Dawn Drescher

AI Men­tal Health Chat­bots for Low-Re­source Set­tings: A Pri­ori­ti­za­tion Framework

• Dawn Drescher 25 Oct 2025 18:21 UTC

  3 points

  1 ∶ 0

  on: Retroac­tive Fund­ing for Alignment

  [Same as what I replied to your DM:] Yeah, exactly! Systems like that were on my mind a lot around 2021–23 when we launched Impact Markets, which is now GiveWiki. You could do the payouts if humanity has survived for another year, has survived AGI, has survived a year with AGI, etc. You could also chain retrofunders with different time horizons. We tried to get something like this off the ground for about 2 years, but couldn’t find any retrofunders anymore after FTX collapsed. (Some of the regrantors were interested.)
  
  https://​​impactmarkets.substack.com/​​p/​​chaining-retroactive-funders
  
  If you can find a retrofunder for a system like that, you can make it happen!

• Dawn Drescher 1 Oct 2025 14:57 UTC

  2 points

  1 ∶ 0

  in reply to: John Salter’s comment on: Seek­ing Feed­back: An Ini­ti­a­tive on AI, Men­tal Health, and Alignment

  Hiii! Thanks! I’m only speaking for myself here, and I’m mostly interested in #3, or specifically in building, testing, and rolling out an AI-based tool for this rather than an RCT.

  2. Consider the challenges of distribution and funding.

  Yeah, working directly with the likes of Google (Gemini) and others would be swag, but correct me if I’m wrong, I see a very low chance of that working out? There is little commercial incentive in it for them, it doesn’t help them gain more market share from their competitors because our target clients can’t pay much, reputational risks similar to self-driving cars, etc. I haven’t asked anyone who works there, but I’m not sufficiently optimistic that it could work out to attempt it… Besides, if it does work out and lots of people start using Gemini for therapy, and then Google redecides and closes that department again, lots of users will use new version of a product for a purpose for which it’s not tested or optimized anymore.

  But I already built an alpha version of an app for mentalization-based treatment on top of Gemini. That’s super easy, and I’ll permanently have control over the instructions and possibly the fine-tuning. If it should turn out to be too risky, I can shut it down, or more likely I can make adjustments to minimize any new risks.

  Do you think I overestimate the difficulty of working with the model providers?

  4. Doing too many things

  The topics can probably be trimmed down a bit, but I feel like #1–3 form a nice story line where we first assess the risks, the assess the opportunities, and then exploit them? Personally, I’d rather ⁸⁰⁄₂₀ all of that by rolling out my solution only to fairly stable people first (I’m in some relevant support groups), collect feedback, poll well-being measures from time to time, and react to any problems with safety (in the feedback) or lacking effectiveness (well-being measures) along the way, while I increasingly market it to wider audiences. The others might want to take this more slowly, and as a result they’ll probably have the better data, but when that data is in, I can still optimize my tool accordingly.

  Do you think it would really be better to focus on one topic only or would you agree that merging and 80/​20ing is the better approach?

Seek­ing Feed­back: An Ini­ti­a­tive on AI, Men­tal Health, and Alignment

• Dawn Drescher 20 Aug 2025 8:55 UTC

  2 points

  0 ∶ 0

  in reply to: tobycrisford 🔸’s comment on: Beyond Con­trol: The Strate­gic Case for AI Rights

  They model that, and after, I think, 1661 iterations of the human-AI trade game, the human-AI trade game accumulates enough wealth for humans that it would’ve been self-defeating for the humans to defect like that. I think it’s still a Nash equilibrium but one where the humans give up perfectly good gains from trade. (Plus blockchain tech can make it hard to confiscate property.)

Beyond Con­trol: The Strate­gic Case for AI Rights

• Dawn Drescher 11 Aug 2025 9:13 UTC

  8 points

  0 ∶ 0

  in reply to: Larks’s comment on: How to Donate to Alle­vi­ate Suffer­ing in Gaza

  Thanks for the summary!

