Hello,

I am a Postdoctoral Researcher at the University of Colorado Boulder, broadly interested in how intelligent systems can better support human reasoning and real-world work. My PhD research investigated the gap between the promised flexibility of crowdwork and the rigid workflows enforced by current platforms. Through empirical studies across smartphones, tablets, speakers, and smartwatches, I examined how crowdworkers adapt their practices to different devices. This work informs my current research on designing adaptive AI systems that align with users’ cognitive processes, goals, and contexts.

HCIC Conference : https://hcic.org/
HCOMP 2026 and Collective Intelligence (CI) 2026 conference are colocated this year
Read more about it here: https://www.humancomputation.com/2026/organizers.html
Qualitative annotation of team communication is essential for understanding collaboration, yet remains labor-intensive and difficult to scale. Large Language Models (LLMs) offer promise as annotation partners, but their reliability varies with interpretive complexity. In this work, we present a mixed-methods evaluation of LLM-assisted annotation using 138 utterances drawn from an empirical, scenario-based disaster-response study, coded across established Team Communication and Metacognition frameworks. By analyzing accuracy, stability (entropy and flip rates), and reasoning quality across progressively scaffolded prompt conditions, we map annotation tasks onto a three-tier spectrum of automatability: fully automatable, prompt-sensitive, and human-critical. Our findings show that LLM autonomy is not binary, but depends on the theoretical and contextual demands of the code. We conclude with a practical framework for calibrating human–AI collaboration in qualitative HCI workflows, offering guidance for responsibly scaling annotation without sacrificing interpretive depth. More: https://dl.acm.org/doi/full/10.1145/3772363.3799244