Quantifying Design Hypothesis Certainty in Early-Stage Design Using Dempster–Shafer Theory

Modern artificial objects have become increasingly complex, and this complexity is mirrored in the design process itself. When critical design changes occur in downstream phases, there is a high risk of deterioration in quality, cost, and delivery owing to rework across related processes. Therefore, potential design changes must be predicted in the early stages of design and proactive measures should be taken. In the early design stage, the process is inherently based on fallible design hypotheses, and the fallibility of these hypotheses can lead to design changes in later stages. Accordingly, the certainty of each hypothesis must be evaluated by considering the evidence that supports it. However, design hypotheses are often supported by multiple heterogeneous pieces of evidence with varying degrees of support, and the information available in the early stage is typically incomplete. As a result, rationally evaluating the certainty associated with each hypothesis is not easy. To address this issue, this study proposes a transparent and systematic method to quantify the certainty of design hypotheses while accounting for incomplete evidential information in the early design phase. It first organizes the conceptual foundation of the evidence underlying hypothesis certainty and models it. Then, by applying Dempster–Shafer theory, a computational framework capable of determining proposition certainty from multiple evidence sources under incomplete information, we propose a method to quantify the certainty of design hypotheses. The proposed method is applied to hypotheses generated in a design experiment, and procedural validity and user evaluation were examined. This study introduces a new approach for managing fallible design knowledge based on Dempster–Shafer theory, suggesting a conceptual basis for the early detection and mitigation of risks associated with potential design changes.