I am so honored and excited to be a 2025 Paul & Daisy Soros Fellow and work on human-AI interaction at @MITEECS! I look forward to joining the incredible community of PD Soros fellows. Huge thanks to my advisors @amyxzh @RanjayKrishna & Rose Novick for their amazing mentorship :).

Excited to present our work soon on how image generation can be site for deep thinking about visual politics and diverse representation at #FAccT2025 with my amazing co-author @andrewbshaw! Thursday June 26, 10:45-10:57 AM at New Stage A.🎨✨.

Our paper on “Agonistic Image Generation”: In collaboration with @andrewshawuw, @RanjayKrishna, and @amyxzh!. Presenting at #FAccT2025 on Thursday, June 26 in "Group Behaviors and User Experiences" with @andrewshawuw — see you there! 6/6.

We tested this with 29 participants. Results: it encourages deeper thinking about representation, and users were more responsive to this approach than to superficial "diversity interventions." 5/6.

We built an interface that researches controversies around user prompts, generates multiple interpretations (especially controversial ones), and encourages users to reflect on them before creating images. 4/6.

Most AI image tools aim to actualize your “mental image” of a prompt. But your mental image may not be the whole picture. What if interfaces helped users engage with broader debates about what prompts could mean—pursuing an interactive, social notion of diversity? 3/6.

Why do we seem to care about some forms of diversity but less about others? The controversy revealed something important: authentic diversity isn't just “correct” demographic representation — it's about engaging with a pluralism of real human conflicts, views, and interests. 2/6.

[FAccT 2025 paper!] 🧵.Remember the March 2024 Gemini controversy? People were upset when it generated "diverse" but historically inappropriate images—like Black Founding Fathers and female Asian Nazis. This sparked our research. 1/6.

RT @LTF_01: This work aims to move beyond the traditional boundaries of the philosophy of science and the science of science. We propose….

RT @uwcse: Inspired by his father’s study of neural networks, #UWAllen and @UWPhilosophy major @andreiskiii taught himself the subject on h….

RT @Plinz: AI alignment that tries to force systems that are more coherent than human minds to follow an incoherent set of values, locked i….

RT @kaysorin: last night i demo’d @openai o1 pro to my most skeptical friend - the results were extraordinary. it was amazing to see o1 p….

RT @amyxzh: Undergrad in the US interested in HCI/HAI research? Apply here to come work with us this summer!.

I’m presenting this #COLM2024 paper as poster #44 in the Tuesday afternoon session! Come talk to me about designing interactions and building LMs to be useful for critical thinking!.

RT @jxmnop: We spent a year developing cde-small-v1, the best BERT-sized text embedding model in the world. today, we're releasing the mo….

You can read our paper here: This work was done with the amazing @jaredlcm, Rose Novick, and @amyxzh!.

We also explore how LLMs help concretely think about metaphilosophical questions – we can get a bit at “what is philosophy?” by asking “can LLMs do/output ‘philosophy’?”. LLMs might even actively address metaphilosophical problems by pushing philosophers' intellectual tendencies!.

✍️≠💭 We also need to think about the incongruence between language and thought – interviewed philosophers often find it difficult and even disruptive to articulate their thinking while it is in development. By default, inputs to LLMs are thoughts "thought enough to be written".

But other interesting problems also arise. For instance, LLMs need to grasp “uncommon sense” ☝️🤨 – seriously considering positions which deviate from commonly ‘correct’ ways of thinking about the world. This might point to different ways of aligning LLMs, e.g. pluralistic.

Many active problems in LLM research turn out to be important to building LLMs as effective critical thinking tools: e.g., output consistency, long-range planning, writing to and drawing from ‘memory’, theory of mind.