Zachary Levonian

@zwlevonian@hci.social
190 Followers
324 Following
170 Posts
Data Science + ML Engineering. HCI PhD @ GroupLens, University of Minnesota. Applied NLP for understanding online social support. Tacoma, WA.
websitehttps://levon003.github.io/
blueskyhttps://bsky.app/profile/zwlevonian.bsky.social
githubhttps://github.com/levon003

INCREDIBLY excited to announce this: I collaborated with the Internet Archive to make a new 24/7 lofi livestream!

By day, you can see scanners scanning and what they're scanning, and when they're not scanning, you will see public domain silent movies and pictures. Sorta like an Internet Archive TV

Watch it LIVE here: https://www.youtube.com/watch?v=aPg2V5RVh7U

lofi Archive radio 🎞️ beats to scan/read microfiche to

YouTube
Wrote about a fun research controversy involving the finding that Irish judges are more likely to cite legal cases in rulings if they have a Wikipedia article. https://levon003.github.io/2025/05/21/wikipedia-event-impact.html
How does Wikipedia article quality impact decision-making?

A randomized experiment on Irish judges and a Wiki Workshop 2025 paper on the NFL draft.

Zachary Levonian’s blog
My team got a nice write-up in a blog post from The Learning Agency. I can't complain about effusive comments from my boss!
https://the-learning-agency.com/the-cutting-ed/article/tackling-the-ai-talent-shortage-how-a-collaborative-model-is-bridging-the-gap/
Tackling the AI Talent Shortage: How A Collaborative Model is Bridging the Gap - The Learning Agency

Ed tech developers are struggling to find quality help amid an AI talent shortage.

The Learning Agency
Violent, dark, and dirty: What Americans think about the Middle Ages | YouGov

Americans have favorable views of chivalry and castles, but not the Crusades or the Inquisition

A few quick notes and a simultation on overfitting in response to Ben Recht's recent blog posts: https://levon003.github.io/2025/02/04/train-test-split.html
Overfitting to train/test splits

Ben Recht writes in a 2023 blog post that train-test splits are shockingly effective for evaluating machine learning models.

Zachary Levonian’s blog
I enjoyed this 2021 CHI paper: "Are Current Voice Interfaces Designed to Support Children’s Language Development?" https://dl.acm.org/doi/10.1145/3411764.3445271 Basic point: most system's conversational scaffolding is rigid and bad, it's probably not even fair to call them "conversations"!
Are Current Voice Interfaces Designed to Support Children’s Language Development? | Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems

ACM Conferences
I've spent the last two years trying to understand how LLMs might improve middle-school math education. I just published an article in the Journal of Educational Data Mining describing some of that work: "Designing Safe and Relevant Generative Chats for Math Learning in Intelligent Tutoring Systems" https://jedm.educationaldatamining.org/index.php/JEDM/article/view/840
Journal of Educational Data Mining

Large language models (LLMs) are flexible, personalizable, and available, which makes their use within Intelligent Tutoring Systems (ITSs) appealing. However, their flexibility creates risks: inaccuracies, harmful content, and non-curricular material. Ethically deploying LLM-backed ITSs requires designing safeguards that ensure positive experiences for students. We describe the design of a conversational system integrated into an ITS that uses safety guardrails and retrieval-augmented generation to support middle-grade math learning. We evaluated this system using red-teaming, offline analyses, an in-classroom usability test, and a field deployment. We present empirical data from more than 8,000 student conversations designed to encourage a growth mindset, finding that the GPT-3.5 LLM rarely generates inappropriate messages and that retrieval-augmented generation improves response quality. The student interaction behaviors we observe provide implications for designers---to focus on student inputs as a content moderation problem---and implications for researchers---to focus on subtle forms of bad content and creating metrics and evaluation processes.Code and data are available at https://www.github.com/DigitalHarborFoundation/chatbot-safety and https://www.github.com/DigitalHarborFoundation/rag-for-math-qa.

Can recommendation systems help people find peer support online? I'll be presenting "Peer Recommendation Interventions for Health-related Social Support: a Feasibility Assessment" at #CSCW2025! I first submitted this work in July 2022, so I'm happy to finally share this. https://arxiv.org/abs/2209.04973
Peer Recommendation Interventions for Health-related Social Support: a Feasibility Assessment

Online health communities (OHCs) offer the promise of connecting with supportive peers. Forming these connections first requires finding relevant peers - a process that can be time-consuming. Peer recommendation systems are a computational approach to make finding peers easier during a health journey. By encouraging OHC users to alter their online social networks, peer recommendations could increase available support. But these benefits are hypothetical and based on mixed, observational evidence. To experimentally evaluate the effect of peer recommendations, we conceptualize these systems as health interventions designed to increase specific beneficial connection behaviors. In this paper, we designed a peer recommendation intervention to increase two behaviors: reading about peer experiences and interacting with peers. We conducted an initial feasibility assessment of this intervention by conducting a 12-week field study in which 79 users of CaringBridge received weekly peer recommendations via email. Our results support the usefulness and demand for peer recommendation and suggest benefits to evaluating larger peer recommendation interventions. Our contributions include practical guidance on the development and evaluation of peer recommendation interventions for OHCs.

arXiv.org
"Baltic Subsea Cables: A Story of Resilience Rather Than Fear" https://pulse.internetsociety.org/blog/baltic-subsea-cables-a-story-of-resilience-than-fear
Baltic Subsea Cables: A Story of Resilience, Not Fear

The narrative of the story should highlight the resilience of subsea cables and the Internet as a whole to continue wor…

Internet Society Pulse
I'm still holding out hope for Mastodon, but if you're an HCI or ML researcher here I'll mention that the Bluesky migration has blown up in the last week. A lot more activity there: https://bsky.app/profile/zwlevonian.bsky.social
Zach Levonian (@zwlevonian.bsky.social)

Human-computer interaction researcher. PhD from University of Minnesota. Tacoma, WA. Mastodon: zwlevonian@hci.social

Bluesky Social