Children as storytellers and mindreaders
How do children learn to see the world through someone else’s perspective? Max van Duijn, assistant professor at the Leiden Institute of Advanced Computer Science (LIACS), receives 24,167 euros from the Elise Mathilde Foundation and the Leiden University Fund (LUF). With this grant he will set up a pilot study, in which he wants to collect children’s stories using a mobile app.
Van Duijn hopes to gain insight into how children develop as storytellers, and how they learn to empathise with the perspective of others, also called 'Theory of Mind' or 'mindreading'. This insight has important applications in the humanities and developmental psychology, as well as in the field of Artificial Intelligence.
Telling stories in an app
Van Duijn uses storytelling to study empathy. ‘In stories, the narrator has to ‘mindread’ the characters who experience all kinds of events in the storyworld,’ he says. ‘Stories thus form a natural environment to study this skill.’ Together with students, Van Duijn will develop a mobile app in which children aged four to twelve can record stories together with their parents. ‘Four-year-olds often tell stories that seem incoherent by our standards as adults. But at the age of ten, they do a lot better,' he explains. ‘What is happening in their development in these years, and how is this reflected in the stories that children can tell?’ After collecting the stories in a database, for his analysis Van Duijn will use a traditional qualitative framework combined with quantitative and computational methods.
Van Duijn uses his grant primarily to set up an infrastructure for the collection of the data. First, he will develop an app for tablets and smartphones, in which children can tell stories to an avatar – a virtual character. In the app, they can safely store the stories, replay them and – possibly with their parents – submit them for research purposes. ‘On the longer term, I hope to have hundreds of stories of about 400-600 words each, spread over the different age groups. But before that, the infrastructure for this project must be put in place.’ In the next phase, Van Duijn wants to scale up the data collection and partly automate the analysis of the stories. ‘With the help of digital tools, it may be possible to convert a spoken story into a transcription automatically, from which various features can then be extracted. But for the time being, the stories have to be annotated manually before starting the data analysis.’
Empathy in computers
One of Van Duijn's inspirations is a classical question from artificial intelligence: How to create an intelligent system that can reason about other intelligent systems, including itself? Or, in other words, how do you model empathy in an artificial brain? ‘One reason why a 'conversation' with Siri easily gets awkward, is that such an artificial assistant is hardly able to read minds,' says Van Duijn. ‘In the near future a lot is about to happen in the field of artificial intelligence in Leiden. Also in that light it is important to better understand how children learn to read others’ minds, because this can serve as a source of inspiration for how to model empathy in an artificial brain.
Van Duijn states to be very happy with this grant. ‘I am extremely grateful to the Elise Mathilde Foundation and LUF. They support me in the crucial phase of this idea: taking the first steps and building the infrastructure for data collection. I appreciate the fact that they have an eye for the potential of a project. When it comes to applying for a VENI grant - which I will be doing in the next year - it is good to have something tangible in advance. So I'm very pleased with the opportunity the Elise Mathilde Foundation and LUF are giving me.’
Once a year the LUF Committee for Academic Expenditure (CWB) awards grants for scientific projects of Leiden University researchers. These grants for academic talent are often an important step towards grants by NWO and other institutions.