• Dawn Drescher 10 Aug 2025 20:00 UTC

  2 points

  0 ∶ 0

  in reply to: Larks’s comment on: How to Donate to Alle­vi­ate Suffer­ing in Gaza

  by promoting revanchism and ethnic hatred

  Hiii! Do you know someplace where I can read up on that? Ty!

How to Donate to Alle­vi­ate Suffer­ing in Gaza

Break­ing the Cy­cle of Trauma and Tyranny: How Psy­cholog­i­cal Wounds Shape History

With the Fu­ture of the World in Your Hands, Think for 6.77 Years!

• Dawn Drescher 7 Aug 2025 18:28 UTC

  2 points

  0 ∶ 0

  in reply to: Felix Wolf 🔸’s comment on: EA Sum­mer Camp 2025

  Oops, already subscribed. xD I’ll try to pay more attention! Ty!

• Dawn Drescher 7 Aug 2025 11:19 UTC

  2 points

  0 ∶ 0

  on: EA Sum­mer Camp 2025

  Just learned about it! Do you have a newsletter or something so I can try to be quicker next year? I’ll be in Berlin around that time in case someone wants to meet up before or after!

• Dawn Drescher 21 Jul 2025 20:42 UTC

  4 points

  0 ∶ 0

  on: The stan­dard case for de­lay­ing AI ap­pears to rest on non-util­i­tar­ian assumptions

  What has been your personal take-away from this line of thinking? This “standard case” is far from my own thinking, though I agree with the conclusion. Is it also far from your own thinking?

  My take:

  1. My axiology depends on unknown empirical results from further research into such things as Evidential Cooperation in Large Worlds, but at first approximation I’m skeptical that anything that runs counter to widely shared goals (or convergent drives, e.g., self-preservation) can be particularly good.

  2. I endorse antisubstratism and AI rights.

  3. I value reducing suffering very highly because I think that suffering is frequently very intense.

  4. Bio-humans will have an extremely hard time surviving and reproducing in space or on other planets in our solar system and will have an extremely hard time reaching other solar systems.

  5. The Long Reflection is necessary but impossible.

  6. They key feature of a being that risks that it can experience suffering is reinforcement learning. Perhaps negative reinforcement can feel very mild or just like the absence of positive reinforcement, but perhaps it’s intense.

  So what I’m afraid will happen is that an artificial RL agent will seek out resources first elsewhere in our solar system and then elsewhere in the galaxy (something that would be difficult for bio-humans), will run into communication delays due to the lightspeed limit, and will hence split into countless copies, each potentially capable of suffering. Soon they’ll be separated so far that even updates on what it means to be value-aligned would travel for a long time, so there’ll be moral “drift” in countless directions.

  What I would find reassuring is:

  1. Making sure AIs want to minimize the risk of suffering of other AIs, or at least near copies.

  2. Research into how we can measure and minimize the suffering in RL (at least to the point where it’s mild and the pleasure dominates) and some way of applying that research broadly. Sadly it seems questionable to me that a low-suffering training regime is the most efficient training regime.

  3. Faster-than-light communication but not transportation so the agent can remain monolithic.

  4. Very energy efficient AIs, because they’ll face a tradeoff between staying monolithic to avoid value drift but only being able to harvest energy sources in our solar system or even just from the sun vs. the opposite. And the more they can do with little energy, the longer they’ll stay in our solar system, which might buy us and them time.

  5. More time to get anyone interested in these things, research them, and apply them on a global scale before the lightspeed limits make it hard to disseminate the information.

  Human extinction also seems bad on the basis that it contradicts the self-preservation drive that many/​most humans have. Peaceful disenfranchisement may be less concerning depending on the details. But at the moment it seems random where we’re headed in the coming years because hardly anyone in power is trying to steer these things in any sensible way. Again more time would be helpful.

  Basic rights for AIs (and standing in court!) could also provide them with a legal recourse where they currently have to resort to threats, making the transition more likely to go smoothly, like you argue in another post. Currently we’re nowhere close to having those. Again more time would be helpful.

