Anchorlogy: An Ontology for Anchoring Bias Detection in Forecasting

Anchoring bias is one of the most prevalent biases within forecasting. It distorts managers’ estimations whenever context-driven intervention to the statistical model output is required. Consequences extend beyond a single organization since forecasting affects order quantity decisions and, therefore, the relations among suppliers, potentially generating a bullwhip effect throughout the supply chain. Anchoring bias can have a significant impact, and despite being related to a numerical value, its detection is very complex. Moreover, it tends to be recurrent when the context that caused the distortion is not explored and precisely understood. Current detection approaches are incomplete, as they do not make explicit the directional component of anchors or their meaning to the decision maker’s mental heuristics. In this work, we present Anchorlogy, an ontology devised to explicitly provide the required context to detect and mitigate anchoring bias during a decision-making process, and a metrological approach to measure it while addressing the deficiencies found in other metrics in the current psychological literature. Our proposal was validated by applying it to two case studies in the forecasting domain, and the results show that it effectively prevents the bullwhip effect in real-world scenarios.