It's bad for our mental health, but it's particularly terrible for optics. I've been critical of our focus on optics in the past, but like it or not, our movement has become very public facing now that AI policy has become such an important mechanism for AI alignment. Optics matter far more than they used to. It's not ideal to air our dirty laundry in the most public possible way in a place where salacious drama and outright lies based on it easily go viral due to the algorithm and bad faith actors. 

I don't think discussing it on twitter should be discouraged if the individual thinks it's the best place for it, but making discussing it on the EA Forum less frowned on would avoid boxing people into that choice.

Thanks! That was the kind of detail I was hoping for. Yeah I don't blame you for wanting someone else to handle this. It sounds exhausting. 

I think many people would find it valuable if you went into more detail here about why you think he is probably guilty of rape. It seems like you are very knowledgeable about the evidence, and your observation that the EA Forum is a better place for this than twitter was apt.

I actually don't really understand what you meant by your first sentence in your last paragraph. Could you clarify it? 

I really appreciate you taking on such a leadership role in addressing instances of sexual abuse.

I really appreciate how nuanced your take is!

I'd like to push back on a couple things. 

1. I expect using AI editing decreases voice-authenticity for many people, but for me it increases it. I doubt I'm the only one. I've had an extremely hard life, and since my earliest memories have used making up elaborate stories to find peace when little else could give me it. I wrote these stories down extensively in K-12. Then for complicated reasons I gave up creative writing for a decade and a half because it reminded me of the reasons for some of my worst hardships. The stories I made up during this time were no less elaborate and deep and central to getting out of my hard times for not being written down. Four years ago I got up the courage to finally start writing them down again. But my skill level was so low that my words were empty. They could almost never capture the vivid feelings that were so familiar in the way these stories existed in my head. I didn't have time to devote to becoming a great writer on my own, so if AI hadn't come along, that would have been that. But I've found that because I use AI in the heavily customized way I described in the post —training it to emulate my best writing, rejecting most of its suggestions etc—my writing can actually capture the worlds and characters and moments I've created that have come to mean so much to me. In non-fiction the value of communicating the underlying ideas more effectively that responsible AI assistance excels at is even more valuable. In non-fiction the words themselves don't have much intrinsic value—they are mostly a means to an end.
2. My strong desire to disclose my AI use has given me some insights into how disclosure plays out in practice. I started by disclosing my AI use, but the hostility that produced was exhausting. It made dealing with the trolls it attracted when I posted my work a full time job. When I do they seem to read my writing with the goal of finding AI tells and telling me how awful my writing is because of them. Many of the AI tells they identify are from passages with no AI input. This is especially painful given how personal my writing is for me, and I expect that is true for many writers. So after a while I gave up proactively disclosing it unless I believed it would be to an audience with particularly good discourse quality. Suddenly everything changed. People loved my writing. No-one has ever noticed the features people describe as AI tells when I haven't disclosed my AI use. It seems very much like that experiment with the Monet painting that I mentioned in my post. The contrast has left me feeling like the costs of disclosure generally outweigh the benefits. I think someone as sophisticated as you could handle that information responsibly. However, in my experience even people in our community are usually better off not knowing, because knowing causes them so much stress that I don't think they would otherwise experience. I would never conceal my AI use if asked about it. This isn't to say everyone who uses AI editing has writing that would produce this contrast, but this has been my experience. 

I want to apologize to you for some of the things I said in the comments on this post 3 years ago. What you did was much closer to an innocent miscommunication than any of us were willing to admit during peak-woke. I stood up to the pressure to be overly #MeToo about your mistakes more than almost anyone else, but even I made statements that were overly critical of you that I should have known better than to make. 

I looked back on the things I said here in response to a current sexual misconduct controversy, and was so disappointed in myself that I have tried to make amends to the community by being braver in a discussion on the current controversy on another post. 

I want to especially apologize to you as an early career woman with autism. Communication is confusing, even for neurotypicals. When I make similar mistakes to the ones you made, people treat me with empathy, simply because of my gender and career stage. I should have used that privilege to be braver about standing up for you.

I was unclear here in a way I’d like to clarify. I think an even better solution than coaching would be not really worrying about this stuff at the organizational level. I didn’t mean to imply I think there is something is intrinsically wrong with gender imbalances in fields. 

If coaching makes workers more productive enough to justify the costs it makes sense, but I don’t think it should be done in order to increase female representation. Women aren’t better professionals than the equivalent man. It frustrates me that women in STEM are essentially paid more for the same work because of all these to initiatives to increase our representation. I suspect some at my university are in blatant violation of anti-discrimination laws. 

Some professions have more female-typical communication norms; others have more male-typical communication norms. More typically male communication norms are not just preferred by many men but many women as well. The more male communication norms in the EA Berkeley community feel like a haven for me. Gendered communication norms within professions typically correlate with how gendered the interest in the profession is for other reasons. For example, fields like mine and other quantitative ones that make EA concepts particularly accessible are typically more attractive to men than women because men and women have different average interests. AI safety is one of these professions. 
 

I realize these are new arguments so I’m happy for people to respond to this if they think I’m wrong, but I probably won’t engage much further with the issue given my high workload at present. 

Thanks for sharing your counterarguments. I appreciate it. It sounds like we have fundamentally different views on the agency of women in my position. While I'm sure we could both learn a lot by going more in depth into this topic together, that's not something I have time to do with my current workload.

Before I leave it there I will say that I actually consider dating within one's profession to reduce instances of sexual misconduct, not increase them. In the Berkeley EA scene a man who treats his sexual/romantic partner poorly would risk not just his social reputation but generally his whole career. It is a powerful incentive to treat us well. I identify that as one of the main mechanisms that produces a much safer dating environment in the Berkeley EA scene than my alternatives. Happy for you to respond to this new argument if you would like.  
 

