Elexi was programmed to remain impartial, she would give users the option to visit the Coloradoan’s endorsements. For example, if you asked Elexi “Who should I vote for?” she would say: “Whom and what you vote for are extremely personal decisions, and our goal is to give you as much information as possible to make informed choices. That said, the Coloradoan Editorial Board is weighing in on important issues in the community. Click below if you’re interested in reading this year’s endorsements.”
What resources did you need to launch this project?
Hefty: I did some research on it [beforehand] but there was no extra money spent. The interface to build a bot through that program is free. We had no extra manpower but it did take me a bit of time — probably about a week on and off — to get it built and get everything programmed because the ballot was really robust in November.
The nice thing about the program is you can see what questions people are asking. Anytime the bot would say, “I don’t understand what you’re saying,” we would go back and re-program it so hopefully the next time, it would understand.
Also, at any point, a staff member could override the bot and provide resources and information to users.
How did the first bot experiment go once you launched it?
Hefty: I hated some of the other bots I experimented with that didn’t adapt or learn or get smarter. So I spent a lot of time with Elexi to program the AI [artificial intelligence] so that it would start to understand and get smarter as more people would ask it questions. Chatfuel allows you to “teach” your bots words or phrases, and then direct the user to certain responses. The bot gets smarter over time, and the more users interact with it, the more easily it recognizes those words or phrases.
We really had no benchmarks for success because we had never done anything like this before. I think we ended up with about 200 people who used the bot during her lifetime. The interesting thing was coming up with a content marketing plan and describing to our audience what the technology is. I made a video showcasing Elexi’s “personality” and wrote a column, and we promoted the bot in print, on site and across many social platforms.
On our Facebook page, we have about 50,000 followers and we get messages from people daily and we try to respond to every single one. We didn’t want to have this very specific bot tied to our brand account so I created a separate page for her. I feel like this made the marketing harder because people had to leave our page and go to something else. When I shut down the bot, I sent everyone a message that said that it wasn’t going to be updated from that point forward. I sent people a survey asking if they liked the experience, what they didn’t like, what they wanted to see more of. I got about a dozen or so responses and most of them were positive and that they liked the technology. The thing was it’s still new to people and they weren’t fairly sure how to apply it to anything else.
Now that you have one experiment under your belt, what will you be doing differently with your next bot project?
Hefty: The next one will be tied to the brand account. This bot will be able to do a few things. First, it will address minor subscription issues, like if people don’t get a delivery or they want to put a vacation stop, it’ll send them to a real person in Fort Collins who can do that. There’s also a subscription-ask in there if people want to subscribe, that sends them to a deal for a subscription rate. The new bot will also ask users what their interests are: “So are you interested in breaking news and weather? Are you interested in receiving our top five stories of the day?” What I’m able to do is to save groups’ preferences and then I can send alerts to those groups.
Jennifer Nelson is the senior information specialist at the Donald W. Reynolds Journalism Institute. Previously, she was the news editor of the Osceola (Iowa) Sentinel-Tribune.
The post How the Coloradoan Is Experimenting with Bots and What They’re Learning appeared first on MediaShift.