Exploring Human-AI Interaction: Insights from Professor Per Ola Kristensson

Professor Per Ola Kristensson

Per Ola is a Professor of Interactive Systems Engineering at the Department of Engineering (University of Cambridge) and a Fellow of Trinity College, Cambridge. He leads the Intelligent Interactive Systems Group at the Cambridge Engineering Design Centre. He is also a co-founder and co-director of the Centre for Human-Inspired Artificial Intelligence at the University of Cambridge.

In Per Ola’s new course with Cambridge Advance Online, Human–Computer Interaction for AI Systems Design, he will share his systems thinking approach to designing human-AI systems with professionals from a wide range of industries. We sat down with Per Ola to ask him about the field of AI, his research interests, and what he hopes learners will take away from his course.

What is your favourite part of being involved in the AI field?

The most exciting part of this field is the sense of adventure in that you have a shot at inventing the future. It is thrilling to be at the forefront of a fast-moving field where there is a real possibility your research results can be felt by hundreds of millions of users in the future.

You’ve previously been involved in some of the world’s most pervasive AI systems, including swipe typing. Can you share what are you currently working on?

My current research interests are threefold. First, I am very interested in understanding how we can create efficient interaction techniques for augmented reality glasses that support depth-based real-time tracking of the wearer’s hands and fingers. It is an exciting research area as it requires considerable use of state-of-the-art machine learning and design flair to get the solutions to work right. However, when they work right, they feel almost like magic. For example, we now have systems that allow users to type in thin air, perform a wide range of 3D gestures that are correctly recognised among a stream of unrelated user motion, and so on.

Second, I have long-standing interest in using AI methods to improve accessibility, specifically AI methods that allow nonspeaking individuals with motor disabilities, such as users with late-stage ALS, to communicate reasonably effectively. For example, we have developed a commercialized system that allow users to type with their eyes without having to rely on dwell timeouts and another research system that automatically generates suitable sentences for a dialogue with a speaking partner based on what the partner has said, a persona of the user, and some information on the context.

Third, I am interested in improving the design of interactive systems by better understanding how we can build better computational design tools that allow designers to explore a wide range of options with the assistance of a computer without losing their sense of agency and control, in other words, to make the most out of both the human designer and the fact that computers can help explore vast design spaces. For example, we have done recent work on understanding how multi-objective Bayesian optimization can be used by designers to arrive at new solutions.

Why do you think human-AI interaction is an important area?

Human–AI interaction is already pervasive. You have it all around you, you just do not notice it as has become commonplace. Auto-complete on phone keyboards, web browser URL fields, search engines recommender systems, spam filters, automatic calendar invites in email apps, speech recognition, parking assistance, and the list goes on.

AI-infused systems bring a lot of potential, but they are also notoriously difficult to design as there are so many variables to consider; not only the complexity of the system itself, but the users’ interaction with an AI. To do this properly requires a systematic approach and this is where human-AI interaction is still in its infancy. However, if we do not get it right then we create systems that are expensive and inefficient at best, and dangerous or even life-threatening at worst.

What do you hope learners will gain from your course?

I want it to be a positive experience and give a sensation that you can change the future for the better. Because human-AI systems actually do this by, for example, automating repetitive or mechanical tasks, amplifying human input, and providing support for better decisions.

However, to do this you need to be clear about the problem you are trying to solve and follow a systematic approach. It may sound like this stifles creativity, but it is actually the other way around: systematic design methods open up so many things to consider and provides a structure for design, requirements, risk analysis, verification and validation.

Join Per Ola for his Human–Computer Interaction for AI Systems Design course, welcoming professional cohorts from July 2023.

Explore our range of online courses led by University of Cambridge academics, and learn from thought leaders at the forefront of academic research. Visit our website for more information: advanceonline.cam.ac.uk(Opens in a new window)