Cooperative Decision-Making to Minimize Biased Perceived Value Effect on Business Process Decisions Using Partially Observable Markov Decision Processes

Deciding what options to take while a business process operate is recurrently challenged by the problem of incomplete information that is available to the decision-makers, for instance, business processes that include manual tasks increase the risk of workarounds due to unsupervised activity. To minimize this problem, this paper explores to what extent a machine decision-making process could support cooperatively the human decision-making process. Provided that the core elements of business processes are humans, and that decision-making is mostly a human-based task, any decision support system design need to account with this human perspective. An agrifood case study is the artifact used to study the design principles for such endeavor, and the following activities are reported: a controlled experiment to assess the human perceived value, a machine generation of value grounded in Markov theory using a partially observable Markov decision process, and a qualitative and quantitative comparison between the former results. Results indicate that significant difference exists between the machine valuation and the human value perception due to the value perception that evolves with experience throughout time. Furthermore, the human value perception collected in this experience could serve as a basis for a training dataset to a new machine decision-making tool.