Effective Altruism Forum
EA Forum

All of CEvans's Comments + Replies

Against Learning From Dramatic Events (by Scott Alexander)

I downvoted this forum post because I think the quoted part of the text, while obviously informal, is an annoying strawman of criticisms EA faced and represents an attitude towards critique that I think is quite counterproductive. I think the rest of the linked post is significantly better though, and agree with the general point.

EA Philippines’ Career Planning Retreat for Students and Professionals

CEvans8mo4

Thanks a lot for posting this! I really enjoyed reading it and the linked google document - would anyone in the EA Philippines team be interested in a short meeting with me about this? I currently run EA Oxford and have some specific questions.

Alethea Faye Cendaña

8mo

Hi there! I'd love to have a chat with you :D You can contact me on my LinkedIn :D

Taking prioritisation within 'EA' seriously

CEvans8mo2

Thanks for the thoughtful comment Amber! I appreciate the honesty in saying both that you think people should think more about prioritisation and that you haven't always yourself. I have definitely been like this at times and I think it is good/important to be able to say both statements together. I would be happy/interested to talk through your thinking about prioritisation if you wanted. I have some other accounts of people finding me helpful to talk to about that kind of thing as it happens frequently in my community building work.

Re. (1), I agre... (read more)

Should I train to become a mediator for the EA ecosystem?

CEvans8mo8

Perhaps another consideration against is that it seems potentially bad to me for any one person to be the primary mediator for the EA community. There are some worlds where this position is subtly very influential. I dont think I would want a single person/worldview to have that, in order to avoid systematic mistakes/biases. To be clear, this is not intended as a personal comment - I have no context on you besides this post.

I am excited about having better community mediation though. Perhaps you coordinating a group/arrangement with external people could be a great idea.

Also I think this kind of post about personal career plans with detailed considerations is great so thanks for writing it.

Severin

8mo

Well, good that my values are totally in line with the correct trajectory for EA then! No, but seriously: I have no idea how to fix this. The best response I can give is: I'd suspect that having one mediator is probably still better than having zero mediators. Let's not make the perfect the enemy of the good. Plus, it's an essential part of the role to just be a catalyst for the conflict parties rather than try and steer the outcome towards any particular direction. (Of course, that is an ideal that is not perfectly aligned with how humans actually work.) So far, every single time I've done ops work without guidance and under precarious financial circumstances has made me miserable and lead to outcomes I was less than satisfied with. I'm definitely not the right person to do this. Plus, I have some evidence this will probably not work within any reasonable amount of effort: One person with an insider perspective of many EA orgs' conflicts said that so far, the limiting factor for hiring an external mediator was having one available who is sufficiently trusted. I.e., being known and trusted in the community is crucial for actually doing this. It's hard enough to build a reputation for myself, even if I'm around at conferences and in the forums a lot. Building a reputation on behalf of external mediators I work with seems like a near impossible task.

Nuclear winter scepticism

CEvans8mo3

Thanks David that all makes sense. Perhaps my comment was poorly phrased but I didn't mean to argue for caring about infohazards per se, but was curious for opinions on it as a consideration (mainly poking to build my/others'understanding of the space ). I agree that imposing ignorance on affected groups is bad by default.

Do you think the point I made below in this thread regarding pressure from third party states is important? Your point "it doesn't matter to them whether it also devastates agriculture in Africa or Australia" doesn't seem obviously true ... (read more)

Davidmanheim

8mo

I don't think pressure from third-party states is geostrategically relevant for most near-term decisions, especially because there is tremendous pressure already around the norm against nuclear weapons usage. I strongly agree that the default should be openness, unless there is a specific reason for concern. And I even more strongly agree that honesty is critical for government and academia - whihc is why I'm much happier with banning research because of publicly acknowledged hazards, and preventing the discovery of information that might pose hazards, rather than lying about risks if they are discovered.

Nuclear winter scepticism

CEvans8mo3

Thanks for the reply and link to the study - I feel quite surprised by how minor the effect of impact awareness is but I suppose nuclear war feels quite salient for most people. I wonder if this could be some kind of metric used for evaluating the baseline awareness of a danger (ie. I would be very interested to see the same study applied to pandemics, AI, animals etc)

Re. The effects on government decision making, I think I agree intuitively that governments are sufficiently scope insensitive (and self interested in nuclear war circumstances?) that it woul... (read more)

