If you hold a longtermist view & you’re a consequentialist (as far as I know, most longtermists are consequentialist), consequentialist cluelessness seems like a major theoretical problem
Basically, trying to assess outcomes occurring 10,000+ years in the future breaks the consequentialism algorithm (because we have no visibility into these outcomes)
Given this theoretical problem, longtermist cause prioritization should include “how robust to cluelessness is this?” as a major factor
Some x-risk interventions seem pretty robust to cluelessness
Also robust: interventions that increase the set of well-intentioned + capable people
The psychedelic experience also seems like a plausible lever on increasing capability (via reducing negative self-talk & other mental blocks) and improving intentions (via ego dissolution changing one’s metaphysical assumptions)
Ergo: under a longtermist view, psychedelic interventions are plausibly in the same ballpark of effectiveness as x-risk interventions & other interventions that increase the set of well-intentioned + capable people
Psychedelics are showing a ton of promise as treatment for a battery of chronic mental health issues (including anxiety, depression, OCD, and addictive disorders including smoking & alcoholism)
Ergo: under a person-affecting view, psychedelic interventions are plausibly in the same ballpark of effectiveness as global poverty interventions
If there’s sufficient interest, I’ll take the time to make the above argument more rigorous.
[Disclosure: I’m helping out Atman Retreat as an advisor; views are my own]
This puzzled me slightly. One reason is that longtermism and person-affecting views are different categories; the former is a view about where, in practice, value lies and the latter is a view about where, in theory, value lies. You could be a totalist (all possible people matter), which is not a person-affecting view, but be a near-termism. I think a better set up would have been: ‘psychedelics look good whether you just value the near-term or the long-term’. I suppose that leaves out the ‘medium-termists’, but I don’t know how many people there are who hold this view, whatever it is, inside or outside EA.
Also robust: interventions that increase the set of well-intentioned + capable people
The psychedelic experience also seems like a plausible lever on increasing capability (via reducing negative self-talk & other mental blocks) and improving intentions (via ego dissolution changing one’s metaphysical assumptions)
I would like you to say more about this. It seems plausible to me that training rationality is orders of magnitude more impactful for the longrun, so this is an objection to counter.
under a longtermist view, psychedelic interventions are plausibly in the same ballpark of effectiveness as x-risk interventions
I don’t think you’ve shown this. It’s more plausible to me that Xrisk is a top tier intervention and rationality and the ‘mindset-changingness’ of psychedelics are in the lower tiers. This would still make them potentially very interesting from a long-termist perspective—in the bucket of ‘things to do take seriously and possibly fund if X-risk has absorbed as many resources as it can’.
Just FYI, I wrote a mammothseriesofarticles on drug policy reform 18 months or so ago where I argued that psychedelics for mental health looks very promising from the near term perspective. In other words, I explicitly claim what you’re claiming! I haven’t had a chance to do more work on it since and I add the usual caveats about not necessarily agree with everything past-Michael wrote.
Also, just because psychedelics are promising as a category of intervention, it doesn’t follow that setting up a retreat of this kind is the best way to go within that (sub)cause area. You’d need to argue for that too.
I don’t think you’ve shown this. It’s more plausible to me that Xrisk is a top tier intervention and rationality and the ‘mindset-changingness’ of psychedelics are in the lower tiers. This would still make them potentially very interesting from a long-termist perspective...
Being in the same tier as rationality interventions is basically what I meant by “being in the same ballpark,” though my language wasn’t very precise.
Also I’ll note that “x-risk intervention” is a pretty loose category:
Doing agent foundations research at MIRI? Definitely an x-risk intervention.
Doing AI safety work at OpenAI? Definitely an x-risk intervention.
Doing AI development work at OpenAI? Maybe that’s an x-risk intervention, though not totally clear how doing AI dev at OpenAI is different from doing AI dev at a FANG company.
Doing AI policy work at CSET? Sure, that’s probably an x-risk intervention, though really what CSET is trying to do is influence policymakers’ opinions about AI (which seems pretty similar to CFAR’s mission, though the target market is different).
Leading rationality workshops at CFAR? In the taxonomy you gave, CFAR is outside the category of “x-risk interventions,” though it’s plausibly doing stuff aimed at x-risk reduction in the same way that e.g. CSET is doing stuff aimed at x-risk reduction.
I would like you to say more about this. It seems plausible to me that training rationality is orders of magnitude more impactful for the longrun, so this is an objection to counter.
Yes, there’s a lot to say on this.
Comparing rationality training programs (e.g. CFAR, e.g. Paradigm Academy) to psychedelic trips is tricky. It’s hard to make an apples-to-apples comparison, because the interventions are operating on very different levels of abstraction.