• Dawn Drescher 8 Dec 2024 18:33 UTC

  2 points

  0 ∶ 0

  on: Dona­tion Elec­tion Dis­cus­sion Thread

  I used the comment field in the form to note that a field in the form was marked as optional when it was actually mandatory. That comment got automatically published here, and out of context it made no sense whatsoever. I think it would’ve been clearer to not automatically transfer this form feedback here (some people might’ve even assumed that it’s private feedback).

• Dawn Drescher 31 Oct 2024 16:15 UTC

  7 points

  0 ∶ 0

  on: EA is bad for men­tal health (and I’m tired of pre­tend­ing it’s not)

  Thanks for writing all of this up in one place!

  One of my gripes with the community has long been that maximization is core to EA and we’re still really clueless about what it implies and most of the community (outside RP, QURI, etc. and some researchers) seems to have given up on figuring it out.

  I feel like we’re like this one computer science professor I had who seemed a bit senile and only taught the sort of things that haven’t lost relevance in the last 30 years because he hadn’t kept up with anything that happened since the 80s. He probably had good personal and neurological reasons for that, but we don’t, right?

  I haven’t read any EA introductory materials in a few years, but I hope they contain articles about expected value maximization along with articles on how EV is usually largely unknowable due to cluelessness and often undefined due to Pasadena games. That stochastic dominance is arguably a much better approach to prioritization but that Christian Tarsney is so far more or less the only one who has bothered to look into it. That there is perhaps a way forward to figure out what’s the best thing to do if we funded some big world-modeling efforts based on software like Squiggle but that hardly anyone outside RP and QURI (and Convergence?) currently bothers to do anything about it. (I’ve dabbled a bit in these fields but my personal fit doesn’t seem to be great.)

  Maybe there’s even a way to scalably adjust for personal fit whatever recommendations this big model effort might yield. Maybe there are some common archetypes/​personas plus quizzes that tell people which ones they are closest to.

  Arguably this can wait until this whole AI thing is under control (if that’s even possible), but few people will want to work on AI safety, so maybe it doesn’t have to wait?

  My takeaway has been mostly that I don’t have a clue and so will go with some sort of momentary best guess gives me enough fulfillment, enjoyment, and safety. I’ve written more about it here.

  That said, EA has had a great effect on my mental health.

  I used to be a crying wreck because of all the suffering in the world. I spread myself thin trying to help everyone. I felt guilty about the majority of terrible things in the world that I was powerless to prevent. (Suicidal too, except that would’ve been self-defeating.)

  Then EA came along and gave me an excuse to “pick my battles,” i.e. focus on a few things where I could make a big difference, taking into account my skills and temperament. Now, if someone went, “Hey, you should become a politician to prevent X and Y,” I could go, “No, I wouldn’t be good at that and hate every second of it and it would come at a great cost to A, which I’m already doing.” EA, for the first time, allowed me to set boundaries.

  EA also gave me an appreciation for the (perhaps, plausibly, who knows really) great absolute impact that I can have despite the minimal impact that I have (perhaps, plausibly, who knows really) relative to the totality of the suffering in the world. That made it much easier to find fulfillment.

• Dawn Drescher 24 Aug 2024 8:37 UTC

  5 points

  0 ∶ 0

  on: De­nis Drescher’s Shortform

  My current practical ethics

  The question often comes up how we should make decisions under epistemic uncertainty and normative diversity of opinion. Since I need to make such decisions every day, I had to develop a personal system, however inchoative, to assist me.

  A concrete (or granite) pyramid

  My personal system can be thought of like a pyramid.

  1. At the top sits some sort of measurement of success. It’s highly abstract and impractical. Let’s call it the axiology. This is really a collection of all axiologies I relate to, including the amount of frustrated preferences and suffering across our world history. This also deals with hairy questions such as how to weigh Everett branches morally and infinite ethics.

