Hi, I am the person who Sapphire chatted with. I worked on the documentary discussed with Larissa Schiavo. Our grant was evaluated by EA Infrastructure Fund in August 2021. I want to be clear that I don't speak for Larissa, and she had no knowledge of my conversation with Sapphire until now.

This post caught me by surprise, Sapphire did not communicate her intent with me prior to publication. I would have rather it came out differently. However,  I agree substantially with Sapphire's position with regards to leadership communication being frustrating and secretive. I think the example of what happened around the incident alluded to does provide some ideas for directions EA organization's leaderships could take in changing  in their communications styles and the style they telegraph to others to take. I will explain this in detail, but first want to clear up some misconceptions and misinterpretations.

• First, Larissa and I were jointly responsible for the project at the time, and now she is the sole director. I have never represented myself as "the person in charge" of the project. 
• Second, Larissa and I were neither asked to exclude Zeynep Tufekci nor Helen Chu from our project, nor did we ever consider rescinding our requests to interview either of them as a result of the conversations that we did have with grantmakers. I will explain why I mentioned them in that text exchange later.

What did happen?

During the grantmaking process, Larissa and I were in fact asked to consider excluding Zvi Mowshowitz from being interviewed for the documentary project. All the communication we had received from EAIF so far emphasized the need for discretion, and this was one of a couple off-record requests.  Of the three reasons I referenced to Sapphire (low visibility, unclear record, lack of politesse), the third was most relevant. Zvi had been consistently vitriolic and unpleasant in his blog posts towards the CDC, FDA, and other public health entities. As CEA and other EA orgs have working relationships with those entities, our grant advisor at EAIF wanted us to mind the potential harm to CEA from association with Zvi's on-record statements via their funding of our project. After some debate between ourselves, we took this suggestion. Zvi was aware that this was the given reason for his exclusion. 

(I want to thank Zvi for helping get the project off the ground, and also for continuing to advise the project after we uninvited him from being interviewed.)

The debate between Larissa and I about ultimately stemmed from a disagreement about how seriously to take funder suggestions, and also the consequences of doing so. As a grantee, it is hard to tell the difference between mere suggestions and implicit commands based on suggestions. As the grant advisor is in a relatively powerful position, and conflict is asymmetrically costly, the payoff matrix is weighted more towards following the grant advisor's suggestions than not. It is still unclear to me whether our grant advisor was telling us "maybe not" or "definitely not", and whether they themselves were taking advice from someone who was telling them "maybe not" or "definitely not" - their payoff matrix is the same as ours, relative to people in positions of power over them. More on this in suggestions.

With regards to consequences, I thought there was more cost to taking the suggestion onboard than Larissa did. I thought that Zvi was only saying a more extreme version of what other interviewees had to say, as in late 2021 it was not exactly brave or controversial to say the FDA and CDC fucked up early pandemic response in the US, and the WHO globally. Zeynep and Helen were people who were on record saying this sort of thing quite loudly, and early, at professional risk to themselves. Would the same exclusion rules apply to them? I personally thought that including such matter was important to the story that we were trying to tell, about the motivations that inspired impactful actions, and excluding it would fail to tell the most insight-generating parts of the story. 

Larissa disagreed, and felt that a perfectly good film could still be made excluding such statements. I realized that she was right, but that one of my key personal motivations for the project was to say things like what Zvi had to say, and I would value the result a lot less without it. This explains my frustration as expressed to Sapphire with regards to what I perceived as attention to politesse rather than to impact, and the risk-aversion that this entailed. More on that in suggestions.

I eventually left the project in April 2022 when Larissa asked me to because I was depressed and distracted and thought I wasn't doing anything really valuable anymore, and recognized that I was being disrespectful to Larissa by being an absentee co-director when she still saw value in the project and wanted to complete it. I wish we parted on better terms, I am grateful to her for incredible generosity and hospitality which I have never repaid sufficiently. 

