PhD application season again :) if u want an extra eye on your PhD application to @ColumbiaCompSci 💙 — submit ur statement of purpose draft for review by **next Monday, Nov 10th** :) — we’ll find current PhD students to help review 🔗 https://t.co/E2HZ0sisVA

🚨New paper on AI and copyright Several authors have sued LLM companies for allegedly using their books without permission for model training. 👩‍⚖️Courts, however, require empirical evidence of harm (e.g., market dilution). Our new pre-registered study addresses exactly this

@jingyitweets @LydiaChilton @ken0324 @YueJiang_nj @sitong0926 @jasonwuishere @savvas_petridis @TuhinChakr @fabulousQian @tywang__ 🦎 many uncertainties and challenges hinder HCI system researchers from doing long-term evaluations… thus, we bring a group of experts to help you jumpstart :) see our full proposal: https://t.co/e6n0sJ7m1i -- see our #UIST2025 workshop website: 📅👉 https://t.co/vn6KyV6Ueq

@jingyitweets @LydiaChilton @ken0324 @YueJiang_nj @sitong0926 @jasonwuishere @savvas_petridis @TuhinChakr @fabulousQian @tywang__ the answer is 📅 "evaluate over time" ⏰ - observing how a system lasts + how users use it over sessions but in HCI, longitudinal doesn’t have to mean 10 years... so the real questions are: - what’s long enough? - how do we design endurable systems + good study protocols? 3/

@jingyitweets @LydiaChilton @ken0324 @YueJiang_nj @sitong0926 @jasonwuishere @savvas_petridis @TuhinChakr @fabulousQian @tywang__ ofc ur tool worked great in that one-time user study… but what abt the 2nd use, day 2, week 2, month 2? will people love it more, or less? skip features? customize? use it for different tasks? or abandon… 📈⌚️✅ this is often unanswered with our common HCI user eval... 2/

Join our #UIST2025 workshop in Busan! 🇰🇷 📅 https://t.co/vn6KyV6Ueq we, HCI people, build systems to address user challenges but beyond that “1-hr single-session” testing, how can we truly show tools’ novelty, UX, impact? how users learn, adapt? 🧩 long-term eval helps 🌱 1/

tomorrow 6/20, i'm presenting this paper at #alt_FAccT, a NYC local meeting for @FAccTConference ✨🎤 paper session #3 🎤✨ 🗽1:30pm June 20, Fri @ MSR NYC🗽 ⬇️ our #FAccT2025 paper is abt “what if ur ChatGPT spoke queer slang and AAVE?” 📚🔗 https://t.co/EiKBKxr6Lp

this was joint work with @jeffreybasoah, Daniel Chechelnitsky; collaborators @taolongg, @_doctor_kat, @chryssaZrv, @KaitlynZhou, and @blahtino; and advisor @MaartenSap at @LTIatCMU, @SCSatCMU, @hdceUW, @stanfordnlp, @GoogleResearch, @istecnico 📚:

& here are some positive comments where users enjoyed the interaction with the sociolectal LLMs: 😊 “It just sounds more fun to interact with” -AAE participant 💅 “I enjoy being called a diva!” -Queer slang participant for more, please see our paper :) [8/9]

lastly, we asked users for explanations for their LLM preference: 👎 “[AAE LLM] using AAE sounds like a joke and not natural.” 😠 “Even people who use LGBTQ slang don’t talk like that constantly. It would be annoying to have an AI constantly use slang phrases.” [7/9]

Q: how does each of the user perception variables impact 🔁 user reliance on LLMs? A: ✨widely the perception variables (trust, satisfaction, social proximity, and frustration) were 🔗🔥 related and ➕📈 positively associated with reliance 🔁 [6/9]

we also observed user perceptions for a sociolectal LLM -- trust, social proximity, satisfaction, preference etc... & we notice... 👋🏾💗AAE participants explicitly preferred the SAELM over the AAELM, whereas this wasn’t the case for 🏳️‍🌈 Queer slang participants 💙💚 [5/9]

In our study we find: 📈👋🏾 AAE speakers rely more on the SAE LLM over AAELM, while no reliance difference for 🏳️‍🌈⚖️ Queer slang speakers this shows that for certain sociolects, users will rely more on a LLM in Standard English over the sociolect they use themselves🤎🩷 [4/9]

We run 2 parallel studies:  1/ the 👋🏾 AAE speakers using AAE LLM  2/ the 🏳️‍🌈 Queer slang speakers using Queer slang LLM participants watched videos 🎬 & l had the option to answer questions❓w/ LLM responses the LLM was either in Standard American English or in their sociolects

Our study (n=985) looks at how AAVE & Queer slang speakers 🗣️ perceive and rely on LLMs’ use of their sociolect (i.e., a dialect centered around a social class).  our main research question: “How do users behave and feel when engaging with a sociolectal LLM?” 🤷🏻🤷🏾‍♀️🤷🏽‍♂️ [2/9]

what if AI played the role of ur sassy gay bestie 🏳️‍🌈 or AAVE-speaking friend 👋🏾 ? you: plan a trip? 🤖 AI: “Yasss queen! let’s werk this babe✨💅” LLMs can talk like us, but it shapes how we trust, rely on & relate to them…🧵 📣 our #FAccT2025 paper: https://t.co/ecQ4z8BP3B

😵‍💫 Long-context human-AI planning with LLMs struggles when users have to manually manage all the context in messy chats (e.g. with ChatGPT). Meet 💡JumpStarter: task-structured context curation for better, collaborative planning with LLMs on complex tasks. 🧵 (1/n)

led also by amazing @Ruyuan_Wan @jiaqi_lll

🇯🇵🌸 #CHI2025 is abt to start! so its time! all queers! all of u on the spectrum! 🏳️‍🌈🏳️‍⚧️ @queerinAI is hosting a breakfast social on day 3 🥯🍘 📅【apr30, 8am】 ☕️【merengue hawaii cafe】 signup: https://t.co/QeUG7gqW6d *limited seats, come fast or cry later lol :) see u then!

updated my site in time #CHI2025 🇯🇵🗼 not presenting, but happy to chat about: 🦆 FeedQUAC: ambient creativity tool with real-time audio feedback (with autodesk) 🎎 DoubleAgents: AI agents that learn human values/policies by simulating test cases & co-planning, executing them!!