Editor’s note: This is one of the occasional posts we do speaking to someone with something to say on topics that we find interesting. If you are seeing this and are not yet signed up to Deepnews, click the button bellow to start receiving our blog posts every week, and a Digest of quality news on an important subject every Friday
By Christopher Brennan
Robots, they’re all around us. From the Alexa on your countertop to the bot on Twitter spewing out nonsense. But can they be used for good, to get people their news in a better way? And how would people even feel about getting their news from a robot?
After previous blog posts on GPT-3 posting on social media or on Substack, I have become increasingly interested in the future where a huge amount of content, including perhaps news, is generated by artificial intelligence tools. Some of this is scary, but there are also efforts that want to use them for journalism.
In fact, we already get much of our news through robots, recommendation algorithms on news apps or social media, which are increasingly being understood as making editorial choices.
But what about getting your news from a robot that is a bit more concrete, not necessarily in the Blade Runner way, but maybe as an account on social media that you could follow and talk to conversationally?
“Newsbots” have not really reached that phase, though newsrooms have given them a shot. A group of prominent outlets made chatbots on Facebook Messenger but then abandoned them. Others have been more focused on specific events like a referendums or elections. In general, users interact with the bots through pre-scripted material, a limited set of questions and answers. (If you were on instant messaging platforms such as AIM in the early 2000s, it may remind you of SmarterChild).
But the rapid technological advances in terms of being able to understand and generate text algorithmically means that some are still looking at a future where robots are the point of delivery for news. With that in mind, one interesting paper from Diego Gómez-Zará and Nicholas Diakopoulos at Northwestern University (as featured in the newsletter RQ1) was focused not just on newsbots but on how people react to them.
“Most tools are conceived as a tool, just as a machine, right? So, the point here was to see if people are able to see news bots as a partner, if they could be a facilitator,” Gómez-Zará told Deepnews in a recent chat.
“Can humans see machines as agents? Can they see them as people or can they see them as someone they can communicate and interact with, not only just a channel or a tool?”
The paper used a bot made for Twitter that would respond to users who tweeted articles from the New York Times. As a sort of added value, the “Anecbotal” bot would select a “personal anecdote”-type comment from the hundreds of comments in the NYT comments section on the story and share it with the user. The goal was to measure how they responded to being chatted up by a robot.
The results were split, with large amounts choosing just to ignore it, a majority of responses that were positive notes of thanks or appreciation to the bot and occasional swear words against it. There was also a split on how Twitter users responded to the bot itself, with some not referencing or talking to it as they engaged with the “anecdote” that it had shared.
Other users, however, interacted with the bot as a bot, as what Gómez-Zará calls an “agent” instead of a tool, even though it didn’t take part in long conversations or use artificial intelligence to generate a response.
A small sample size and the specifics of “Anecbotal” (what it was sharing, its few followers, a lack of connection to a trusted news outlet) limit what we can really know, though the paper provides an interesting jumping off point for thinking about how humans being ready to accept news from robots.
Gómez-Zará views the next steps as having a better (algorithmic) understanding of information in news and then providing information for context, traditionally one of the key roles of a journalist.
“If I asked Alexa, “tell me the weather,” that’s a very simple question. But if I say Alexa, “Do I need an umbrella today?” it’s associated with the idea that it’s going to rain and because of that we need to get information on the weather,” he said.
In terms of news, the researcher believes a smarter newsbot could use its knowledge (taken from the outlets’ news articles) to help answer specific questions that someone has about the news. If its writing could achieve a certain level of quality (which at Deepnews we are of course always interested in) there could be more extended conversations.
The goal would be to plug gaps in the readers’ knowledge and keep them informed beyond reading one article. It is a way off from what is currently available from newsrooms, but also from current chatbots, which are more developed for customer service or telemedicine. But there are hopes that with advances it would ultimately allow users, in a personalized way, to go more in-depth into news.
“[News readers] would be able to have a conversation with someone. Not necessarily with a journalist because it’s very hard to address thousands and thousands of replies, but a machine can do that,” Gómez-Zará says.