Hi! I’m a PhD student studying computer science at Stanford University. My research interests lie in human-centered artificial intelligence and intelligent tutoring systems.
Adaptive Assessment and Curriculum: In this project, we aim to 1) develop knowledge tracing models to predict students answers to GMAT questions based on their previous actions and responses; 2) devise ways to build models upon them to sequence material in ways maximally useful for students; 3) use the text of questions and answers to perform natural-language analysis and experiment with augmenting knowledge-tracing and other algorithms to be able to incorporate new questions (which were not included in training data) into the software; 4) use official score reports to study to what degree it is possible to predict students real test scores based on their interactions with the software.
Intelligent Tutoring Chatbot: In this project, we’re designing, building, and evaluating a tablet-based tutoring chatbot for children that teaches through compelling narratives based on each child’s current ability, interest, and characteristics.
Enabling Natural-Language Interactions in Educational Software: In this project, we’re collecting a data set with tens of thousands of student interactions. Using the collected data, we build NLP-powered RL models to predict the most effective sequence of interactions. We also evaluate the quality of inference into student understanding by using the trained models to target feedback to students based on their natural-language responses and measure knowledge gains from their beginning of the activity to the end of the activity.
Previously, I worked with Professor James Landay, Professor Jacob Wobbrock, and Professor >Andrew Ng on a study comparing the text entry performance of speech-based dictation versus small touch screen keyboards. I also worked with Professor Maneesh Agrawala on automatically extracting references between text and tables from academic papers and with Professor >Percy Liang on robust machine learning at the Stanford Artificial Intelligence Lab (SAIL). At McGill University, I worked on computing bisimulation metrics for Markov Decision Processes with Professor Prakash Panangaden and Professor Doina Precup and defining and justifying well-founded recursion principles over LF specifications with Professor Brigitte Pientka.
- I’m currently leading the Smart Primer project at Stanford.
- I’m organizing an AI Learning Series with Glynn Capital Management. More info can be found at https://thefutureofai.splashthat.com/.
- I hosted an innovation mixer featuring how to build AI startups on Feb 1st, 2017! More info can be found at our event page.
- Check out this NPR news report on our input study!
- Our study was also covered in several Chinese mediaoutlets 🙂
- A fun video demonstrating our work: Stanford experiment shows speech recognition writes texts more quickly than thumbs
- My interview with Tech News Today
- Sherry Ruan, Jacob O. Wobbrock, Kenny Liou, Andrew Ng, James Landay. Comparing Speech and Keyboard Text Entry for Short Messages in Two Languages on Touchscreen Phones. Ubicomp. 2017.
- Sherry Ruan, Gheorghe Comanici, Prakash Panangaden, Doina Precup. Representation Discovery for MDPs Using Bisimulation Metrics. The Twenty-Ninth AAAI Conference on Artificial Intelligence (AAAI). 2015.
- Sherry Ruan. Bisimulation Metric Computation for Markov Decision Processes. NIPS Women in Machine Learning Workshop. 2014.