An approach to the extraction of preference-related information from design team language

The process of selecting among design alternatives is an important activity in the early stages of design. A designer is said to express design preferences when assigning priorities to a set of possible design choices. However, the assignment of preferences becomes more challenging on both a practical and theoretical level when performed by a group. This paper presents a probabilistic approach for estimating a team’s overall preference-related information known as preferential probabilities that extracts information from the natural language used in team discussion transcripts without aggregation of individual team member opinions. Assessment of the method is conducted by surveying a design team to obtain quantitative ratings of alternatives. Two different approaches are applied to convert these ratings into values that may be compared to the results of transcript analysis: the application of a modified Logit model and simulation based on the principle of maximum entropy. The probabilistic approach proposed in the paper represents how likely a choice is to be “most preferred” by a design team over a given period of time. A preliminary design selection experiment was conducted as an illustrative case example of the method. Correlations were found between the preferential probabilities estimated from transcripts and those computed from the surveyed preferences. The proposed methods may provide a formal way to understand and represent informal, unstructured design information using a low overhead information extraction method.