17 Apr Three practical uses for AI argumentation
Argumentation is an incredibly complex area of language – so much so that philosophers have devoted millennia to studying it. Simple questions, such as what makes a debate successful, and which assumptions are legitimate, have very challenging answers.
This is why the field is so interesting to computational linguistics. If computers can understand the building blocks of argumentation, mirror it back to humans, and learn from the ideas presented, then it could have a broad range of applications. Francesca Toni, Professor in Computational Logic at Imperial College London, explains three areas that will benefit from argumentation technology when it is perfected.
“Medics face huge challenges,” Francesca says. “They are overstretched with an increasingly wide array of clinical trials and medical guidelines, and patients are more and more involved with the decision making. There’s a lot of conflict in the information available to them.” Francesca explains that argumentation could help medics navigate this information and come up with the best treatment for patients. By taking the human error and irrational emotions out of medical decisions, doctors can give an authoritative and logical response.
2. Product reviews
There are few places on the internet more toxic than a comments section, but AI argumentation may be able to cut through the noise. “When we buy a product, we look for reviews and it’s very time consuming to go through all of them. Often, they are full of conflicts,” says Francesca. “Argumentation can help here by identifying the arguments and conflicting opinions and resolving them.”
3. Customer support
“Another way AI can help is in customer support, which is a key ContactEngine mission,” says Francesca. “In customer support, you need to identify the best arguments to satisfy a customers’ request. You may also want to be able to learn from the arguments a customer puts forward to give better service in the future.” If computers are to succeed at the front line of customer service, then they need to be sensitive to the changing mindsets of the general public.
How close are we to getting there?
According to Francesca, this isn’t the stuff of distant science fiction. “We’re almost there, we already have systems that can argue and help in a range of settings,” she explains. There are just two hurdles to making the technology more available for everyone. One is being able to identify a broader range of arguments and process them and the other is to find out the best way to show the explanations for certain behaviour.
“Humans relate to argumentative explanations so well but it’s not clear what form these explanations should take. For example, in product recommendations, should we present them as a ‘for and against’ or the whole debate about why those arguments are valid?”
Argumentation is key to being able to get computers to better understand humans. As data sets become more available and AI becomes increasingly transparent, we should expect to see these powerful programmes more and more in everyday life.