Sviatlana Höhn, pictured, has set up a chatbot laboratory Mike Zenari

Sviatlana Höhn, pictured, has set up a chatbot laboratory Mike Zenari

The train is late. While it sits stationary on the tracks just metres from Luxembourg’s central train station, I decide today is the day to test the government’s new consumer rights chatbot, Poli.

Clicking on the website, I explain the situation and ask what I can do. “Hello, I’m Poli. Were you informed of the delay when you bought your ticket?” a chat bar reads. I click “no” and explain the length of delay is less than 60 minutes.

Within seconds, Poli replies that I’m not entitled to compensation, signposts me to transport information and consumer rights websites, and reminds me this information is not a legally binding text.

Poli may be in the early stages, and can currently only handle questions in French about telecommunications, travel and commerce, but, because the chatbot uses machine learning, these sorts of interactions will improve over time. That’s good news for anyone seeking to know their rights in Luxembourg.

The fact that the Luxembourg government is jumping on the chatbot train speaks volumes about the possible uses for these kinds of computer programs. “Before, we produced chatbots because we could. Now, the really interesting time starts as people are looking for meaningful cases where chatbots are useful,” says Sviatlana Höhn, a computer science post-doc who set up the chatbot laboratory at the University of Luxembourg and a meetup group on the topic.

Poli is certainly a useful case, providing quick access to information, which potentially opens the door for other language versions in future. But it is not without risks, and Höhn stresses providers and users must consider who is responsible if the wrong information is given. “You need to understand it’s a machine and there’s a chance it doesn’t understand,” she says.

Poli is all the more surprising given Luxem­bourg’s reluctance to embrace chatbots. This cautious approach comes from the dominance of the financial sector, where established banks and insurance firms generally deem customer service bots as high risk. Startups may employ them as a “distinguishing feature”, Höhn suggests.

But, on the whole, customer-focused bots are mostly confined to proof of concepts, or put to work on internal tasks such as supporting HR functions, booking appointments or managing IT help desk requests. That could be about to change with the launch this year of Henri, a communications and reputation management chatbot simulating human conversation.

“Businesses will use bots. It’s not a matter of if but when,” Laurence Ponchaut of Distinct Communication says, explaining that her firm and Nowina Solutions created Henri to support Luxembourg companies in developing their own bots on company premises. “This is especially important for the financial sector as banks and insurance companies treat sensitive data,” Ponchaut explains.

She suggests it only needs one major Luxembourg player to embrace chatbots for the business community to follow suit. The social and educational sectors also offer plenty of opportunities for chatbot development.

Höhn is developing a language teaching interface while a masters student at the university is working on a testing chatbot for other computer science students. Then there are the headline grabbers like ­LuxAI’s smart social robot companion, QT.

As is often the case, however, ideas develop more quickly than technology and Höhn says we are a long way off having a conversational chatbot that recognises and understands dialects of Luxembourgish, for instance. Another challenge is getting end users to accept chatbots with their current limitations. Höhn says: “This is a first step for things to get accepted and for people to not be afraid.”

This article was originally published in the September 2018 edition of Delano Magazine.