The rationality training programs I know of operate almost entirely on the conceptual level (though I believe Paradigm uses some bodywork modalities also). The basic structure of conceptual rationality training is something like:
Instructor says some words about a rationality topic
Trainee hears these words & tries to internalize the topic
Trainee practices their internalized version of the rationality topic (by themselves, with other trainees, or with the instructor)
Instructor provides feedback to trainee to improve the trainee’s internalized model of the topic
I think this structure can work really well for information & technique transfer, especially when the trainee is engaged & the instructor is skillful.
The basic structure of a psychedelic trip is very different:
Tripper thinks about and articulates the intentions & expectations they have about their upcoming psychedelic experience (to themselves, or to a facilitator)
Tripper ingests a psychedelic (by themselves, or with a sober facilitator present)
Tripper has a psychedelic experience. A wide range of subjective experiences can happen:
Old memories can come up and/or become salient
New perspectives about friends, family, one’s immediate environment can be adopted
Emotions can be felt very intensely, especially emotions about salient people & topics in the tripper’s life
Insights (or more skeptically, “insights”) can be had about the tripper’s psychology, social assumptions, epistemic assumptions, and metaphysical assumptions
New personal narratives (“this is the story of my life; this is what my life’s about”) can be adopted
Once sober, tripper integrates the experience (by themselves, or in dialogue a facilitator)
How did the actual trip match up to your expectations about the trip?
What came up? What was interesting? What was trivial, or silly?
Did anything come up that’s worth incorporating into your everyday life?
I think this structure can be very helpful for surfacing emotional blocks (e.g. akratic feelings), as well as for resolving known emotional blocks.
The psychedelic experience can also help change one’s assumptions + internal monologue + personal narrative, usually in ways that are considered helpful. (Note that it doesn’t do this automatically, it can just help “loosen you up.” You still have to opt in to making changes about these things.)
So, to the extent that we’re limited by information & technique transfer, I’d expect conceptual rationality training to be more leveraged.
And to the extent that we’re limited by emotional blocks & unhelpful personal narratives, I’d expect the psychedelic experience to be more leveraged.
(I’m not claiming that the psychedelic experience is the only thing that helps with emotional blocks, etc. Many other techniques can also help, though the other techniques I know of tend to take much longer to bear fruit.)
I agree this is plausible, but I think you would accept that this is conjecture and still quite a long way from what we want, which I assume is some sort of quantified, evidence-based, comparative analysis.
For sure. This is a just a theoretical comparison – some kind of quantified comparative analysis would be great, though as far as I know nothing like that exists at present.
Edit: Also note that there aren’t many analyses that compare across EA cause areas. (e.g. comparing animal welfare interventions to global poverty interventions, e.g. comparing x-risk interventions to animal welfare interventions.)
Michael Dickens’ cause prioritization app attempts this, though as far as I know it hasn’t been used to drive much decision-making.
For additional background, here’s a short (& lossy) argument for psychedelics as Cause X:
Almost everyone in EA holds either a longtermist view or a person-affecting view
If you hold a longtermist view & you’re a consequentialist (as far as I know, most longtermists are consequentialist), consequentialist cluelessness seems like a major theoretical problem
Basically, trying to assess outcomes occurring 10,000+ years in the future breaks the consequentialism algorithm (because we have no visibility into these outcomes)
Given this theoretical problem, longtermist cause prioritization should include “how robust to cluelessness is this?” as a major factor
Some x-risk interventions seem pretty robust to cluelessness
Also robust: interventions that increase the set of well-intentioned + capable people
CFAR & Paradigm Academy are aimed at this
The psychedelic experience also seems like a plausible lever on increasing capability (via reducing negative self-talk & other mental blocks) and improving intentions (via ego dissolution changing one’s metaphysical assumptions)
Ergo: under a longtermist view, psychedelic interventions are plausibly in the same ballpark of effectiveness as x-risk interventions & other interventions that increase the set of well-intentioned + capable people
If you hold a person-affecting view, mental health seems like a cause area on par with global poverty (see Michael Plant’s cause profile on mental health)
Psychedelics are showing a ton of promise as treatment for a battery of chronic mental health issues (including anxiety, depression, OCD, and addictive disorders including smoking & alcoholism)
Ergo: under a person-affecting view, psychedelic interventions are plausibly in the same ballpark of effectiveness as global poverty interventions
If there’s sufficient interest, I’ll take the time to make the above argument more rigorous.
[Disclosure: I’m helping out Atman Retreat as an advisor; views are my own]
A couple of comments
This puzzled me slightly. One reason is that longtermism and person-affecting views are different categories; the former is a view about where, in practice, value lies and the latter is a view about where, in theory, value lies. You could be a totalist (all possible people matter), which is not a person-affecting view, but be a near-termism. I think a better set up would have been: ‘psychedelics look good whether you just value the near-term or the long-term’. I suppose that leaves out the ‘medium-termists’, but I don’t know how many people there are who hold this view, whatever it is, inside or outside EA.
I would like you to say more about this. It seems plausible to me that training rationality is orders of magnitude more impactful for the longrun, so this is an objection to counter.
