Politibot and the value of conversational interfaces

Politibot is a conversational news chat project, created as an experiment by a group of Spanish journalists and technologists and currently available on Facebook Messenger and Telegram.

Politibot is not a “smart” robot, in the sense that it does not have natural language processing interface: it simply offers replies from a menu. On some occasions it offers several alternatives, with different answers or conversational routes, while on other occasions it simply offers an answer so as to simulate a conversational interface and break the dynamics of linear information. The idea is to provide an interaction with a news item different to traditional reading in which the user has the feeling of interaction with an entity that is providing information in small doses.

Eduardo Suárez and María Ramírez are two of its creators: I introduced them to my friends at BigML with the idea of increasing the analytical conversational interface to allow for more freedom in the conversation, but realized this would pose problems by exceeding the pretensions of the application. In reality, Politibot is simply a different way to read a news story, allowing the reader to interact with the information in a more enjoyable way, making it more likely to be remembered.

The idea, basically, has a lot to do with what Confucius said, and that all teachers know:

“I hear and I forget. I see and I remember. I do and I understand.”

Listening to or reading content provides lower levels of retention than if you engage in it via a conversation, even if that conversation is simulated and comes down to choosing a few options from a menu. And after a few weeks of news consumption in Politibot, usually only one a day, the experience reflects exactly that: Politibot offers me the opportunity to “talk” about the French elections, about which I know little, meaning I can acquire an understanding of the main issues, who is who, etc. Politibot does not just tell you a story, but instead offers details — phrases, graphics, podcasts, news and comment from other media, etc. — so that it builds a story in which the user plays a slightly more active role, feels more involved, and retains content better, simply by pressing a button.

The applications are many and varied: firstly, in education. I know from experience that my students get a better grasp of a complex subject when they interact with it, as opposed to me simply explaining it or telling them to read an article. I can now see that I could propose simple interaction and improve understanding of a subject, an approach already part of simulations and business games.

Secondly, the interface: there are days when I am cheered up by Politibot’s “hello human”: I’m not sure if I’ll ever regard Politibot as a friend, or whether someone could fall in love” with a chatbot as in the film Her, but the idea does not seem completely out of the question :-) And all this, thanks to an interface that borrows from instant messaging, something we are all familiar with, along with an app we already have installed.

Then there is the question of manipulation or indoctrination, not in the case of Politibot, which seems to me to be balanced, but through other developments: in principle, the more varied the sources of information we read, the less room there is for indoctrination, but an interface like this raises the possibility of making certain information more palatable through conversation.

In any event, Politibot is an interesting exploration of the possibilities of conversational interfaces or chatbots, both for my work as a technological analyst, and in my professional development as a teacher. A conversational interface helps retain information through a conversation, leaving me with the feeling of knowing more. All this, thanks to a simple interface and some trivial options that make me an apparently active part in a conversation.

Chatbots are an interesting interface applicable to many areas: Politibot is proof that there is no need technical sophistication or especially ambitious natural language processing for surprisingly good results. We now have the choice of reading a story, as opposed to getting involved in it, even if it is very superficial. Time to think about the many doors that this potentially opens.

(En español, aquí)