How to use AI to enhance user experience
But many of the most promising examples of AI focus on people,designed around human interactions and serving to augment jobs rather than replace them. Stanford University Associate Professor Michael Bernstein conducted an on-demand webinar to discuss the ideal intersection of people and AI software, focusing his attention on the ways in which AI can create a better user experience (UX).
Avoid overpromising and under delivering
There are limits to what existing AI platforms can accomplish. When incorporating AI into a user interface (UI), it’s easy to fall into the trap of exaggerating AI capability to make a platform seem more user-friendly. Virtual assistants and customer support bots might have an intuitive UI, but if the underlying AI can’t do much more than point users to help pages or call center contact information, then people might not bother asking questions at all.
Instead of overpromising, it’s better to use the technology’s strengths to support enhanced UX. For instance, Professor Bernstein highlighted visual design solutions that use AI to create a heat map detailing which parts of the screen viewers are most likely to look at. Web and app developers can use those tools to lay out interfaces that place the right information and on-screen features in the best locations.
Account for uncertainty through adaptive interaction
Many everyday examples of AI in action have a high degree of certainty. For instance, if you ask Alexa, Siri or some other virtual assistant to set an alarm for a specific time, there’s really no way for the AI system to misinterpret that situation. Both the request and the desired action are clear.
That’s not always going to be the case, though. For tasks that don’t necessarily have such a definitive outcome, and UX designers need to understand how to properly use AI in these instances. Some developers have attempted to address this issue by incorporating adaptive interaction into their interfaces. The latest version of Microsoft Office, for example, uses an adaptive layout to customize the toolbar according to specific situations and past actions. The platform is, in essence, anticipating what might be relevant and helpful to the user and then pushing those features forward so they’re readily available.
Another common approach to uncertainty is to suggest or recommend actions to the user rather than have the system act on its own. Perhaps the most visible example of this is the content recommendation tools included in media streaming platforms like Netflix and Hulu. Their underlying AI systems gather information about users’ preferences and viewing history to suggest other content they might enjoy, including multiple options to choose from. Not all of those suggestions are necessarily going to hit the mark, but there’s a good chance that the user will find something new to watch.
Help your machine learning model produce better results
AI systems rely on machine learning models to recognize different input, carry out various tasks and continually improve their performance. There’s a lot of variability with these models that goes unseen by researchers and engineers, making it difficult to anticipate what results you’ll get.
App and web designers can help these systems perform better by looking at data and input through the eyes of their AI and machine learning platforms. If you understand how AI systems view different inputs, you can reinforce the data that’s most salient to them.
There’s a lot of ground to cover with a subject as rich and complex as AI. Anyone interested in this exciting field should watch the full webinar to learn more tips from Professor Bernstein on how to use AI to build better UX.
Thinking about switching fields to develop AI-enabled solutions of your own? Stanford’s Digital Transformation online program can prepare you for a career in AI, including a UX-focused course taught by Professor Bernstein. Enroll today to start pursuing your professional dreams.