Understanding Interpretability in Modeling Biotechnological Processes: A Critical Review

In the field of process engineering, mathematical models play a fundamental role in better understanding systems and developing model-based systems for decision-making. These models are representations used to capture the key features of a system or process for a specific purpose, such as knowledge generation, estimation, control, optimization, and others. However, the validity of these models depends in some cases on the scope defined by the applicable theory, law, or principles, the experimental dataset used for their development and validation, as well as on the ability of these models to interact with their end users. In this sense, interpretability is an essential property that determines to what extent a human operator can interpret the model. However, the definition of this property is not clear, as there is no consensus on the concept in the literature. In this paper, we provide a state-of-the-art review of the various definitions of interpretability available in the literature as well as a critical analysis of the lack of uniqueness of the interpretability concept for the mathematical modeling spectrum, specifically in applications associated with biotechnological processes. This work is not an exhaustive study of the literature, but rather an approach to the existing differences between the definitions of interpretability and a proposal for the generalization of the concept.