People detect painful expressions more easily in members of their racial ingroup than outgroup. Here, we wanted to investigate this racial bias with a machine learning model trained to detect activations of different action units of painful facial expressions. We examined whether the model detected higher action unit activation for European than African faces when trained on datasets with mostly White faces. To control for confounding variables, pictures of faces were generated with the FaceGen Modeller. Results revealed that there exist differences in the visual detectability of some facial muscle activations due to skin color or other race-dependent facial features. Despite the bias towards European looking faces in the training data, some activations were more easily detectable in African faces. Thus, neither the perceptual detectability, nor the larger exposure to own-race faces seems to solely explain the racial bias in pain detection.

Learning to write is characterized by bottom-up mimicking of characters and top-down writing from memory. We introduce a CNN-RNN model that implements both pathways: It can (i) directly write a letter by generating a motion trajectory given an image, (ii) first classify the character in the image and then determine its motion trajectory `from memory', or (iii) use a combination of both pathways. The results show that, in one-shot and few-shot learning, the model profits from different combinations of the pathways: The generation of different character variants works best when the top-down is supported by the bottom-up pathway. Refilling occluded images of efficiently learned characters works best when using the top-down pathway alone. Overall, the architecture implies that a weighted merge of bottom-up and top-down information into a latent, generative code fosters the development of compositional encodings, which can be reused in efficient learning tasks.