People Marketing

Making Artificial Intelligence More Intelligent When It Comes to Language

, by Fabio Todesco
Dirk Hovy, Associate Professor of Computer Science, has won an ERC Starting Grant of 1.5mln euros. His project introduces demographic factors into language processing systems, which will improve algorithmic performance, avoid racism, sexism, and ageism, and open up new applications

What if I wrote that "winning an ERC Grant, Dirk Hovy got a sick result?". Those familiar with the use of "sick" as a synonym for "great" or "awesome" among teenagers would think that Bocconi Knowledge hired a very young writer (or someone posing as such). The rest would think I went crazy. Current artificial intelligence-based language systems wouldn't have a clue. "Natural language processing (NLP) technologies," Prof. Hovy says, "fail to account for demographics both in understanding language and in generating it. And this failure prevents us from reaching human-like performance. It limits possible future applications and it introduces systematic bias against underrepresented demographic groups".

Hovy, an Associate Professor in the Department of Marketing, obtained a €1.5mln Starting Grant from the European Research Council with his project INTEGRATOR (Incorporating Demographic Factors into Natural Language Processing Models). The research project wants to make the design of demographically aware NLP systems possible. "NLP is now used in translation, search, chatbots, personal assistants, business analytics, and many other common applications. So demographically-biased models create uneven access to these vital technologies", he says.

In a sense, current artificial intelligence language systems are not that intelligent. "They miss the intent of a speaker, and they don't adapt their language to the user," Prof. Hovy says. "If we ask a smart assistant why it gets dark in the evening, it always explains it in the same way, no matter whether the question was asked by a five-year-old or by a scientist. But a child and an adult require very different responses. Current systems assume that everyone speaks in the same way. That assumption and inability to adapt to different user groups encode demographic biases, leading to sexism, racism, and ageism. With this project, we will make systems demographically smarter. We will develop formal approaches to describe and measure demographic bias and prototype better systems".

"I want to render language technology about people again," Prof. Hovy concludes, "as when the field was initially influenced by linguistics and philosophy. Later, it restricted itself to technical aspects and feasible applications. It's like the person who lost their car keys but only searches under the streetlamp because it's the only place where they see something. If the keys are elsewhere, we'll never find them. We need to expand the light." That's why, in his ERC-winning project, Hovy will work with scholars in machine learning, sociolinguistics, sociology, demography, ethics, and philosophy, including three post-doc researchers he will recruit over the next five years.

The ERC was established in 2007 by the European Commission to promote basic research in Europe, with three highly-competitive grants – Starting, Consolidator, and Advanced – for top academics at different career stages. It does not impose specific topics, but requires high-risk/high-reward projects with a high scientific and social impact.

Professor Hovy's is the 37th ERC Grant hosted by Bocconi since the inception of the European program in 2007, and the second one in Computer Science.

Making AI Linguistically More Intelligent

Watch video