Designing the Characteristics of Design Teams via Cognitively Inspired Computational Modeling

Abstract

Teams are a ubiquitous part of the design process and a great deal of time and effort is devoted to managing them effectively. Although teams have the potential to search effectively for solutions, they are also prone to a number of pitfalls. Thus, a greater understanding of teams is necessary to ensure that they can function optimally across a variety of tasks. Teams are typically studied through controlled laboratory experiments or through longitudinal studies that observe teams in situ. However, both of these study types can be costly and time-consuming. Months, if not years, pass between the initial conception of a study and the final analysis of results. This work creates a computational framework that efficiently emulates human design teams, thus facilitating the derivation of a theory linking the properties of design problems to optimized team characteristics, effectively making it possible to design design teams.This dissertation first introduces and validates the Cognitively-Inspired Simulated Annealing Teams (CISAT) modeling framework. The central structure of CISAT is modeled after simulated annealing, a global optimization algorithm that has been shown to effectively mimic the problem-solving process of individuals. Specifically, a multi-agent analog of simulated annealing is used in CISAT to mimic the behavior of teams. Several additional components, drawn from the psychology and problem-solving literature, are then included in the framework to enable a more accurate description of individual activity and interaction within the team. CISAT is then used to investigate the relationship between design problem properties, team characteristics, and task performance. Multiple computational simulations are conducted in which simulated teams with various characteristics solve a variety of different configuration problems. These simulations are then post-processed to produce a set of equations that make it possible to predict optimal team characteristics based on problem properties, thus enabling the optimal design of design teams. To validate these equations a behavioral study is designed and conducted in which teams of engineering students interact at different frequencies while designing a complex system. Results of the study offer a limited validation of the predictive equations.This dissertation further highlights the resource efficiency and versatility of CISAT by demonstrating its use in two additional applications. In the first, a new numerical optimization algorithm is derived directly from CISAT by stripping away all but the most quintessential team-based characteristics. The team-based characteristics of this algorithm allow it to achieve high performance across a variety of objective function with diverse topographies. In the second application, CISAT is used in conjunction with Markov concepts to examine the order in which designers make changes to their solutions. Although it has been demonstrated that humans apply changes in a specific order (called a sequence) when solving puzzles, such patterns have not been examined for engineers solving design problems. It is shown that operation sequences are used by designers, and improve solution quality. This dissertation demonstrates how characteristics of individual designers and design teams can be captured and accurately reproduced within a computational model to advance our knowledge of design methodology. Future extensions of this work have the potential to inform a deeper and more holistic understanding of the search process.

Modelling the Dynamics of Influence on Individual Thinking during Idea Generation in Co-design Teams

“…Agent-based modeling often involves imbuing agents with the ability to learn about their environment. Some researchers have modelled social learning by means of mental models [15], or learning from doing the task [17] [13]. While simulating learning, it is often assumed that agents know the design solution space and therefore pursue optimal solutions.…”

Mining Process Heuristics from Designer Action Data via Hidden Markov Models

“…These observations show that designers in the cooling system study frequently moved between strategic states, whereas designers in the truss study tended to remain within a state for longer periods. A similar degree of operational restlessness is apparent in nonhidden Markov chain models that are trained on the same data [28].…”

“…This is known as first-order sequencing. While this may initially seem to be a limitation, other work has demonstrated that including multiple previous states does not significantly increase model veracity for Markov-based models in design applications [28]. In addition, evidence from studies of humans solving tavern puzzles has shown that first-order sequencing is important for effective problem-solving [29].…”

Influencers in design teams: a computational framework to study their impact on idea generation

“…Most of the models described in the literature deal with some form of learning in their agents to accomplish the purpose of their work. The most common logic implemented in many models listed above is in the form of learning from experience (McComb, 2016;Lapp et al, 2019). However, while simulating learning it is often assumed that the agents are aware of the design solution space and they thrive for the optimal solution (McComb et al, 2017).…”