SUMMARY

The objective of this proposal is to study the framing of design decisions and its impact on design choices and outcomes. For this proposal, a design frame is defined as the combination of a specific design space, set of modeling assumptions, and search strategy. Design frames are important because they influence which artifact is selected. In current design practice, designers commonly use optimization methods to select the best design artifact according to one or more specific technical objectives, such as efficiency or performance. However, such technical attributes rarely reflect the true preference of the designer. Alternatively, the objective could be a comprehensive, scalar value function, reflecting the decision maker’s preference as a function of multiple technical and economic attributes. In addition to the value from the artifact, such a value function should also reflect the considerable costs of modeling and solving the optimization problem. Because the framing of a design decision affects both the value of the artifact and the cost of the design process, the design frame should be selected to obtain an acceptable tradeoff. This knowledge about how best to select a design frame is currently lacking.
A model is developed to study how a design frame influences the decision maker’s preferred outcomes. This model, called the design decision framing model (DDFM), allows for an unbiased comparison of design frames. Avoiding bias is non-trivial due to the potentially biasing impact of modeling assumptions on the estimated artifact value. The DDFM achieves a fair comparison using a simulated, omniscient supervisor who asses the value of design frames based on the same “true” value model.
Using the DDFM, experiments will be conducted to study the influence of design frames on design process value. The experiments will focus on gaining design framing knowledge for real options problems. In real options problems, the option to modify the artifact at some point in the future is explicitly considered. Although a real option may require a substantial upfront investment, it often results in a substantial increase in the expected value. This is due to the flexibility of the design to react to uncertainty. However, considering real options greatly increases the computational complexity of the design problem, substantially adding to the design process costs. For this reason, one area where heuristics can add value is that of real options problems.