To establish a successful integration of humans into a cyber-physical system, many challenges have to be overcome in the context of the Tactile Internet with Human-in-the-Loop. For a seamless integration of humans, communication latency constraints and safety requirements have to be met. Those can only be ensured when there is a clear understanding of human interactions, which directly calls for computational models of human goal-directed and habitual behaviours. In this chapter, we report on recent insights from cognitive neuroscience and formal methods towards models of human behaviours. These cover contextual inference as a cornerstone for faithfully modelling the assumptions under which a human operates stochastic operational models to improve reasoning about human behaviours with rigorous formal techniques, and how low-latency constraints can be formally guaranteed within these assumptions and models. We also propose new modelling principles from neuroscience that bridge the gap between the human brain and computation.