Suggestions for improvement

I want to be clear that I do not want to assign blame to EA-sphere communications norms for a value conflict between Larissa and I, or how it was handled. Where communications norms came into play was whether or not to ask our grant advisor for clarification.

All the communication we had received from EAIF so far emphasized the need for discretion, and this was one of a couple off-record requests. Larissa and I did not realize that we had interpreted our grant adviser's off-record words differently until after the meeting was over. Larissa felt it was bad form to ask for clarification, and I did not want to do so unilaterally. Different communication norms could have defused this portion of our debate. So I have a few suggestions. 

• The first is to defuse the risk of grantees or reports appearing insubordinate by not taking suggestions that are actually meant as suggestions by giving more explicit credences on suggestions, especially those given orally. On the other hand, implicit requests should not be phrased as suggestions. I am planning on writing a separate post on this on LW. 
  • I give a 70% chance probability that this makes communication around suggestions better if implemented, and we don't all just go up one more simulacrum level. 
  • 20% this actually catches on. 
  • At any rate, something about how to telegraph suggestions versus requests should be  part of the norms around internal communications to make this more clear for both parties.
• The second is to give grantees explicit ways to follow up on off-records conversations like the one we had with our grant advisor. If more grant advisors were aware of this issue, and could clarify that "if you have any confusion about what we discussed, please please please ask me" written explicitly in a followup email, then grantees might feel able to request clarification more easily. 
  • 90% this would have made it easier to ask for clarification. 
  • 70% asking for clarification would have helped. 
• Thirdly, where uncertain, it would be good to delineate at what points in the process oversight is to occur. Both Larissa and I understood that EAIF (and by implication CEA) getting final say was a condition of the grant funding. It was unclear how much permission we would have to request from EAIF and the parties they worked with in order to tell this story, with or without Zvi. It would have been easier if we had known more explicitly what was wanted, rather than ultimately feel forced to choose the more risk-averse route. 
  • 20% we would have made different decisions as a result of more clarity here. Still would have made a difference.
  • 90% sure our grant advisors also weren't sure what the appropriate oversight steps would have been for this project - something that could have used further discussion. 

I think that if what happened in our case is typical, it helps to explain the culture of secrecy and risk-aversion. Because of EA's increasing proximity to external power, like major governmental entities such as the FDA and CDC, potential political conflicts become key considerations in decisionmaking. This means people interfacing with them want to be less open and less risk averse in order to minimize the damage from mistakes they might make, but their doing so ends up causing other people in EA orgs to have do the same at all levels. It is SUCH a rationalist trope to blame everything on mysterious inadequate equilibria but is what I am going to do here! I hope that we actually exchange this equilibria for a better one rather than just acknowledge it sadly. Some final points towards that goal.

Overall, I think my vision of what I'd like EA-sphere communications to be like is most similar to Oli Habryka's vision - more transparency, less party line. Some elements of that: 

• More unscripted public discussion between respected figures in EA, rather than only after everyone has agreed. 
• Major EA orgs find ways to publicly voice disagreements with each other and still cooperate effectively on projects. 
• Major EA orgs reorient their strategic vision so that they do not depend on other EA orgs having positive PR. Currently, the risk of negative PR to any major org is a situation which forces cooperation between all major orgs on communications objectives. 
• Associated with that, less focus on PR, especially damage control. Damage control fosters secrecy. 
• More forum discussion like this.
• More focus on the individual problems in the world outside EA-sphere, rather than meta-EA. Yes, I know I am not contributing to the solution here! 

I do not know if this is the direction which leads EA to the most success, but it is the one which I'd feel most comfortable with. I wish I'd submitted some of this to the redteaming contest. 

Hey, I'd like to look into this but I'm having a hard time figuring out who might have given this kind of feedback to the documentary maker, or what part of CEA might have been advising on the project. DMing you.

Would it be a good norm that people contact organizations they plan to criticise before publishing such posts? I can only think of this as beneficial when the post is based on not easily verifiable, private information. While it is legitimate to use such information as basis for criticism, there are usually two sides to a story.