PhD candidate @UofTCompSci w/ @ashton1anderson · Measuring how AI shapes human learning, creativity & wellbeing · ex @MSFTResearch
Toronto
Joined September 2018
- Tweets105
- Following656
- Followers284
- Likes1.4K
Pinned Tweet
I’ll be at #ICML2026 next week to present our position paper on LLMs optimized for well-being.
Would love to meet folks thinking about measuring + improving AI’s long-term effects on people (e.g., creativity, learning, wellbeing).
Poster: Wed Jul 8, 10:30am, Hall A #3215
Harsh Kumar retweeted
Looking for postdocs+PhDs to help lead a large-scale study of the effects of long-term AI use, a collaboration between Berkeley+@TransluceAI+others!
Funding+positions available (postdocs/visiting researchers/Transluce affiliations). Please reshare; application in next tweet!
Harsh Kumar retweeted
1/ As frontier LLMs continue to improve, their reasoning is getting harder for us to follow and monitor.
Ensuring we can understand what models are doing is critical for safe AI progress.
Since 2023, we've been developing a paradigm to build more intelligible models. 🧵
Harsh Kumar retweeted
🚨The AI value alignment problem isn't one problem—it's many. The field bundles together distinct conceptual, ethical, social, and technical challenges in a messy and unhelpful way.
Progress in AI alignment starts by recognizing which problem we're actually trying to solve. A mature science of alignment needs clear concepts and structure.
📢 Excited that our paper "A taxonomy of AI Value Alignment problems" has been accepted at @AIESConf 2026!
Harsh Kumar retweeted
Looking forward to sharing this work with @DavMicRot, @jakehofman, and @dggoldst — see you in Philly!
Come see _the_ Mohammed Alsobay ( @malsobay ) present at AOM Tuesday August 4th in the session "From Tools to Social Actors: How Conversational AI Is Reshaping Work" 8-9:30 in 107A.
events.aom.org/og/sessions/8…
Harsh Kumar retweeted
I'm at ACL presenting these papers! Also excited to chat about developing LLMs that promote *human* agency and well-being, by measuring and accounting for user impacts and implicit assumptions made in human-AI interaction :-) or anything else :P
#ACL2026 #ACL2026NLP
Joint work with Ashton Anderson (@ashton1anderson ), Louis Tay (@LouisTaySC ), and Karina Vold (@karinavold).
Paper: openreview.net/pdf?id=LCsPkh…
Poster page: icml.cc/virtual/2026/poster/…
Harsh Kumar retweeted
Grateful for featuring our paper about AI’s impact on critical thinking under time constraints (w/ @1harsh_kumar @MinaLee__ )
More on the project website: criticalthinking-ai.app
Check our full paper 👉: doi.org/10.1145/3772318.3791…
#AI #humanthinking #cognition
As AI becomes an active participant in our cognitive lives, the line between thought partner and substitute is becoming harder to maintain. time.com/article/2026/04/30/…
Harsh Kumar retweeted
New paper: "In Your Own Words"! We propose a computational framework for identifying interpretable themes from free-text survey data, and demonstrate its benefits on a new dataset of self-described race, gender, and sexual orientation. 🧵1/
New #CHI2026 paper led by @JiayinZhi0728 that I'm grateful to be part of! Most AI cognition research assumes sufficient time. But real life? We're either rushing at the last minute or spiraling over a project due next year. Time context changes how AI impacts our thinking.
✨New paper alert
Does AI help or harm critical thinking? It’s one of the most pressing questions around AI. But maybe the better question is: under what conditions? In our #CHI2026 paper, we find that time can change everything.
🌐 criticalthinking-ai.app/
#AI #criticalthinking #humancognition #decisionmaking #civicparticipation
(1/10)
Harsh Kumar retweeted
We were doing positive results on AI in education before positive results on AI in education was cool:
dl.acm.org/doi/abs/10.1007/9…
🚨🚨 Excited to share our first *positive* results on AI in education!
Most AI tutor work focuses on making the chatbot better. We suggest another lever: deciding what students should practice next to improve learning.
We combine an LLM tutor with reinforcement learning to personalize problem sequencing using signals from student-chatbot interactions and solution attempts.
