New BIDA Safety Docs

October 22nd, 2019
bida, contra, safety
Over the past several months BIDA's safety committee has been working on a new version of our documents and policies. The previous set represented a mixture of several different rounds of "we should probably have documentation about X" that didn't fit together in a coherent way and didn't explain what people should expect if they talked to the committee. The new docs try to fix this, and are:

Most of the changes are explanation, wording, and trying to make things more coherent, but there is one policy change. When taking actions in the supporting the reporter category, it no longer includes uneven splits or other asymmetric responses. This doesn't apply to actions in the protecting the community category: responses like banning someone who is a danger to others are still on the table.

This is something I've thought about a lot, and gone back and forth on. When I wrote up how I was thinking about this in December 2018 I was in favor of only doing even splits. Discussion on that post and in person convinced me that splits didn't have to be even. Discussion on that post, with other members of the safety committee, and experience trying to apply the policy then convinced me back to even splits and other symmetrical responses.

The core problem here is that safety committees are groups of volunteers who have limited ability to figure out what's actually happened. In setting up policies, we need to be aware of our limitations and be careful not to set things up in a way where we are an attractive tool for abusers. When I did think asymmetric splits were a good idea I wrote:

The worst people can do with this is turn this back into a mutual exclusion situation.
I now think this was missing a very important aspect of the problem. If a reporter knows from the beginning "this committee only implements even splits" and can decide whether to ask for one, that's very different from expecting an asymmetric split and then later having the other person be able to push things into a symmetric split.

This is all very tricky, but I think the new policies put safety committee in a much better place for taking reports in the future.

Referenced in:

Comment via: facebook, substack

Recent posts on blogs I like:

Exercises in benchmarking and evals, part 7: DeepSWE, Senior SWE-Bench, napkin math, and winter tires

This is part of a series of exercises on benchmarking, evals, and experimental design (1, 2, 3, 4, 5, 6)1. We're going to look at three questions, which are presented before the answers to give you time to think about the questions before seeing the a…

via Posts on July 23, 2026

Donation recommendations for effective altruists (July 2026)

My recommendation for donors who are American citizens or permanent residents is that they donate directly to high-value political candidates.

via Thing of Things July 22, 2026

Let's Taboo the V-Word

Read this post on Substack.

via Home July 12, 2026

more     (via openring)