Midcourse Correction
A new decision-making tool can help guide engineering students toward more innovative designs.
By Nur Özge Özaltin, Mary Besterfield-Sacre, and Renee M. Clark
Today’s organizations must meet or exceed customer expectations through product innovation to capture and retain market share. Thus, learning to design innovatively is a crucial skill for engineering students to acquire. Given this, might engineering educators benefit from a decision-support tool that offers recommendations on student design-process activities that are associated with innovative outcomes?
In our research, we developed a Bayesian network decision tool using GeNIe modeling software, which can assist instructors by identifying activities at various points in the design process that are more apt to yield a breakthrough product. Our tool is based on actual design-process and outcomes data reflecting roughly 18,000 hours of design work collected from senior bioengineering capstone teams at two universities. Since Bayesian networks allow both downstream and upstream inference, the likelihood that a design-process category (conceptual design, for example) was performed at a certain level at a certain point can be determined for an innovative (or non-innovative) design outcome. This type of what-if analysis can be used by instructors to advise students on desirable design-process activities.
Design-process activities were chronicled using a survey system over a 24-week period as teams advanced from an initial concept to a working prototype of their medical-product designs. Activities included determining the target customer, customer needs analysis, brainstorming, research activities, design modifications, design and prototype reviews, and documentation of the design. The process-level data were used to build (set parameters for) the Bayesian network model. The structure of the network was based on Dym’s design framework, with the addition of two product-realization categories.
We connected each team’s process-level data to the innovativeness of its product, as assessed by the capstone instructors. Using both innovative and non-innovative outcomes data and the associated process-level data, our model was developed to contrast processes associated with innovative versus non-innovative products. This type of information can be used by design educators to guide their teams and make formative assessments and recommendations. Given the length of the capstone projects, we developed separate Bayesian networks for the individual phases of the design – early, middle, and late. The cross validation and sensitivity analysis we performed showed our model to be both accurate and robust.
To demonstrate the GeNIe software and our models, we created two videos. The first demonstrates how to use a GeNIe model to input information, such as the knowledge or assumption that a product is innovative (or not), and subsequently obtain an output – for example, the most probable usage level of a particular design category in a given design phase. The second video describes how educators can use our Bayesian network to guide their design students. It demonstrates the switching of a model between innovative and non-innovative outcomes settings to uncover contrasts in the most probable usage levels of the design categories. These comparisons provide a means to directly advise a design team. For example, if a hypothetical design outcome is toggled from non-innovative to innovative and there is an associated change in the most probable usage level of, for example, problem definition in a certain design phase, empirical insight into recommended process changes can be obtained and used to inform team activities.
Since the GeNIe software is free to download, design educators can experiment with our models and conduct what-if analyses to determine the probable impact of various usage levels of Dym’s design categories on the innovativeness of the product and vice versa. Our GeNIe models are available upon request from the authors. Educators also can use our methodology as described in our paper to develop models that may be more specific to their disciplines or situations.
Nur Özge Özaltin received her Ph.D. from the industrial engineering department at the University of Pittsburgh, where Mary Besterfield-Sacre is an associate professor and director of the Engineering Education Research Center in the Swanson School of Engineering, and Renee M. Clark is director of assessment. This article is excerpted from “An Engineering Educator’s Decision Support Tool for Improving Innovation in Student Design Projects” in the Summer 2015 issue of Advances in Engineering Education. (This research was supported under an NSF BES RAPD collaborative grant, award number 0602484.)