We tested this in a 5-month randomized field experiment in a Python course across 10 high schools in Taipei. All students had the same course material and the same AI tutor. The only difference was adaptive vs. fixed problem sequencing.
Result: across 770 students, adaptive sequencing improved performance on an in-person final exam taken without AI assistance by 0.15 SD, with larger effects for beginners. Our evidence suggests the gains came from stronger engagement and more productive AI use.
Harsh Kumar retweeted
What if we boost LLM output diversity by fusing base and aligned models?
Excited to share this work led by @YichenZW, whose passion for understanding LLM diversity really shows here. More insights and fun ideas on LLM creativity coming soon as well!💡
Lack of diversity in your LLM generation?
(also noted by Artificial Hivemind, best paper @NeurIPSConf)
Time to bring your base model back!
An inference-time, token-level collaboration between a base and an aligned model can optimize and control diversity and quality!
Harsh Kumar retweeted
My big problem with low-powered studies is that CONDITIONAL on obtaining a significant result, the estimate is often exaggerated or the wrong sign, which misleads entire subfields for years.
Doubt it? Run a simulation.
Harsh Kumar retweeted
Please RT:
Reminder that I'm recruiting MSc and PhD students to join my lab at UAlberta focusing on Human–AI Interaction. Come shape the future of AI-assisted programming tools for better productivity and learning!
Apply here: ualberta.ca/en/computing-sci…
Deadline: December 15
Harsh Kumar retweeted
Want to be an intern at Microsoft Research in the Computational Social Science group in NYC (Jake Hofman, David Rothschild, Dan Goldstein)
Follow this link and do your thing! Deadline approaching soonish!
apply.careers.microsoft.com/…
Harsh Kumar retweeted
Hello colleagues and friends! ✨Please help repost and share ✨: I am officially on the job market for 𝘁𝗲𝗻𝘂𝗿𝗲 𝘁𝗿𝗮𝗰𝗸 𝗮𝘀𝘀𝗶𝘀𝘁𝗮𝗻𝘁 𝗽𝗿𝗼𝗳𝗲𝘀𝘀𝗼𝗿 and 𝗶𝗻𝗱𝘂𝘀𝘁𝗿𝘆 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗿𝗼𝗹𝗲𝘀 where I can apply my research to real-world AI safety challenges!!
My work combines empirical studies and system building to explore topics such as 𝗿𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗹𝗲 𝗔𝗜 (𝗥𝗔𝗜), 𝗔𝗜 𝘀𝗮𝗳𝗲𝘁𝘆, 𝗮𝗻𝗱 𝗵𝘂𝗺𝗮𝗻-𝗔𝗜 𝗰𝗼𝗹𝗹𝗮𝗯𝗼𝗿𝗮𝘁𝗶𝗼𝗻. To this end, I develop tools and processes to support responsible AI in real world industry settings, particularly in the contexts of 𝗔𝗜 𝗮𝘂𝗱𝗶𝘁𝗶𝗻𝗴, 𝗿𝗲𝗱-𝘁𝗲𝗮𝗺𝗶𝗻𝗴, 𝗮𝗻𝗱 𝗶𝗺𝗽𝗮𝗰𝘁 𝗮𝘀𝘀𝗲𝘀𝘀𝗺𝗲𝗻𝘁.
Harsh Kumar retweeted
A subtle aspect of predictive modeling is target variable construction: translating an unobservable concept like "healthcare need" into a prediction target
But how does target variable construction unfold in practice, and how can we better support it going forward? #CSCW2025🧵
Harsh Kumar retweeted
Sharing our initial leaderboard and open data release for 🎶Music Arena⚔️!
Music is subjective and multi-dimensional. A key goal of Music Arena is to provide insights beyond binary preferences! 🧵
Harsh Kumar retweeted
🔊 Easy to lose hours to live jamming w/ Magenta RT. Feels like a glimpse at something magical
Try it yourself! Some setup involved (we're working on improving that!) but worth it: g.co/magenta/rt-audio-inject…
Harsh Kumar retweeted
If you’re at @AOMConnect #aom2025, come check out our panel on NLP in organizations, featuring @daveholtz, Sameer Srivastava, @william__brady, @emyeom & me! Our symposium presents four perspectives of exploring organizational behavior through language💬 7/29 @ B4-m10, 1:45pm!