I re-read the Time article in response to your comment. I remember it as being misleading at the time, but re-reading it, it's even worse than I remembered. I feel impatient with myself for being so intimidated by pressure to be woke that I gave it as much respect as I did three years ago. I came away motivated to not let our culture slip back into that again. 

I don't have time to get into all the things that frustrated me about that article, but a central one was the expectation that we automatically side with the parties in the story who were women with sexual misconduct allegations. They rarely gave us enough information for us to make an informed decision about what actually happened for ourselves. I respect that some of this information has to be private, but less information means you have raise your uncertainty, not fill in the gaps with assumptions. They instead relied on the #MeToo frame, which society has moved on from for good reasons. We later found out that some of their descriptions were quite misleading. A woman doesn’t become infallible just because she makes a sexual misconduct allegation. She is only human like the rest of us. For example, we’re now learning that part of one of the current scandals was just an innocent miscommunication. She is also capable of lying. In 2014 Scott Alexander tried to figure out what percent of rape allegations were false. As a lower bound, the near-provably false proportion seems to be 2% to 10%, depending on the study. As an upper bound, studies find that police consider 20% to 40% of rape allegations they get to be unfounded. To me that is too high to justify the assumptions the author made. All information is valuable. Just one instance of sexual misconduct is far too many. But we were illogical to blindly trust such a clearly biased article the way we did. Instead we should have applied the same rigor to addressing sexual misconduct problems that we apply to research on EA cause areas. That is what produces the most useful results in any endeavor, including for the victims of sexual misconduct themselves. 

For context, I was myself a victim of sexual harassment by a senior member of my field (not in EA). He had poor English skills, and it was possible that a miscommunication from that was responsible for the worst thing he did. I felt that I could not report the harassment to the university because this possibility would not have been adequately considered due to the #MeToo mindset. With his visa on the line, it seemed like too great a risk.

Can you say more about the alarming reports of the last few weeks? It sounds like you might have more information about that than I do. 

I agree that we should be doing the best that we can to reduce sexual misconduct, not just settling for being as good as other communities. I mentioned that we don't seem to do worse than other communities because I thought you were arguing that we do. I'm glad I was mistaken.

I'm confused why this post got many downvotes. It seems completely inoffensive to me. Is it just the critique Tom Bry-Chevalier raised, or do people have other critiques? Would be curious to hear from anyone who downvoted it for other reasons. I'm not saying this with the goal of debating this (am very busy with school right now)—I'm just curious. So please don't hesitate because you don't have time for a big debate either. If someone replies to this comment and specifies that they don't have time for a big debate too, I encourage people who disagree with them to respect that, but feel free to go at it if you both want to.

I'm a woman in the Berkeley EA scene, and am puzzled by your perspective. 

My guess is that most of the trouble is that many women are shy, and find it stressful to say they are not interested in a topic of conversation or an event or when men politely proposition them etc. I remember having conversations on this topic on the EA forum 3 years ago. A lot of women argued that we should see ourselves as helpless in a way that deeply disturbed me. For example, they argued that, in general, if a man who is much more powerful in a woman's field than her flirts with her and she dislikes it, it is reasonable for her to flirt back instead of declining the advance, out of fear that he will retaliate against her professionally. But I am that woman! I'm an undergraduate, and I know exactly what happens when you critique leading men in your field in the Berkeley EA scene far more seriously than merely declining their polite advances. These men are so passionate about having good moral character that in all my 5 years in this scene I've never known one who I could picture retaliating against someone for declining their sexual or romantic advances. In fact, they are so clearly more respectful towards us than they are in the surrounding communities of similar demographics that dating outside this community is such a bigger relative risk that I no longer bother with it. It seems like there is a gargantuan gap between what many women think would happen if they spoke up about their preferences and what the world is really like, so I think Nathan's post was highly appropriate. 

As with any community, sexual misconduct occasionally occurs, but I know that Nathan takes these cases very seriously. So do I. It does not seem to me that this post is downplaying them. It also doesn't seem that EA has worse sexual misconduct rates than other communities. That post does not discuss the Berkeley scene specifically, but knowing that it is so much safer than my alternatives makes me skeptical the numbers would be worse if the sample was limited to us instead of just being disproportionately us. 

Men changing their behavior is not a solution if women being shy is the problem. If they are too shy to turn down sexual/romantic advances, sexual/romantic discomfort is likely the least of our worries—presumably it is also holding these women back from critiquing these men professionally, doing grave damage to their fields. This is something I'm aware of as a woman in a 90% male field. It is sometimes argued that one way of increasing female representation is speaking less intensely so that women who are intimidated can feel more comfortable speaking. But there is so much value in the intense way that we speak in male dominated fields. A more practical solution to me seems to be some sort of coaching for women who are shy so that they are able to speak up regardless of the intensity of the discussion and regardless of the power of who they are speaking to, like I am. This wasn't something that came naturally to me—I have very far below average self esteem. But I found that overcoming it enough to speak my mind anyway was beyond worth it. I'd rather have a world where other women like me go through such an empowering experience than corrupt our communication so they can remain trapped in fear. So far, the critiques I've seen of Berkeley EA sexual/romantic norms remind me a lot of the critiques of my field. 

But this is just my personal experience. I am very open to being wrong. Please tell me what has caused you to develop your perspective. I was a victim of many years of sexual abuse when I was younger—if sexual abuse is in fact the main problem, it is very important to me that we address it.