iiinova — where your digital self keeps evolving

Andre Ye

@andreiskiii.bsky.social

109 followers 86 following 9 posts

Incoming PhD @ MIT EECS. ugrad philosophy & CS @ UWa

Posts Replies Media

Andre Ye @andreiskiii.bsky.social · 08/05/2026

If you're interested in additional perspectives on this work, check out Jenny's blogpost on "slow AI" jennyhuang19.github.io/slow-ai-ai-t... and my blogpost on AI for "work *on* norms" andreiski.substack.com/p/ai-working...

andreiski.substack.com

AI working within norms, AI working on norms

Rethinking AI tasks beyond "objective" and "subjective"

030

Andre Ye @andreiskiii.bsky.social · 08/05/2026

Thank you so much to my incredible collaborators Jenny Huang, @aliciaguo.com, Rose Novick, @tamarabroderick.bsky.social, and @mitchellgordon.bsky.social!

120

Andre Ye @andreiskiii.bsky.social · 08/05/2026

Read our paper here arxiv.org/abs/2604.17815 and see our project website here multiverse.csail.mit.edu (best on desktop).

120

Andre Ye @andreiskiii.bsky.social · 08/05/2026

We think multiverse-like interactions and artifacts are exciting to investigate further as models get better: they leverage the strengths of LLMs in mapping and evaluating decision spaces, while empowering humans to understand and choose how they navigate these spaces.

230

Andre Ye @andreiskiii.bsky.social · 08/05/2026

We also built an interface to help people understand how upstream decisions condition downstream outputs. In user studies, participants who interacted with the conceptual multiverse thought they engaged with questions more deeply and thoroughly than with a chat interface.

120

Andre Ye @andreiskiii.bsky.social · 08/05/2026

We worked with domain experts in philosophy, AI alignment, and poetry to articulate what good decisions look like in each domain, and used this to calibrate generator LLMs that build out the multiverse for any query and verifier LLMs that evaluate the quality of the multiverse.

150

Andre Ye @andreiskiii.bsky.social · 08/05/2026

The “conceptual multiverse” is inspired by multiverse analysis methodology, which reports all results from systematically varying data analysis choices, giving analysts a more robust picture of a result's significance.

120

Andre Ye @andreiskiii.bsky.social · 08/05/2026

Answers to open-ended questions, like in philosophy or in AI alignment, are structured by decisions that should be surfaced to the user for many reasons, e.g. to avoid sycophancy or to enable more precisely specification of how a set of model alignment rules applies to output.

140

Andre Ye @andreiskiii.bsky.social · 08/05/2026

“Should I fear death?” Ask an LLM and you get one answer or a big bag, but little visibility into the decisions and assumptions that produced them. We built the "conceptual multiverse": a system that surfaces those decisions and lets users change them. multiverse.csail.mit.edu

141

Reposted by Andre Ye

Amy Zhang @axz.bsky.social · 24/06/2025

This upcoming #FAccT2025 paper was w/ an amazing duo of undergrads @andreiskiii.bsky.social @andrewshawuw.bsky.social & deeply fuses philosophy with human-AI interaction design. "Unsettling the hegemony of intention" indeed! 😛 It also won the undergrad thesis award at UW 🏅 arxiv.org/abs/2502.15242

arxiv.org

Agonistic Image Generation: Unsettling the Hegemony of Intention

Current image generation paradigms prioritize actualizing user intention - "see what you intend" - but often neglect the sociopolitical dimensions of this process. However, it is increasingly evident ...

0256

Reposted by Andre Ye

Allen School @uwcse.bsky.social · 11/04/2025

Inspired by his father’s study of neural networks, #UWAllen CS and Philosophy major @andreiskiii.bsky.social taught himself the subject on his way to enrolling in UW at age 15. He just won a @pdsoros.bsky.social Fellowship to pursue a Ph.D. at MIT in #AI. pdsoros.org/fellows/andr... #HuskyExperience

pdsoros.org

Andre Ye – Paul & Daisy Soros Fellowships for New Americans

043