Nur Yildirim

Nur Yildirim headshot

Carnegie Mellon University

Advances in Artificial Intelligence (AI) have enabled a myriad of unprecedented technical capabilities. Leveraging this collection of technologies with output uncertainty that we broadly refer to as AI to build and deploy applications has never been easier. While building AI systems is getting easier, situating them in the real world in ways that are beneficial for humans remains extremely challenging. Today, the majority of AI initiatives fail, often because innovation teams do a poor job of envisioning and selecting concepts to develop. They choose high-risk projects that may or may not be valuable while overlooking low-risk, high-value opportunities. As AI capabilities become readily available and commoditized, critical questions arise: What should we build with AI? How do we effectively identify AI use cases? I argue that this breakdown stems from the current innovation process for AI – a place where design innovation can play a critical role in successfully transitioning these technologies into everyday life.

My research investigates the use of design in early phase AI development to discover the right things to build with AI. I develop resources and processes to support innovation teams in ideation, selection and formulation of AI concepts. These resources sensitize teams to AI’s capabilities and limitations, so that team members and stakeholders can effectively collaborate and engage in the conceptualization and development of AI technologies, regardless of their background (e.g., product manager, UX designer, software engineer, domain expert, policymaker, and others). This approach grounds AI innovation in existing AI capabilities and actual human needs. It enables teams to systematically explore the problem-solution space to identify low-risk, high-value concepts before selecting what to build. To inform my research, I often collaborate with innovation teams in industry and academia. In turn, this contributes to design practice and curricular development to foster best practices for envisioning AI products and services.