I’ve been working with Dr. Rob Jacob at the Tufts Human-Computer Interaction Lab for 3 years, exploring how neuroimaging tools, such as electroencephalogram (EEG) and functional near-infrared spectroscopy (fNIRS), could be used to classify cognitive workload states as people use a variety of interfaces. Initial work looked at chess puzzles, but a bulk of my contributions were on two studies in collaboration with Microsoft Research on Microsoft Copilot in Word.
Study #1: Journal Paper
Decoding Chess Puzzle Play and Standard Cognitive Tasks for BCI: A Low-Cost EEG Study
Using a low-cost Muse 2 EEG device, we explored whether we could classify high/low workload as users completed chess puzzles of varying difficulty. A majority of my contribution was in running participants through the study and literature review.
Currently in peer review, preprint can be found here
Study #2
Neural and Cognitive Impacts of AI: The Influence of Task Subjectivity on Human-LLM Collaboration
Participants completed 4 different tasks, each with and without the Microsoft Copilot tool. Using an fNIRS device and the NASA Task Load Index, we explored whether there was a significant difference in cognitive workload between AI and No-AI conditions. I was involved with every aspect of this study, including design, execution, analysis, and writing.
Will be submitted along with Study #3, but preprint can be found here
Study #3: Currently Analyzing
Extension of the previous study, using the same measurement tools (fNIRS and NASA-TLX). However, we shortened the study to two tasks focused on real business challenges such as investment or merger decisions. We found a significant difference in fNIRS cognitive workload between AI and No-AI conditions, something that was not apparent in the previous study.