B. Cui, P. J. Bex, N. Ross
We adapted the genetic-algorithm face paradigm introduced by Carlisi et al. (2021) and extended by Binetti et al. (2022), combining it with a three-dimensional morphable face model, to investigate how healthy young, healthy older, and CVL-affected individuals internally represent facial emotions.
Central vision loss (CVL) from age-related macular degeneration (AMD) profoundly impacts daily life. While much research focuses on visual discrimination and reading, understanding how CVL affects emotional perception remains unexplored. We adapted the genetic-algorithm face paradigm introduced by Carlisi et al. (2021) and extended by Binetti et al. (2022), combining it with a three-dimensional morphable face model, to investigate how healthy young, healthy older, and CVL-affected individuals internally represent facial emotions. Participants evolved faces they perceived to best express 13 distinct emotions (or 4 for the older healthy group). Using a three-cohort design, we separated age effects from CVL-specific effects. Results revealed four significant group differences: Awe Convergence (H = 10.97, p = 0.001), Shame Intensity (H = 10.04, p = 0.005), Interest Range (H = 6.35, p = 0.016), and Interest Stability (H = 4.86, p = 0.026). Additionally, cosine similarity analysis in 199-dimensional face space showed cohorts use fundamentally different facial feature configurations (mean cosine {approx}0.006 for YNG vs CVL), suggesting CVL may reorganize emotion representations rather than simply scaling them. These preliminary findings raise questions about whether all emotions require uniform amplification in CVL and point toward the potential value of selective rehabilitation over broad caricaturing approaches.