Generative AI strategy, user experience, and consulting. San Francisco, California

Design Theory

The design community has been slow to adopt the changes wrought by generative AI. I find this unusual but not entirely surprising — and not because designers harbor some specific fear and loathing vis-a-vis AI and its synthetic "creativity." Language models don't present a design surface, navigable information space, interactive user experience, product or even software interface that the design community is used to working with and on. Conversation, arguably a valid user experience, navigation and negotiation of functions, tasks, and actions, and qualitative stretch of time and cognitive investment of attention and interest, is not easily diagrammed or wireframed. What formalization has been applied to conversation as a form of action and interaction, NLP has handled, not UX.

And yet the success or failure of generative AI commercially, and with end users and in customer facing applications especially, will likely come down to user experience. How well do users understand the AI, what do they expect it to do, how well does it perform, and does it easily satisfy user needs? These and many more AI-specific user experience issues will shape and inform reception of AI generally.

Designers therefore ought to be invested in framing their own persepctives on generative AI. We should have heuristics ready for different kinds and uses of AI. We should have modified customer journeys, personas, service design blueprints, and more that reflect and capture the unique uses of AI. Our very design theories need to be updated to reflect the intrusion of AI into the digital domain, where generative models are disrupting long-standing applications and services from search to customer service chats, and image, video, and music creation. The core concepts of human-centered design, from utility, use, needs, requirements, to ease of use and satisfaction need to be reconsidered and reframed. It's possible that AI will soon challenge us to consider it human-adjacent or even post-human. When intelligence is post-subjective, non-human but human-informed and steered, what concepts will have have for design and knowledge, design and trust, design and relationships?

While design theory at the macro level needs to accommodate AI, I find that conversational AI in particular presents a unique need for new design concepts. I believe that conversational AI should be designed to mirror, reflect, and employ learned human linguistic competencies as much as possible. That is, humans should be able to communicate (texting or speaking) to AI naturally. AI should be designed to "understand" users as best as possible. We know this is more than a straightforward challenge. Humans don't communicate clearly, and AI doesn't understand its own use of language. And yet language models perform adequately in many situations and in cases can be designed to sustain long interactions with users with reasonable success.

A communication-centric approach to the design of language models takes the view that AI should at least try to approximate the two-sided, or mutual and intersubjective, nature of human communication. Current research into reasoning in language models focuses on the model's own reasoning and reflections on its reasoning. In human communication, we reflect not only on our own reasons but on those of the Other as well. There's what's called a "double contingency" in human communication, captured in the image that "I know that you know that I know..." Language models have been pre-trained to assume a common ground lingustically with users, but much research takes a monological not a dialogical view of user interaction. Design theory could be useful here if it can explicate the ways in which user experience needs to be better reflected in language model customizations. Aspects of communication related to understanding, empathy, emotion, common ground, interest and interestingness, relationship and trust, even co-presence — all of which belong to human social interaction — can be of use in shaping conversational AI design.