LAUSR

Affective computing has been shown effective and useful in a range of use cases by now, including human–computer interaction, emotionally intelligent tutoring, or depression monitoring. While these could be very useful to the younger among us—including in particular also earlier recognition of developmental disorders, usually research and even working demonstrators have been largely targeting an adult population. Only a few studies, including the first-ever competitive emotion challenge, were based on children’s data. In times where fairness is a dominating topic in the world of artificial intelligence, it seems timely to widen up to include children and youth more broadly as a user group and beneficiaries of the promises affective computing holds. To best support according to algorithmic and technological development, here, we summarize the emotional development of this group over the years, which poses considerable challenges for automatic emotion recognition, generation, and processing engines. We also provide a view on the steps to be taken to best cope with these, including drifting target learning, broadening up on the “vocabulary” of affective states modeled, transfer, few-shot, zero-shot, reinforced, and life-long learning in affective computing besides trustability.