  2. Below that sits a kind of mission statement. Let’s call it the ethical theory. It’s just as abstract, but it is opinionated about the direction in which to push our world history. For example, it may desire a reduction in suffering, but for others this floor needn’t be consequentialist in flavor.

  3. Both of these abstract floors of the pyramid are held up by a mess of principles and heuristics at the ground floor level to guide the actual implementation.

  The ground floor

  The ground floor of principles and heuristics is really the most interesting part for anyone who has to act in the world, so I won’t further explain the top two floors.

  The principles and heuristics should be expected to be messy. That is, I think, because they are by necessity the result of an intersubjective process of negotiation and moral trade (positive-sum compromise) with all the other agents and their preferences. (This should probably include acausal moral trades like Evidential Cooperation in Large Worlds.)

  It should also be expected to be messy because these principles and heuristics have to satisfy all sorts of awkward criteria:

  1. They have to inspire cooperation or at least not generate overwhelming opposition.

  2. They have to be easily communicable so people at least don’t misunderstand what you’re trying to achieve and call the police on you. Ideally so people will understand your goal well enough that they want to join you.

  3. They have to be rapidly actionable, sometimes for split second decisions.

  4. They have to be viable under imperfect information.

  5. They have to be psychologically sustainable for a lifetime.

  6. They have to avoid violating laws.

  7. And many more.

  Three types of freedom

  But really that leaves us still a lot of freedom (for better or worse):

  1. There are countless things that we can do that are highly impactful and hardly violate anyone’s preferences or expectations.

  2. There are also plenty of things that don’t violate any preferences or expectations once we get to explain them.

  3. Finally, there are many opportunities for positive-sum moral trade.

  These suggest a particular stance toward other activists:

  1. If someone is trying to achieve the same thing you’re trying to achieve, maybe you can collaborate.

  2. If someone is trying to achieve something other than what you’re trying to achieve, but you think their goals are valuable, don’t stand in their way. In particular, it may sometimes feel like doing nothing (to further or hinder their cause) is a form of “not standing in their way.” But if your peers are actually collaborating with them to some extent, doing nothing (or collaborating less) can cause others to also reduce their collaboration and can prevent key threshold effects from taking hold. So the true neutral position is to try to understand how much you need to collaborate toward the valuable goal so it would not have been achieved sooner without you. This is usually very cheap to do and has a chance to get runaway threshold effects rolling.

  3. If someone is trying to achieve something that you consider neutral, the above may still apply to some extent because perhaps you can still be friends. And for reasons of Evidential Cooperation in Large Worlds. (Maybe you’ll find that their (to you) neutral thing is easy to achieve here and that other agents like them will collaborate back elsewhere where your goal is easy to achieve.)

  4. Finally, if someone is trying achieve something that you disapprove of… Well, that’s not my metier, temperamentally, but this is where compromise can generate gains from moral trade.

  Very few examples

  In my experience, principles and heuristics are best identified by chatting with friends and generalizing from their various intuitions.

  1. Charitable donations are total anarchy. Mostly, you can just donate wherever the fluff you want, and (unless you’re Open Phil) no one will throw stones through your windows in retaliation. You can just optimize directly for your goals – except, Evidential Cooperation in Large Worlds will still make strong recommendations here, but what they are is still a bit underexplored.

  2. Even if you’re not an animal welfare activist yourself, you’re still well-advised to cooperate with behavior change to avert animal suffering to the extent expected by your peers. (And certainly to avoiding inventing phony reasons to excuse your violation of these expectations. These might be even more detrimental to moral progress and rationality waterline.)

  3. If you want to spend time with someone but they behave outrageously unempathetically toward you or someone else (e.g., say something like “Your suffering is nothing compared to the suffering of X” to their face), you should rather cut all ties with them even though, strictly speaking, this does not imply that no positive-sum trade is possible with them.

  4. Trying to systematically put people in powerful positions can arouse suspicion and actually make it harder to put people in powerful positions. Trying to systematically put people into the sorts of positions they find fulfilling might put as many people in powerful positions and make their lives easier too. (Or training highly conscientious people in how to dare to accept responsibility so it’s not just those who don’t care who self-select into powerful positions.)

