Intelligent Adaptation of Difficulty and NPC Behavior in Serious Video Games for Learning

User-centric adaptation of serious video games continues to be a very important issue because of its benefits, such as enhanced motivation and engagement of individual players. It is based on player/learner characteristics that can be measured, estimated, recognized, or found by classification or clusterization. The paper suggests a new approach for dynamic, user-centric tailoring of task difficulty and the behavior of non-player characters. The approach is based on the emotional state and the shown outcomes of the individual player. Recognizing the current emotional state is based on facial expression analysis by convolutional neural networks and on an analysis of physiological data measured by sensors while playing the game. There are provided examples of serious games of learning with a dynamic adaptation of task difficulty and non-player character behavior. Copyright (C) 2024 The Authors.