Models play a central role in understanding complex systems such as humans. However, cognitive science and psychology face important challenges in the development and testing of large-scale, complex models. The approach advocated here tackles these challenges by embedding the models in a real-world context that focuses on the functional aspect and the generation of overt behavior.
This Research Topic centers on the methodology of understanding systems by building them, specifically the construction of autonomous computer-generated humans as a research methodology. It is directly supported by the dramatic increase in the graphical quality of computer-generated humans: virtual humans appear indistinguishable from real humans, providing a unique opportunity to push more realistic cognitive and behavioral models. We aim to promote a novel paradigm for developing and testing large-scale complex models. In building models that drive artificial humans, we are asking questions relevant to the understanding of the human mind. These complex models will support a larger scope and wider-ranging predictions compared to reduced models, connecting the phenomenology of psychology and neuroscience’s aetiological view.
Modeling in this field faces two important challenges: Firstly, purely theoretical models tend to be unspecified and inconsistent. Secondly, the complexity of large-scale models and the fact that many models do not yield predictions at the neuronal level makes them difficult to test. The standard procedure to test a model is to measure reference data from the modeled system and compare it to the behavior of the model. One key problem with this approach is missing behavioral reference data, mainly because (analytical) empirical investigations ask different questions than the (synthesizing) modeling approach. Unlike similar approaches that use virtual humans as a means of social stimulus delivery, the emphasis of this Research Topic is on the use of virtual humans to embody models and testing them in real-time interaction. The cornerstone is that the model’s quality is assessed by the quality of the interaction between the virtual human, controlled by the model, and the biological human.
Examples of this approach can be seen in applying generalized gesture generation systems, emotional models, social interaction models, to more specific phenomena such as modeling empathy, social gaze, or rapport. They have significantly improved our understanding of human behavior by allowing us to test and refine our hypotheses in a much-controlled way that would not be possible with human subjects. With the recent advances in computational power and efficient algorithms, the next step in virtual human research should be creating more complex models to combine many methods and phenomena in virtual agents for implementing and evaluating models of human behavior.
We welcome original research, reviews, commentaries, and perspectives on topics which advance our understanding of cognitive processes via computational modeling techniques, especially in virtual humans, such as
• Work that showcases the approach of embodying models in systems for models testing
• Methods for implementing realistic virtual humans and embodying models
• Methods for quantifying human to virtual human interaction, i.e. the evaluation of cognitive, affective, and social models.
• Theoretical papers contributing to the proposed testing methodology
The Topic Editors would like to thank
Dr. Özge Nilay Yalçin for her contribution during the initial stages of creating this Research Topic.
