With an audience of over 100 students, Zachary Stine, an assistant professor in the Department of Computer Science and Engineering, discussed the challenges and insights of using artificial intelligence and machine learning to study culture.

Stine’s Sept. 24 lecture was one of several events hosted by the Schedler Honors College during its Challenge Week, where enrolled students were required to attend at least two events. 

During his lecture in the Doyne Health Sciences Center, titled “Mirrors of Meaning: AI as a New Instrument in the Study of Culture,” Stine discussed how AI acts as a reflection.

“AI reflects our own intelligence back at us, sometimes in weird ways, but sometimes in ways that usefully give us new vantage points from which we can view ourselves,” Stine said.

He said society obsesses over some of the new advancements with AI.

“Everyone is really captivated with so-called generative AI systems, large language models, image generating models,” Stine said. “In some sense, I think that’s just a weird by-product of what is scientifically interesting to me about these technologies.”

Stine discussed the problems and benefits of using AI to study culture.

“Studying culture is hard because you’re in the thing you’re trying to study, and that self-referentiality is weird,” he said.

Stine said studying with AI hypothetically pulls the camera lens back and allows for a broader picture of different world views.

“If we could do that kind of comparison, outside of our brains, which are situated in culture, then we might say this is the mathematical relationship between this world view and that world view,” Stine said.

Stine analyzed some of AI’s faults and limitations.

“If you were somehow to make culture perfectly empirical to get rid of the self-referential nature of studying it, you wouldn’t actually be studying culture anymore,” he said.

Stine said AI and machine learning are math and logic-based.

“The computational realm is maddingly simplistic. I like that because it lets us boil down a particular problem into a really simplistic form,” he said. “We lose a lot of richness when we do that, but what we gain is a little bit of precision to think about what this stuff means.”

Stine said language is ambiguous and ultimately contributes to its importance in our culture.

“If every text had somehow encoded within it its correct meaning, would we have arguments? We certainly wouldn’t have interpretive arguments,” he said.

Stine said there exist hypothetical games an individual can play in their lives to make an evaluative judgment about the definition of culture. He said people do this often, but he warns against it.

“I think that when you do that, you flatten the object of study in such a way that we probably shouldn’t call it culture anymore,” he said. “I think that’s kind of a key feature of culture — this kind of resistance to being definable in a precise way.”

He said these games are similar to looking through the lens of AI to try and see something like culture and meaning. 

He then emphasized the importance of investing in pure humanities programs and not just humanities programs that are interested in computational work.

“We have this weird problem where we use AI to take a picture of something, something that’s alive, and we kind of kill it when we flatten it into a picture,” he said. “But when we look at it through a lot of different lenses, I think it provides us useful information that serves as a complement and not a replacement to the ways that we have traditionally studied culture.”

After the lecture, Madison Nicholson, a senior in the Honors College, said, “I didn’t realize this was even a field. I didn’t realize computational linguistics was a thing.”