A practitioner’s guide to creating useful, usable, desirable, and responsible AI products.
How do you design an AI-infused product? What are the best practices? Doesn’t designing an AI-infused product just mean designing a product? How do you design those products ethically and responsibly?
We, a group of designers and researchers posed these questions to ourselves individually, then collectively, as we experienced designing various AI products and realized that we have to question some fundamental assumptions about interaction design and product design when it comes to AI.
Why? Because AI systems are probabilistic and prone to errors, they change over time as they learn and as the models get better or evolve, and as a result they can violate established principles of traditional interaction design. Designing AI products require a holistic approach including consideration of data, models, systems and UX. Missing any element can cause harm. Yet existing principles for responsible AI are often one dimensional and fail to revisit and update the fundamentals …
So, we dove into over 20 years of AI-focused design guidance literature-research papers, magazine articles, and blog posts. We synthesized over 200 guidelines, best practices and recommendations and distilled them into a set of eighteen best practices for human-centered AI design, then validated the guidelines through rigorous research. the work is already framing the conversations across industries, and was recognized with an honorable mention award at CHI 2019, a conference that historically has defined the interaction paradigms of the future.
The guidelines provide recommendations on how to create meaningful AI-infused experiences that leave users feeling in control, and respect their values, goals, and attention.
The talk will tell the story of the guidelines’ creation, give an overview of the guidelines with examples, and most importantly, share our methods for using the guidelines in our fortune500 company’s product teams.
• How are they used to spawn a new product vs. How are they used to evaluate an existing one?
• How do the guidelines fit if you are designing a consumer AI product, vs a productivity tool?
• How do you balance the seamlessness of the interaction with the need for transparency of the AI system?
I’ll end the talk with a call for community engagement – we’d love people to share experiences of using the guidelines with the design community and help with pushing this work and our discipline further.
Designing AI-infused products should be done thoughtfully and responsibly, this talk offers everyone tools and methods to achieve it.
