As AI becomes more widespread and deep learning models are integrated into various aspects of industry, scientists and engineers face the challenge of integrating human operators and end users into the design and operation of cognitive cyber physical systems. This challenge is exacerbated by the fact that many deep learning models employ black box approaches that lack transparent, human-interpretable algorithms. This paper addresses the challenge faced by proposing a human cyber physical network model, inspired by the human brain. In contrast to contemporary deep learning models, ours leverages a vector symbolic architecture to interactively learn human behavior and to develop cognition within the networks. To test this exploratory model, a two-part simulation is conducted. Using a modified industrial human-machine interaction dataset, we create structural networks of human, cyber, and physical object representations. These objects and their situational contexts are then encoded as hyperdimensional vectors. With context-dependent thinning, our model builds analogical episodes featuring distributed, parallel associative memories. The proposed model is shown to have analogical reasoning capabilities, with object nodes learning structural characteristics such as their hierarchical, or part-whole relationships within a network. Functional characteristics, such as human motion patterns and is-a relationships are learned as well. The model’s accuracy and performance can be transparently audited using established algorithms from network science. The results in this paper indicate that by designing systems as brain-inspired networks of human, cyber, and physical objects, vector symbolic architectures can be used to learn their structure and function by human-interpretable methods. Thus, the accountability inherent in the proposed model increases AI explainability in human cyber physical systems.