‘Not on the Exam’ with Prof Huili Chen

In this ongoing series, we ask our professors not just about their work, but also to tell us about their favourite cuisine and whether they prefer cats or dogs. This latest installment features Assistant Professor Huili Chen, who arrived at the Faculty of Information last January from Princeton University where she was a Presidential Postdoctoral Research Fellow. 

What courses are you teaching? 

I taught one course last term, INF2169H Exploratory User Research. And this fall, I’ll be teaching the introduction to UX design course, which is the required course for first year students. During the winter term, I will be teaching my elective, which will be about human cognition, behavior and culture in AI-mediated environments – or how AI impacts how we think, create and connect as human beings. 

What’s a skill that students leave your classes with? 

We are learning how to learn and to answer those timeless questions about the human condition. Technologies are only the tools that may help us to answer or shed light on those timeless questions.  

What big question drives your research right now? 

From the beginning of human civilization, we have invented tools – from pens and pencils to cave drawings, and, later, the internet, archives and libraries. We call these cognitive tools because the tools we create, in turn, shape how we think, create and connect. The big question I study is this interrelationship between the cognitive tools we invent and our cognition and culture, how they co-shape each other.  

What first sparked your interest in this field? 

I’ve always been interested in human behavior and cognition – how we navigate an incredibly complex world with limited cognitive resources. The world is huge, and our brains can only process so much, yet we still manage to make sense of such a complex world and move through it. At the same time, I didn’t want to study these questions in only a traditional way. I’m interested in using technology and innovative methods to explore fundamental questions about human experience, especially cognition and culture. 

What are some of those innovative methods? 

Large language models are one example. They can be used as experimental settings to study how humans think, learn and communicate. The way we talk to these models can reveal something about how we think and learn. 

During my PhD, I developed a robotic platform and deployed it in people’s households. We studied how people’s conversational dynamics and behaviours changed when they interacted with robots. The larger question was a social science question, but we developed different technologies and systems in order to study it. 

What did these robots do? Would they do housework?  

I wish, but they actually didn’t have arms so they couldn’t do housework. They were conversational agents, which means they could display various non-verbal cues. The robot could listen, and then when it listened, it could nod its head, things like that. It had language ability and conversing capability, as well as non-verbal communication skills. 

Did they look like robots? 

They looked like cute pets. 

Do you have a recent success story, a project that you worked on that you’re especially proud of? 

I’ll share my latest publication, which is in press at Trends in Cognitive Sciences (TICS). I’m picking this project because it’s a truly, truly exemplary example of interdisciplinary research. Our collaborators are computer scientists, cognitive scientists and philosophers. And what we did is try to draw from philosophy – particularly epistemology and philosophy of science – and cognitive science on the concept of understanding.  

We then bring that into the current debates on whether intelligent machines have genuine understanding. It’s a very important topic because nowadays we are always wondering whether intelligent machines actually understand or they are just pretending to understand something via pattern matching. This paper sketches out this conceptual framework to shed light on the question of what it means if we attribute understanding to machine. 

Would you describe yourself as an AI cheerleader or an AI doomer? 

My role is to study human experience so that we can safeguard human experience in the age of AI. I don’t think our future with AI is predetermined or deterministic.  

Do you have a favorite resource you would recommend? 

There is a recently published book called The Laws of Thought by Thomas Griffith, which is quite good. It’s a several hundred-year history of how we try to use mathematical models to understand the human mind. It also discusses how the original concept of AI was not to replace humans, but it was proposed as a way to simulate and study the human mind.  

And now for the lightning round. Do you prefer print or digital? 

Digital. 

Dogs or cats? 

That’s a complicated question. I have a cat, so I’d say cat. 

Favorite cuisine? 

Fusion. I like any sort of fusion.  

Favorite podcast? 

Maybe Huberman’s Lab. He is a neuroscientist at Stanford and has been doing podcasts on neuroscience and human development for several years. The topics are very wide ranging from physiology to neuroscience to behavioral science. 

Secret hobby or fun fact? 

I love reading literature of all sorts from sci-fi all the way to canonical Western literature like Dostoyevsky. And I’m currently writing a novel.