Zachary Levonian (@zwlevonian) header
Zachary Levonian (@zwlevonian) profile picture

Data science + ML engineering. HCI PhD @grouplens. Researching: health, social support, recommendation, applied NLP. Mastodon: @zwlevonian@hci.social

Tacoma, WA
Joined August 2017
It looks like the academic migration to Bluesky will be successful this time (RIP to Mastodon). So I have also migrated there, hopefully permanently: bsky.app/profile/zwlevonian.…
1
94
Zachary Levonian retweeted
Ok, I wrote this up (link below)
2
6
1
19
3,037
Come join my research lab in @CSatMines and pioneer a ✨✨brand new field ✨✨for #HCI in online spiritual care! I am looking for an excellent PhD student for Fall 2025. Find me at #CSCW2024 to introduce yourself in-person. I’d love to meet you! (See topics in my reply thread.)
2
8
2
47
4,029
So cool to see my friend @joelmjaeger quoted in the Washington Post today as a climate and energy expert! washingtonpost.com/climate-e…
1
61
Zachary Levonian retweeted
Been waiting to see clear studies on this: “LLMs can mark open text responses to short answer questions, across various by subject areas and grade levels. We found that GPT-4 performance (with minimal prompt engineering) was in line with the performance of expert human raters.”
Can AI grade students? New paper examines GPT-4 ability to score real student answers. Professors and teachers should take notice. l.sslevine.com/kCEO
13
61
7
409
104,960
I built a demo tool that uses GPT-3.5 to generate math hints. Walton's survey reveals that teacher perceptions of tools like this are surprisingly positive.
What do teachers, parents, and students think of #AIinEducation? Usage is growing and support is high. New polling provides helpful insights on how these groups think about and use AI in and out of the classroom. Learn more: ow.ly/P6Fp50SfpO7 #EdTech
1
4
486
Fun seeing an article I did a bit of work on get featured! I wish it was a higher-quality article 😅
2
298
Zachary Levonian retweeted
Hello CSCW Community! 📣🌞☺️ Please consider submitting to the Poster Track of #CSCW2024. Posters are an opportunity to describe new work or work that is still in progress. The deadline is May 23, 2024. For more information, please check cscw.acm.org/2024/index.php/… #PuraVida
11
18
5,610
Zachary Levonian retweeted
Replying to @UMNComputerSci
@UMNComputerSci is offering a 10-week summer @NSF REU program for undergraduate students who are interested in research on Human-Centered Computing for Social Good. Stipend is provided. reu.cs.umn.edu Apply by Apr 10; applications will be accepted until filled
13
1
15
5,667
Dream postdoc!
You’re unable to view this Post because this account owner limits who can view their Posts. Learn more
1
374
What's the best way to advise my first PhD student to pick #CHI2024 workshops to submit to? Biased answers only, please! 😇 (Explanations or "other" reasons invited in replies, please!)
45%Relevance of topic to PhD
15%Interesting tangent topic
9%Organizers are my friends
32%Strategic networking opps
47 votes • Final results
3
1
1
918
Love this approach. Prompt engineering for LLMs isn't mature or stable enough to *not* look at prompts created by abstractions. "Show me the prompt!"
New blog post 📰: Understanding what LLM frameworks are doing is difficult (digging through code, docs, etc). I show how intercept API calls with @mitmproxy to quickly understand how popular frameworks work and find interesting results. I walk through examples with Guardrails, LangChain, DSPy, Instructor and Guidance. I illustrate the value of saying "Fuck you, show me the prompt" before starting to use any tool. hamel.dev/blog/posts/prompt/
178
I think this is basically true in HCI as well!
Replying to @samth
3. In PL, the biggest reason reviewing is frustrating for authors is that we do not have field- or area-wide consensus on methods, goals, standards, or approaches. This is obviously not fixable in the publication process.
1
1
8
2,588
Interesting interview. This point about biting your tongue exactly mirrors @lastpositivist's argument about when people value "speaking their mind" in online political communication sootyempiric.blogspot.com/20…
206
A researcher writes a creative theory predicting a complex pattern. She runs a preregistered study that supports all hypotheses. HOWEVER, you realize she ran a secret pilot to learn the pattern BEFORE writing the theory. Does this new info change the evidence for the theory?
26%Less evidence now.
43%Same evidence.
31%More evidence now.
683 votes • Final results
5
10
9
18
19,189
Zachary Levonian retweeted
Do you wish you could test a different social media algorithm? Now you can. Announcing the Prosocial Ranking Challenge. $60,000 in prize money -- and we will test five winning algorithms on real users and real platforms using a browser extension. humancompatible.ai/news/2024…
3
50
10
129
29,103
Zachary Levonian retweeted
New dataset for anyone interested in social media beyond the West👇 We provide 400m+ user interactions and 1.4m user profiles for the microblogging platform Koo which has gained traction in India, Brazil and Nigeria. Paper: arxiv.org/abs/2401.07599 Data: zenodo.org/records/10476212
5
28
2
77
10,785
I'm reposting this as a general note but also because I'm looking for someone to come talk to my class about socialization processes within the family - can be about gender, race, class, disability - anything goes! Please reach out if interested!
Replying to @Dr_jlgreenlee
Graduate students, postdocs, and trainees are welcome if interested! Non-academics are welcome! Clinicians who work with families are welcome! Come talk with my students about your great work with families! @PsychChatter @APA_43 @SPPDiv54 @ncfr
1
373