Sometimes we have to depend on philosophy to explain to us why something apparently simple is in fact extremely complicated. The way we use referring expressions – things that pick out the entities we want to talk about, such as “Mary”, or “that guy over there” – falls into this category, but is no longer just a matter for the philosophers; it’s complicated enough to require highly interdisciplinary explanation. In his book, Computational Models of Referring: A Study in Cognitive Science (MIT Press, 2016) Kees van Deemter approaches the problem from a computational angle, asking how we can develop algorithms to produce referring expressions that are communicatively successful, efficient, and potentially even human-like in their performance. He draws on a broad range of work from across cognitive science to address this question, and in doing so, also gives us an excellent example of how computational thinking can inform linguistic theorising. In this interview, we discuss several aspects of this work, including the role (and limitations) of the Gricean maxims, the challenge of audience design and shared knowledge, and how the salience of different properties of an entity can and does enter systematically into our choice of referring expression.