Human age estimation performance based on facial images: Potential implications for refugee processing

Abstract

Government agencies responsible for refugee processing are often challenged to develop an accurate identification profile for asylum seekers when reliable documentation is unavailable. As processing and support systems are designed differently for children and adults, it is critical to determine an accurate age estimate. However, current methodologies to estimate an individual’s age lack accuracy, verification or violate ethical standards, presenting the need to explore an alternative age estimation procedure. The present study aimed to explore human performance of age estimation, using facial images, for refugee processing purposes. As a within-subjects design, participants (N = 46) undertook a perceptual task to estimate the age of both children (11–17 years) and adults (18–24 years) from a facial image. Estimates were more accurate (i.e., closer to the true age) for children than for adults, although there was a consistent tendency to overestimate the true age. If this methodology was utilised for refugee processing, trends of over-, as opposed to under-estimation, increase the likelihood of a child being incorrectly labelled as an adult, than an adult labelled as a child. Future research could aim to further develop a task performance baseline by defining group-specific biases of estimators and estimations. The results of this study provide government agencies with an initial understanding of task performance and the potential biases in human perception.