I read an interesting article about AI where 6 UX Principles were laid out, that article is AI and cognitive offloading: sharing the thinking process with machines
And the 6 principles are:
- Understanding user pain points
- Promoting collaborative problem-solving
- Using existing mental models and patterns
- Leveraging progressive disclosure
- Ensuring graceful failure
- Contextual guidance
The article caught my attention because it simply and succinctly laid out principles to work with. This is very useful information, specially when UX is not my expertise. I’m just an engineer. But the article put in front of me an approachable way to tackle the problem.
We all know how every company out there is slapping GenAI to their products. We are no different. And like everybody else, we are finding very hard to make up a good use case for it. Where and how should we integrate this new technology to truly add value for our customers? That is the million (billion?) dollar question.
One aspect of the answer has to do with UX. How should the user interact with the technology to get the most value out of it. And to me, and my colleagues, this is a very hard answer to come up with. Putting a chat interface on top of some existing feature just doesn’t cut it, really. And this is not even about how to prevent hallucinations and making sure that the outputs of the LLMs are accurate. No, this is “just” about how to build a product experience that makes sense.
Here is a short summary of the 6 principles:
Understanding user pain points First off, you need to solve an actual problem for the user. Seems obvious, but in this AI FOMO times it has to be re-stated. The article suggest focusing on reducing cognitive load for the user. So make sure not to add more by forcing them through a clumsy experience.
Promoting collaborative problem-solving Collaboration that reduces cognitive load is very valuable. Look to help the user get unstuck, rather than adding barriers to achieve their goal. Help the user get in the flow of what they are trying to achieve, and keep them in the flow with timely help.
Using existing mental models and patterns Leveraging existing mental models of how things should work reduces friction to adopt your product. This could be as simple as using industry established clues, like the ✨ icon.
Leveraging progressive disclosure Now, this is a new one for me. You don’t need to overwhelm the user with options and tools all at once. Show them the tools they need at the time they need it. It can be shortcuts to get started with a higher goal. Or leveraging wizard-like flows.
Ensuring graceful failure We all know about this one. But in the age of hallucinations it becomes that much more important. Specially when the product relies on the output of an LLM to move forward. You might encounter that the response did not conform with the requested format. Or be incorrect altogether.
Contextual guidance Finally, don’t expect your users to be prompt engineering experts. Understanding how to better interact with a GenAI system is not easy. Guiding the user on how to make the most of the system at a given point goes a long way.
In conclusion, to make the best of this GenAI technology we need to pay attention to UX more than ever. All these principles are nothing new, but they take a new meaning when applied to a new technology. Overall a great article and a lot for me to learn from. It’s always good to have map to navigate new terrains.