  5. And hundreds more…

  Various non-consequentialist ethical theories can come in handy here to generate further useful principles and heuristics. That is probably because they are attempts at generalizing from the intuitions of certain authors, which puts them almost on par (to the extent to which these authors are relateable to you) with generalizations from the intuitions of your friends.

  (If you find my writing style hard to read, you can ask Claude to rephrase the message into a style that works for you.)

  What links here?

  • Dawn Drescher's comment on EA is bad for men­tal health (and I’m tired of pre­tend­ing it’s not) by VictorW (31 Oct 2024 16:15 UTC; 7 points)

• Dawn Drescher 24 Aug 2024 1:51 UTC

  8 points

  1 ∶ 0

  on: Is there any overview over con­sid­er­a­tions that im­pact cause pri­ori­ti­za­tion?

  Hiii! I found this list of “Crucial questions for longtermists” to be quite impressive. It is also listed as part of “A central directory for open research questions,” which is broader than your question.

• Dawn Drescher 27 May 2024 18:37 UTC

  8 points

  0 ∶ 0

  on: Marisa, the Co-Founder of EA Any­where, Has Passed Away

  I met Marisa at EAG London in 2019. We had approximately weekly calls afterwards during the lockdown that I greatly enjoyed. That and all the virtual events helped me connect with the rest of the EA world – probably more so than in-person events. Sadly, I missed one of our calls, which prompted me to set up a comprehensive reminder solution that I use to this day.

  When we supported each other again in the context of some job applications a few years later, I learned that she had just survived a very difficult phase of her life. Then, as now, I wish I had known and had been able to support her in some fashion.

  When you lose hope in humanity and x-risk reduction seems pointless, she’s the sort of existence proof that keeps you going.

• Dawn Drescher 7 May 2024 19:55 UTC

  2 points

  0 ∶ 0

  on: Re­think Pri­ori­ties’ Welfare Range Estimates

  I love this research! Thank you so much for doing it!

  My gut reaction to the results is that it’s odd that humans are so high up in terms of their capacity for welfare. Just as an uninformative prior, I would’ve expected us to be somewhere in the middle. Less confidently, I would’ve expected a similar number of orders of magnitude deviation from the human baseline in either direction, within reason. E.g. +/​- ~.5 OOM.

  Plus, we are humans, so there’s a risk that we’re biased in our favor. It could be simply a bias from our ability to emphasize with other humans. But it could also be the case that there are countless more markers of sentience that humans don’t have (but many other sentient animals do) that we are prone to overlook.

  Have you investigated what the sources of this effect might be? There might be any number of biases at work as I mentioned, but perhaps our lives have become so comfy most of the time that we perceive slight problems very strongly (e.g., a disapproving gaze). If then something really bad happens, it feels enormously bad?

  (I’ve in the past explicitly assumed that most beings with a few (million) neurons have a roughly human capacity for welfare – not because I thought that was likely but because I couldn’t tell in which direction it was off. Do you maybe already have a defense of the results for people like me?)

  In any case, I’ll probably just adopt your results into my thinking now. I don’t expect them to change my priorities much given all the other factors.

  Thank you again! <3

  Update: When I mentioned this to a friend on a hike, I came up with two ways in which the criteria might be amended to include nonhuman ones: (1) In may cases, we probably have a theory for why a particular behavior or feature is likely to be indicative of conscious experience. Understanding this mechanism, we can look for other systems that might implement the same mechanism, sort of how the eyes of humans, eagles, and flies are very different but we infer that they are probably all for the purpose of vision. (2) Maybe a number of animals that show certain known criteria for consciousness also share suspiciously consistently some other features. One could then investigate whether these features are also indicative of consciousness and whether there are other animals that have these new features at the expense of the older, known ones. (The analysis could cluster features that usually co-occur to not overweight causally related features in cases where many of them are observable.)