Post

Clément Canonne

@ccanonne.github.io

Preparing a blog post, and as a side product I wrote some short thoughts on AI in reviewing, in TCS and beyond.

9/10/26, 8:51 pm5 reposts41 likes

Thomas Steinke

@stein.ke

> nobody forces anybody to be on a Program Committee, or to be a (sub)reviewer

Clément Canonne

@ccanonne.github.io

I don't see ICLR as TCS, tbh, and also I think that policy is wrong. (But even then, my comment stands, as technically nobody forces anyone to submit a paper to ICLR if they'd rather not.)

Suresh Venkatasubramanian

@geomblog.bsky.social

I agree with your point on the merits. The problem that @stein.ke raises though has its own cascading effect. More people decline reviewing, fewer reviewers, more motivation to use AI in reviews, and so on.

Clément Canonne

@ccanonne.github.io

Well: more people are coerced into reviewing, more motivation to use AI in reviews+less confidence in peer-review, and so on :) As I said, mandatory reviewing sounds to me like a very bad and self-defeating solution to the problem ML conferences have (too many submissions)

Prasanna Ramakrishnan

@pras17.bsky.social

Thinking about this, I wonder if the situation could be helped by having reviewers include AI disclosures like authors do in papers. If people don't want to be dishonest, or say that they did an obviously bad thing,...

Prasanna Ramakrishnan

@pras17.bsky.social

...it could encourage them to use the tools in a way that's an improvement (better understanding of the paper, better review, etc.). Also helps the PC calibrate how seriously they should take a review.

Clément Canonne

@ccanonne.github.io

For conferences which decide to allow AI in reviews, this type of disclosure should indeed be a bare minimum!

Clément Canonne

@ccanonne.github.io

Here's said blog post: bsky.app/profile/ccan...

Clément Canonne@ccanonne.github.io

I wrote a blog post on "TCS, AI, and the Adults in the Room," addressed to the senior researchers and established academics in our field: minorfree.github.io/Clement/ Thanks to @hunglv.bsky.social for lending me his blog!

Clément Canonne

@ccanonne.github.io

(this won't make it to the blog post itself, which is on a related topic, but with a near-orthogonal focus)