Making the Weather - Reflections on Design for AI Symposium

By Phil van Allen - IDE Programme Coordinator

On October 14, 2022, the TU Delft Faculty of Industrial Design Engineering presented the first ever Design For AI symposium at the Next Delft building on the TU Delft campus. This event featured internationally known speakers and focused on the importance of defining the future of AI by “Making the Weather” through the most effective strategies for designing AI. In addition to thought-provoking keynotes, the day included informative and lively workshops on topics such as HCI in The Netherlands, discovering design issues in AI, and AI-enabled design research. There were also numerous opportunities for in-person discussion and networking in this growing, active research and commercial community.

The symposium was sold out and attended by over 150 designers, entrepreneurs, researchers and students, plus about 50 people watching the live-stream from 8 different countries. Sponsors included: Kickstart AI in collaboration with TU Delft AI Initiative and Mondai | House of AI at NEXT Delft.
Read more

The day started with a short introduction by Dean Caspar Chorus who discussed how designers are uniquely positioned to address the coming storm of technology innovation. Rather than “weathering the storm,” he suggested that we should instead, as designers, take the more proactive role of “making the weather.” We can do this by using design as an approach to ensure that the AI we get is AI that we need. So rather than regulating AI to fix it, we can design it better at the start to create preferable outcomes. He emphasised that this is possible because transdisciplinary designers like those at TU Delft IDE understand the people who will use AI, as well as the people who are making it.

Our keynote speakers provided a diverse set of perspectives on the design of AI, from the tech focused world of Google, Elizabeth Churchill questioned the very definition of AI, encouraging designers to be thoughtful about the language and methodologies they use for AI. Matt Jones brought years of deep design experience to argue that the metaphors we use for the design of AI are key. For example, he suggested that we think of AI not as a thing that has artificial agency, but rather as providing “Pedal Assist” like an electric bike does. Caroline Sinders argued that design should be an essential part of any policy development, since the application of policy will naturally involve design. As an example, she showed how GDPR failed frequently early on because some websites implemented “dark patterns” that seem to follow the policy by are in fact coercive. John Zimmerman discussed the reasons for a glaring gap in AI innovation, where about 85% of AI projects fail to deploy. In many cases, this is often because, his research at CMU found that the easier “low hanging” fruit of an application context are missed by the design team because the team has ambitions for AI that are too high.

Spread throughout the day were well-attended workshops. The Human Centered AI Systems Deep Dive explored the challenging issues of designing for AI through provocative experimentation with prompt-based text-to-image generators such as Dall-E 2. AI-enabled Design Research: Opportunities and Challenges explored the challenges and opportunities for AI-based design research, where AI can be used for the advancement of the research. The CHI Nederland workshop opened a dialog about a national agenda for Human Computer Interaction research in The Netherlands.