Decision-Guidance Systems


Why do we need them?


An increasing number of decision-guidance systems need to guide human decision makers to move a complex process toward desirable outcomes. Examples include finding the best course of action in emergency, deciding on business transactions within a supply chain, and developing public policies guided by most positive outcomes. 


The abundance of information is a blessing and a curse at the same time for a modern decision maker. Contemporary information systems are relentlessly efficient to collect and process huge amounts of factual data, but they are weak in terms of insights and wisdom. While human decision makers have plenty of useful intuitions, insights, judgments, and preferences, they can be quickly overwhelmed by the amount of dynamically available data. Typical operations research approach involves mathematical optimization to find the best course of action, but this requires a formally defined model, which defines the search space and the objective. Such a formal model is often not available a priori, but rather need to be extracted from historical data, and iterative dialog with human decision makers. Approaches to move information systems toward intelligence comparable to human's have only seen small progress, while humans are notoriously hard to change and upgrade!


What do they do?


Decision-guidance systems support an iterative process of giving actionable recommendations to and extracting feedbacks from a human decision-maker, with the goal of arriving at the best possible course of action. This needs to be done (1) in the presence of large amounts of dynamically collected data, (2) while learning objectives or decision preferences from historical data and decision maker's responses, and (3) under diverse constraints that capture complex real-world processes, composite services, and supply and demand.


Research focus


The technical tools involve (1) mathematical and constraint programming, (2) database management, and (3) statistical learning and data mining to extract user preferences and objectives.  Research on Decision-guidance focuses on the development of models, languages, and, most importantly, algorithms toward the first Decision-Guidance Management System (DGMS), which is productivity tool for fast development of decision-guidance applications. This is analogous to Database Management Systems (DBMS) serving as a productivity tool for fast development of database applications.