Collecting Design Knowledge for Generative AI Systems

Sunday 7 July 2024, 9:00 am to 12:30 am

Workshop Chairs

Ye Wang (Autodesk Research) | ye.wang@autodesk.com

Voho Seo (Hyundai Motor Company) | voho@hyundai.com

Nicole Damen (University of Nebraska Omaha) | ndamen@unomaha.edu

Workshop Committee

Kosa Goucher-Lambert, Assistant Professor (UC Berkeley)

Alex Tessier, Director in Simulation, Optimization and Systems Research (Autodesk)

Murtuza Shergadwala, Responsible AI Research Scientist (Workday)

Aim and Content

Generative AI relies heavily on substantial and high-quality datasets to address domain-specific challenges. However, acquiring hand-crafted design knowledge proves challenging (Erichsen, Pedersen, Steinert, & Welo, 2016; Grudin, 2020), particularly due to the time constraints faced by designers. Designers often hesitate to invest additional time in documenting and organizing explicit design knowledge, such as sketches, models, and inspirational images. Additionally, tacit knowledge, such as trade-offs and prioritization, is typically shared informally among designers working on the same project, lacking proper documentation. Figure 1 illustrates three types of challenges in documenting design knowledge.

Figure 1: Challenges in documenting design knowledge

1. Sorting Challenge: Sorting design documents becomes a formidable task due to the varied and abstract design workflows, shaped by individual designers with diverse perspectives.

2. Processing Challenge: Handling car design documentation proves to be intricate, given the complex nature of the product comprising over 20,000 parts and employing diverse data formats throughout various developmental stages.

3. Collecting Challenge: The imperative of confidentiality and the preservation of originality in car design results in a decentralized and somewhat closed design process, presenting difficulties in gathering comprehensive and openly accessible documentation. 

The process of designing products is influenced by a range of factors, spanning from clear specifications to personal experiences. Assuming a structured design process, we can categorize these factors impacting the workflow and outcomes. The outer circle of the chart in Figure 2 represents instances of design knowledge, organized by sources: public data, confidential real-world data, and tacit knowledge. Public data, abundant and easily gathered, stands in contrast to the scarcity and difficulty of collecting confidential real-world data. The tacit knowledge category involves personal knowledge, inherently ambiguous and elusive. In our workshop, we'll delve into these knowledge categories, exploring techniques to collect and leverage them for generative AI.

This workshop aims to explore innovative interaction methods and techniques to address the difficulty of collecting design knowledge. Using a practical case study focused on designing a new car, the workshop will delve into various dimensions of design knowledge, such as metacognition (Lawrie, Hay, & Wodehouse, 2022), tacit and explicit (Tsoukas, 2012; Grandinetti, 2014), scientific and emotional Bratianu (2014), confidential and public (Ahmad, Bosua, & Scheepers 2014).

Workshop format

Participants will be divided into groups to collaboratively brainstorm, consolidate, and synthesize new approaches for collecting design knowledge within a design company context. The workshop will progress through phases of idea generation, refinement, and presentation. Each group will articulate a cohesive narrative illustrating how their proposed techniques can enhance the collection of design knowledge and subsequently benefit future design processes.

Schedule outline

9:00 Introduction (problem statement, key challenges, workshop objectives & agenda)

9:30 Discussion of existing challenges for collecting design knowledge

10:00 Idea generation in groups

10:30 Discussion & voting of ideas

11:00 Idea refinement in groups

11:30 Presentation of refined ideas

12:00 Workshop end

Submission

Interested participants are encouraged to submit the following ideas to the workshop chairs:

1. What aspects of design knowledge would you like to delve deeper into during the workshop?

2. Which design workflows do you envision being further enhanced by AI capabilities?

3. How can we overcome challenges related to collecting design knowledge?

These ideas will be integrated into the workshop, fostering a collaborative exploration of innovative solutions to the challenges of collecting design knowledge.

References

Ahmad, A., Bosua, R., & Scheepers, R. (2014). Protecting organizational competitive advantage: A knowledge leakage perspective. Computers & Security, 42, 27-39.

Bratianu, C., & Orzea, I. (2014). Emotional knowledge: The hidden part of the knowledge iceberg. Management Dynamics in the Knowledge Economy, 2(1), 41-56.

Erichsen, J. A., Pedersen, A. L., Steinert, M., & Welo, T. (2016, April). Using prototypes to leverage knowledge in product development: Examples from the automotive industry. In 2016 Annual IEEE Systems Conference (SysCon) (pp. 1-6). IEEE.

Grandinetti, R. (2014). The explicit dimension: What we could not learn from Polanyi. The Learning Organization, 21(5), 333-346.

Grudin, J. (2020). Evaluating opportunities for design capture. In Design Rationale (pp. 453-470). CRC Press.

Lawrie, E., Hay, L., & Wodehouse, A. (2022, July). A Classification of Methods and Constructs in Design Cognition Research. In International Conference on-Design Computing and Cognition (pp. 97-114). Cham: Springer International Publishing.

Tsoukas, H. (2012). How should we understand tacit knowledge? A phenomenological view. Handbook of organizational learning and knowledge management, 453-476.