Designing for AI
Creating Minimum Viable Intelligence & Setting User Expectations
We are entering the age of intelligence—a time when technologists imbue artificially intelligent components into many products without a clear framework for how such intelligence is delivered to users consistently. It is a product designer’s job to make AI feel human-like and magical, not overwhelming and scary to users. When designing for Artificial Intelligence scenarios, whether for a large enterprise or small startup, setting user expectations is critical to deliver a reliable product. I will share some best practices from designing for AI in both large and small organizations. No matter the company size, a minimum viable product is important to design and not to be the result of unplanned feature cuts. I will share what Minimum Viable Intelligence is for an AI product, and how designers can deliver a clear UX when solving problems efficiently.
When thinking of how to design for intelligent products, first and foremost it needs to seem competent. Users must trust the AI agent or service with information and believe that it can achieve their goal. The bar for this depends on the expectations the designer sets. The most difficult thing about breaking out of scenario-focused AI is the lack of clear boundaries. Are you aspiring to create an entire conversational AI agent? Then the bar will be high, being that it is human-like in every way, including what it can respond to. A less intelligent Bot, however, will teach users the rails of its conversation early on to avoid disappointment. In this talk we’ll dig deeper into setting appropriate expectations when designing for AI across large and small applications.
Designing AI touchpoints from conversational interfaces to more traditional UI leads a designer to solve for how to best explain the capabilities of AI without overwhelming or frightening the user. This can be achieved by drawing on human interaction models. Responsiveness when users expect it is only one part of this equation. Apps that explain processes in human ways, like thinking, seeing, or reading, can benefit from showing users where they are in a process, while explaining it in natural ways. Emulating true intelligence takes more than just seeming alive and being basically competent though. To surpass users’ expectations can be a delightful moment when the product seems truly and independently intelligent.
I am the VP of Product at London-based AI startup, Context Scout. For the company, I have delivered a UX architecture to provide a connected browser that proactively searches using NLP and Machine Learning. Before that I spent the bulk of my career in Seattle, Washington, designing Cortana across devices, particularly Natural Language interactions with the assistant. I have designed multi-modal solutions for Windows, Holographic & VR, following my Masters in Interaction Design from Carnegie Mellon University. Before that I earned my BFA from Parsons, as well as a psychology degree from the New School.