(6ii) Systems Approach to Advanced Decision-Making in Chemical Engineering, Biomanufacturing, and Society

Like natural resources, data can be scarce or abundant. When experiments are costly to perform, one can use the design of experiments (DoE) to generate data efficiently. While DoE is a well-studied subject, from my experience working with practitioners in the biopharmaceutical industry, most experimental designs are still heuristic-based. This gap between practice and available DoE techniques creates opportunities for the industry to adopt the more systematic, automatic, and model-based approaches. The abundance of historical data and the lack of data/metadata management, on the other hand, create challenges to reproduce experiments and retrieve process parameters for model development. The transition toward better data infrastructure requires a collective effort from chemical, systems, and software engineering.

Complex systems are my core research interest. I am particularly interested in the identification and prevention of unintended negative consequences in complex systems. The positive benefits of rapid technological developments often come with unintended negative consequences to society. It is generally difficult to predict unintended consequences. Nonetheless, chemical engineers have been practicing techniques such as process hazard analysis (PHA) to identify potential vulnerabilities in a process system. The healthcare sector also adopts similar “checklist” approaches to prevent unintended negative consequences. I plan to further expand this approach to policy-making and design of emergency response protocols.

The human factor is something we engineers tend to overlook, yet it is a crucial part of engineering systems. Two aspects of this topic are worth studying. First, when stakeholders respond to signals in an engineering system, the response can become feedback and create a closed-loop situation. It requires both control theory and game theory to study the closed-loop behavior and design mechanisms such as variable pricing that are both robust and efficient. Second, one can actively involve human decision-making in the human-in-the-loop computing configuration where algorithms proactively suggestion solutions while humans safeguard the integrity of machine-generated decisions and provide feedback.