🐴 AI is evolving fast - is UX ready?
My discoveries and reflections in August
☀️ Happy Tuesday,
I’m back from vacation! I traveled to many places, ate a lot of food, and reconnected with old friends I hadn’t seen in a decade.
It was also a timely break after finishing the first cohort of my AI course. I received great feedback and felt both grateful and relieved.
One thing I didn’t forget in August was to keep up with the ever-changing AI world. In today’s issue, I’ll share some recent thoughts—how UX is adapting to AI.
1. New Components
Use Perplexity as an example. What design patterns do you see on the main page?
There’s a search box, sample prompt questions, and even the “New Thread” button in the left navigation bar.
After I asked a question, more patterns appeared: the Sources section at the top and the follow-up question prompt in the search bar.
All these components are new patterns that have become more common with the emergence of AI.
Let’s take another look at ChatGPT’s interface:
More subtle patterns have emerged due to the growth of AI:
The ability to generate options and toggle between them.
The ability to switch between different AI models.
The ability to provide feedback based on personal preferences.
The source of the response.
The disclaimer about AI’s accuracy.
In the below interview, Henry Modisett also discussed some specific design challenges that Perplexity has faced at the intersection of AI and UX.
2. New Interfaces
The adaptation at the component level is just part of the equation. There are even more possibilities for the entire interface.
Palash Nandy, an Engineering Director at Google, gave a demo on Gemini’s adaptive interface:
Gemini was able to understand the context and Palash’s intent, allowing it to create personalized interfaces based on AI’s understanding of both.
Some examples:
Interesting, isn’t it?
And there’s more to come. As multi-modal technologies evolve, we’re going to see more tailored experiences with AI’s interpretation of voice and vision.
3. New Principles
User interfaces are what appear on the surface. Yet fundamentally, principles are what can help AI scale in a sustainable and responsible way.
A few years ago, Google released some case studies of real-world AI-driven products. The case studies themselves are somewhat outdated (over three years ago), but the human-centered principles they outlined are still relevant.
Here are some principles:
Determine if AI adds value.
Even the best AI will fail if it doesn’t provide unique value to users.
Set the right expectations.
Be transparent with your users about what your AI-powered product can and cannot do.
Explain the benefit, not the technology.
What do users need to know about how the system works to understand and use your product?
Be accountable for errors.
Understand the types of errors users might encounter and have a plan for resolving.
Earlier this year, a team from IBM Design also took a deeper dive into design principles for Generative AI applications.
One intriguing example was around co-creation—giving users the ability to control the outcome.
For instance, in DreamStudio’s left sidebar, users can control the amount and proportion of the generated images. This feature is also common in other generative AI tools.
So instead of letting AI take the control of everything, it gives human some degree of control.
It might feel overwhelming to keep all these principles in mind because of AI, but I believe there are no set rules. We need to use our best judgment and thoughtfully outline the principles based on our specific use case.
4. Further Reading
UX Design in AI by Ericsson
Transforming UX with Generative AI by Marc C. Seefelder
Artificial Intelligence Driven Design by Joël van Bodegraven & Collaborators
Thanks for reading! Hope you find it helpful.
☀️ Have a great week!
—Xinran
P.S. This is something I discovered and ate during my trip.
Guess what it is?
Such a great read! In your first section (New components), you shared an example of "switching between different AI models".
Do you have any additional thoughts on how that is useful to the user, especially someone who is not proficient about each model and its characteristics?
Answer revealed: It's raw walnut, freshly grown in the yard!