I don’t think you’ve shown this. It’s more plausible to me that Xrisk is a top tier intervention and rationality and the ‘mindset-changingness’ of psychedelics are in the lower tiers. This would still make them potentially very interesting from a long-termist perspective—in the bucket of ‘things to do take seriously and possibly fund if X-risk has absorbed as many resources as it can’.
Just FYI, I wrote a mammoth series of articles on drug policy reform 18 months or so ago where I argued that psychedelics for mental health looks very promising from the near term perspective. In other words, I explicitly claim what you’re claiming! I haven’t had a chance to do more work on it since and I add the usual caveats about not necessarily agree with everything past-Michael wrote.
Also, just because psychedelics are promising as a category of intervention, it doesn’t follow that setting up a retreat of this kind is the best way to go within that (sub)cause area. You’d need to argue for that too.
Being in the same tier as rationality interventions is basically what I meant by “being in the same ballpark,” though my language wasn’t very precise.
Also I’ll note that “x-risk intervention” is a pretty loose category:
Doing agent foundations research at MIRI? Definitely an x-risk intervention.
Doing AI safety work at OpenAI? Definitely an x-risk intervention.
Doing AI development work at OpenAI? Maybe that’s an x-risk intervention, though not totally clear how doing AI dev at OpenAI is different from doing AI dev at a FANG company.
Doing AI policy work at CSET? Sure, that’s probably an x-risk intervention, though really what CSET is trying to do is influence policymakers’ opinions about AI (which seems pretty similar to CFAR’s mission, though the target market is different).
Leading rationality workshops at CFAR? In the taxonomy you gave, CFAR is outside the category of “x-risk interventions,” though it’s plausibly doing stuff aimed at x-risk reduction in the same way that e.g. CSET is doing stuff aimed at x-risk reduction.
Yes, there’s a lot to say on this.
Comparing rationality training programs (e.g. CFAR, e.g. Paradigm Academy) to psychedelic trips is tricky. It’s hard to make an apples-to-apples comparison, because the interventions are operating on very different levels of abstraction.
The rationality training programs I know of operate almost entirely on the conceptual level (though I believe Paradigm uses some bodywork modalities also). The basic structure of conceptual rationality training is something like:
Instructor says some words about a rationality topic
Trainee hears these words & tries to internalize the topic
Trainee practices their internalized version of the rationality topic (by themselves, with other trainees, or with the instructor)
Instructor provides feedback to trainee to improve the trainee’s internalized model of the topic
I think this structure can work really well for information & technique transfer, especially when the trainee is engaged & the instructor is skillful.
The basic structure of a psychedelic trip is very different:
Tripper thinks about and articulates the intentions & expectations they have about their upcoming psychedelic experience (to themselves, or to a facilitator)
Tripper ingests a psychedelic (by themselves, or with a sober facilitator present)
Tripper has a psychedelic experience. A wide range of subjective experiences can happen:
Old memories can come up and/or become salient
New perspectives about friends, family, one’s immediate environment can be adopted
Emotions can be felt very intensely, especially emotions about salient people & topics in the tripper’s life
Insights (or more skeptically, “insights”) can be had about the tripper’s psychology, social assumptions, epistemic assumptions, and metaphysical assumptions
New personal narratives (“this is the story of my life; this is what my life’s about”) can be adopted
Once sober, tripper integrates the experience (by themselves, or in dialogue a facilitator)
How did the actual trip match up to your expectations about the trip?
What came up? What was interesting? What was trivial, or silly?
Did anything come up that’s worth incorporating into your everyday life?
I think this structure can be very helpful for surfacing emotional blocks (e.g. akratic feelings), as well as for resolving known emotional blocks.
The psychedelic experience can also help change one’s assumptions + internal monologue + personal narrative, usually in ways that are considered helpful. (Note that it doesn’t do this automatically, it can just help “loosen you up.” You still have to opt in to making changes about these things.)
So, to the extent that we’re limited by information & technique transfer, I’d expect conceptual rationality training to be more leveraged.
And to the extent that we’re limited by emotional blocks & unhelpful personal narratives, I’d expect the psychedelic experience to be more leveraged.
(I’m not claiming that the psychedelic experience is the only thing that helps with emotional blocks, etc. Many other techniques can also help, though the other techniques I know of tend to take much longer to bear fruit.)
I agree this is plausible, but I think you would accept that this is conjecture and still quite a long way from what we want, which I assume is some sort of quantified, evidence-based, comparative analysis.
For sure. This is a just a theoretical comparison – some kind of quantified comparative analysis would be great, though as far as I know nothing like that exists at present.
Edit: Also note that there aren’t many analyses that compare across EA cause areas. (e.g. comparing animal welfare interventions to global poverty interventions, e.g. comparing x-risk interventions to animal welfare interventions.)
Michael Dickens’ cause prioritization app attempts this, though as far as I know it hasn’t been used to drive much decision-making.
Good point – I agree that the near-term / long-term distinction is better for this.