June 3, 2024

Sterling Williams-Ceci is a doctoral student in information science from Ithaca, New York. She earned her B.A. in psychology and the College Scholar Program at Cornell University and now studies the influence of AI on people’s thoughts about societal issues under the guidance of Michael Macy and Mor Naaman at Cornell and Cornell Tech, respectively.

What is your area of research and why is it important?

People are constantly asking AI technologies, like ChatGPT, to give them information about important issues. AI is unable to think on its own, so it instead produces information by linking words that it predicts will occur together based on past human-written text (e.g. Wikipedia). Because AI is so new and produces information in this unique way, it is unclear how much AI influences our thoughts. In my work, I investigate how much AI-generated information about societal issues affect people’s attitudes and beliefs about these issues. I’ve found that when AI-generated suggestions are biased, they can make people’s writing and opinions about societal issues more biased; likewise, balanced summaries of controversial issues can make people’s attitudes more moderate.

What are the larger implications of this research?

There are several risks involved when using AI to answer questions: For example, some AI models are trained on text that over-represents one viewpoint, which my work has shown can lead people to adopt these biases. Another deceptive aspect of AI is that it often acts human-like (such as referring to itself with first-person pronouns), so people may think AI is as trustworthy as human experts, which could be very harmful (for instance, people may ask AI for medical advice instead of doctors). My research is showing that people’s beliefs are indeed influenced by what AI tells them, and that these risks are important to educate people about.

You were named second place winner in the 2024 Three Minute Thesis (3MT) competition. Can you describe the process of condensing your research into a three-minute presentation?

I initially found it very challenging to present an entire, multi-experiment research project in three minutes. There were so many separate results that were important to the conclusion of my work, but I knew I could not possibly fit them all into my presentation. I eventually had to think about the speech as more like a storytelling challenge: I used a relatable example about political polarization to capture the audience’s attention, then quickly transitioned to my work by showing how people use AI suggestions to express their polarized views but that these suggestions can be biased, too, so what happens to our attitudes when we use them?

How will you take what you’ve learned from participating in the 3MT and use it in future academic and professional contexts?

I found this challenge very rewarding, mainly in terms of encouraging me to be succinct. I’m someone who tends to over-talk, over-detail, and over-write, and this isn’t ideal for communicating effectively with other people about my work. From the 3MT, I’ve learned that people will actually understand my message better if I use broader statements instead of ones that include every detail. I also enjoyed getting practice in the actual logistics of presenting, such as how to vary my vocal tone to best engage the audience. I will no doubt use these skills in my future presentations.

President Pollack has designated this academic year’s theme as freedom of expression. What does freedom of expression mean to you?

To me, freedom of expression means that we should be open to reading diverse perspectives and talking with people of all societal groups. A lot of the work in my discipline is concerned with fact-checking and preventing misinformation and has resulted in certain news media sources being labeled as being “low factuality” or “unreliable.” While this work has been instrumental in preventing people from adverse outcomes of believing in conspiracy theories and false claims, we must not completely exclude information from these sources – even if someone says something that is deemed “misinformation,” we can still learn something valuable about how others think.

What are your hobbies or interests outside of your research or scholarship?

I like outdoor running and am always challenging myself to run longer distances. I also enjoy learning statistics and coding in a variety of languages (my favorite is R), which has allowed me to do some cool things like create a clone of ChatGPT to use in my research. I really like traveling. I also love sugar in all forms (cookies, ice cream, candy, etc.), so anything involving dessert is automatically on my schedule.

Why did you choose Cornell to pursue your degree?

The information science program at Cornell is unparalleled, in my opinion – we have so many amazing faculty and students who are spearheading new subfields and discoveries. I currently work at Cornell Tech in NYC, which is a nice change of pace from Cornell in Ithaca. At both campuses, I feel well-supported and like I have all the resources I need to have a successful career as a researcher.