- Thursday, February 28, 2019 - 2:00pm
- Gates-Hillman Center 4405
John Zimmerman, HCII, CMU (Co-Chair)
Aaron Steinfeld, RI, CMU (Co-Chair)
Carolyn Rosé, HCII and LTI, CMU
Saleema Amershi, Microsoft Research Designing AI Group
Technology often comes into being in the form of a technical advance, and without much concern for user experience (UX) design. As the technology matures, designers work to re-understand how it might be situated in the real world and to invent new product forms not imagined when it was first invented in an engineering lab.
It is difficult to see this kind of design innovation happening with AI. It is hard to trace how an AI technical advance produces such design innovations that operationalize the tech for new target users, new activities, and/or new contexts. When working with AI, designers reported challenges in almost all archetypal design activities (e.g., sketching, fast and iterative experience prototyping), on many familiar design issues (e.g., ethics and trust), spanning a wide variety of subject domains.
This thesis focuses on understanding AI as a design material. I ask: What exactly makes AI distinctive, and why is it so difficult to design? How can HCI and design research make AI easier to understand and manipulate into the innovation of products and services?
I identify two key characteristics of AI that shapes how it is designed and experienced: (1) multi-layered capabilities and (2) the decoupling between capabilities and system behaviors. The two characteristics together form a useful framework and provide structure for more effectively describing AI's design space.
Using the framework as a scaffold, I describe a set of Research through Design projects and discuss how the tools and knowledge emerging from these concrete design situations contribute to the broader challenges of innovating AI products and services. These projects include (1) designing a prognostic decision-support system that helps clinicians decide whether and when to implant an artificial heart to a heart-failure patient; (2) designing an intelligent text editor; and (3) for my proposed work, supporting the fan fiction writing-and-consumption ecology with language intelligence.
All are welcome! Please forward to others you think may be interested.
- Queenie Kravitz