Another exciting project and finding enabled by #Sotopia ! We find that LLM Agents' social capabilities are sometimes over-estimated when agents have too much access to other agents' social goals, and that this effect differs in cooperative vs. competitive scenarios!

Let’s talk about social simulations! Do you know that term could refer to various settings? Our new work suggests that you might want to double-check before being “amazed” by those simulations.
📜: arxiv.org/abs/2403.05020
🌐: agscr.sotopia.world 1/

More exciting work that is enabled by Sotopia!

Excited to be involved in this project! If you're queer, and artist, and have thoughts about gen AI, please do consider participating!
QueerInAI

🚨 Announcement:
DLATK is now available as a colab notebook!
github.com/dlatk/dlatk/bl…
DLATK is a suite for end-to-end human language analysis, used in 100+ AI/NLP/Psych papers. This was lead by Shashanka Subrahmanya, joint with Salvatore Giorgi, Adithya V Ganesan, johannes Eichstaedt, and World Well-Being Prj 🧵1/4

Very excited to share this paper!
There's a lot of work on AI/LLM-based agents, but with Sotopia we're hoping to enable scientific inquiries around LLM capabilities in interactions and what that means for social AI ! We already are working on 2-3 different follow ups 👀

I'm hiring a postdoc at UVA to edit and detoxify LLMs!

This is an exciting opportunity to join a vibrant research community and to collaborate with Steven L. Johnson (he/him) @[email protected] and Maarten Sap (he/him)

Feel free to get in touch and please help spread the word!

tinyurl.com/yeywuu3b

Wondering whether LLMs have enough social intelligence for real-life scenarios? Me and Graham Neubig Yonatan Bisk Daniel Fried are presenting our latest SOTOPIA benchmark at ALOE Workshop poster session (12/15 12:45pm-1:45 Room 211) Happy to chat about human-AI interaction, social and…

🏆 SODA won the outstanding paper award at #EMNLP2023 ! What an incredible journey that was! I'm grateful for this journey with you Jack Hessel Liwei Jiang Peter West Ximing Lu Youngjae Yu Pei Zhou Ronan Le Bras Malihe Alikhani Gunhee Kim Maarten Sap (he/him) Yejin Choi & Allen Institute for AI💙

Hello! Our work on *counterspeech strategies* will be presented at #EMNLP2023 by my wonderful lab-mate and coauthor Akhila Yerukola!!
arxiv.org/abs/2311.00161 (more details in the 🧵)

Please go check it out at
Findings 7 (East Foyer / Virtual) Sunday 12/10 11:00-12:30PM

📢Please RT!📢

I'm recruiting a postdoc to work on Editing and Debiasing LLMs at UVA in collaboration with Steven L. Johnson (he/him) @[email protected] (UVA) and Maarten Sap (he/him) (CMU)!

More info on my website: tomhartvigsen.com/join

Come chat with me about this at #NeurIPS2023 !

I am recruiting PhD students to work with me at Illinois Computer Science at the intersection of HCI and NLP! I am especially excited about understanding how we use language to communicate and building tools that help us communicate better.

More: talaugust.github.io/students/

Understanding others’ beliefs plays a significant role in our daily social interactions. But can LLMs achieve this same Theory of Mind? In a paper by AI2's Hyunwoo Kim, a new benchmark is introduced that offers answers: blog.allenai.org/does-gpt-4-hav… #emnlp2023

I wish we could stop worrying about whether something is intelligent or whether something is general, and just focus on whether something is useful, whether something helps people, and whether something moves society in a positive direction.

>

Our #EMNLP2023 work on 'Modeling Empathic Similarity in Personal Narratives' explores a shift in task from semantic to empathic similarity in text 🧵(1/5)

📋Paper: arxiv.org/abs/2305.14246
💻 Code and dataset: github.com/mitmedialab/em…

Don't settle for generic rewrites! Existing stylistic rewriting methods and eval metrics miss the mark by ignoring context.
📣In our #EMNLP2023 work, we highlight the importance of context, especially in evaluation.
Excited to share my first PhD paper🎉
📜arxiv.org/pdf/2305.14755…

Who knew water-cooler chitchat could be so useful? AI2's Hyunwoo Kim & collaborators did, which is why they're introducing SODA, the first million-scale high-quality social chitchat dataset. Learn about it on the blog: blog.allenai.org/are-you-thirst… #EMNLP2023

People ask why it matters to have spaces like this 🧵

I’m often in rooms where no one looks like me, and yet the tech + policies + norms shaped in those rooms disproportionately impacts people who are largely underrepresented in these spaces.

How does 'AI bias' relate to 'existential harm' -- a tech-oriented term for machine-made murder? Here's a quick jump with me deeply into the heart of the math of misunderstanding: False Positives, False Negatives, and You: 🧵