Infant categorisation in naturalistic conditions: The role of language in real-world concept formation

Categorisation is a fundamental skill underlying generalisation of learned knowledge.

But how do infants form categories over experiences that are separated in space and time, where direct comparison is not possible?

While these processes, evident as early as 4 months after birth [1], are well-understood when studied in rigid familiarisation / novelty preference procedures [2], we know little about learning in naturalistic contexts and the role of the caregiver in structuring such scenarios, in particular using verbal labels.

In my PhD, I propose to investigate these processes in an interdisciplinary project that combines innovative experimental infant studies with advanced quantitative methods (AQM) leveraging machine learning.

One particular aim is to develop methods for the analysis of complex audiovisual recordings of parent-child interactions that will be made available to researchers in the field.Study 1 will use head-mounted eye tracking to examine the toddlers' first-person experience when encountering objects belonging to the same category across different contexts (exploring spatially separate rooms).

Advanced methods for investigating sequential properties of the caregiver-child-object interactions will be employed to understand links between experiences over time and the relationship between attention and learning outcome.

Study 2 aims for a big data approach using parent-contributed videos of natural play sessions at home and will employ machine learning and techniques from Natural Language Processing for analysis. This part of the project will allow investigating individual variation and differences in socio-economic status.