Vasco Grilo

8mo

You are welcome! It is worth having in mind that the intervention was only 1 min, so it is quite low cost, and even a small effect size can result in high cost-effectiveness. Right, to be honest, that sounds plausible to me too (although I would rather live in the world where nuclear winter was not a thing!). The countries with nuclear weapons "only" have 55.6 % of global GDP[1], so third parties should still exerce some reasonable influence even if it may be limited by alliances. In that case, finding out nuclear war had negligible climatic effects would counterfactually increase, in the sense of continuing to fail to decrease, the expected damage of nuclear war. Another important dynamic is the different climatic impacts across countries. Here are the results from Fig. 4 of Xia 2022 for the 27 Tg scenario (closest to the 30 Tg expected by Luísa): The climatic effects are smaller in the US than in China and Russia. So the US may have a military incentive to hide that nuclear winter is real, while secretely preparing for it, such that China and Russia are caught unprepared in case of a nuclear war. Tricky... In general, I still abide by principles like: * Having less weapons of mass destruction (e.g. nuclear weapons) is almost always good. * Governments being honest with their citizens (e.g. about nuclear winter) is almost always good. 1. ^ From The World Bank, the countries with nuclear weapons except for North Korea had a combined GDP in 2022 of 55.8 T$ (25.46 T$ from the US, 17.96 T$ from China, 3.39 T$ from India, 3.07 T$ from the UK, 2.78 T$ from France, 2.24 T$ from Russia, 0.52203 T$ from Israel, and 0.37653 T$ from Pakistan), and the global GDP was 100.56 T$.

Nuclear winter scepticism

CEvans8mo4

Thanks for writing this - it seems very relevant for thinking about prioritization and more complex X-risk scenarios.

I haven't engaged enough to have a particular object-level take, but was wondering if you /others had a take on whether we should consider this kind of conclusion somewhat infohazardous? Ie. Should we be making this research public if it at all increases the chance that nuclear war happens?

This feels like a messy thing to engage with, and I suppose it depends on beliefs around honesty and trust in governments to make the right call with fuller information (of course there might be some situations where initating a nuclear war is good).

Davidmanheim

8mo

I think it's not at all a reasonable place to worry about infohazard, for 2 reasons related to incentives. First, the decisionmakers in a nuclear conflict are very likely going to die regardless - they are located in large cities or other target areas. Whether or not there is a nuclear winter is irrelevant. Second, the difference between hundreds of millions dead and billions dead is morally critical, but strategically doesn't matter - if the US is involved in a large scale nuclear exchange with Russia or China, it's going to be a disaster for the US, and it doesn't matter to them whether it also devastates agriculture in Australia and Africa. On top of this, I think that this is a bad situation to argue for infohazard risk, since hiding the damage or lack thereof of an extant risk is imposing ignorance on an affected group. This wouldn't be critical if the infohazard creates the risk, but here it does not.

Vasco Grilo

8mo

Thanks for commenting! Paul Ingram has done some relevant research about the effects of increasing awareness for the impact of a nuclear war. The study involved 3 k people, half from the US (1.5 k), and half from the UK (1.5 k). The half of the citizens (750 from the US, and 750 from the UK) in the treatment group were shown these infographics for 1 min: The numbers from the 1st graphic are from Xia 2022, whose 150 Tg scenario (3rd row of the 1st graphic) considers the 4.4 k nuclear detonations studied in Toon 2008. These were the effects on opposition/support for nuclear retalation: I am not a fan of this way of analysing the results. I believe it would have been better to assign a score to each of the levels of opposition/support (e.g. ranging from -2 for strong opposition to 2 for strong support). Then one could calculate the effect size of the intervention from ("mean score in the treatment group" - "mean score in the control group")/"standard deviation in the 2 groups", and report the p-value. Having lots of categories allows one to test many different hypotheses, and then selectively report the statistically significant ones. In any case, the 1st infographic contains information about both the direct and indirect deaths, so it is not straightforward to interpret whether the increase in opposition to retaliation illustrated just above was due to gaining awareness of the direct deaths from the detonations, or indirect ones from the nuclear winter. The survey asks one question to understand what motivated the change in the level of opposition/support (see Q2 in the last page). One of the potential answers was "Avoid risk of killing civilians in other countries, or triggering global famine" (4), which refers to nuclear winter. The study does not report the results of the answers to Q2. However, I speculate the difference between the fraction of people answering 4 in the control and treatment group was not statistically significant at a 95 % confidence level (

My independent impressions on the EA Handbook

CEvans9mo9

Thanks for writing this post Victor, I think your context section represents a really good and truth-seeking attitude coming into this with. From my perspective, it is also always good to have good critiques of key EA ideas. To respond to your points:

1 and 2. I agree that the messaging about maximisation has the danger of people taking it too far, but I think it is quite defensible as an anchor point. Maybe this should be more present in the handbook, but I think it is worth initially saying that >95% of EAs' lives don't look like some extre... (read more)

VictorW

9mo

Thanks for the clarification about how 1 and 2 may look very different in the EA communities. I'm not particularly concerned about the thought that people might be out there taking maximization too far, the framing of my observations is more like "well here's what going through the EA Handbook may prompt me to think about EA ideas or what other EAs may believe. After thinking about your reply, I realized that I made a bunch of assumptions based on things that might just be incidental and not strongly connected. I came to the wrong impression that the EA Handbook is meant to be the most canonical and endorsed collection of EA fundamentals. Here's how I ended up there. In my encounters hearing about EA resources, the Handbook is the only introductory "course", and presumably due to being the only one of its kind, it's also the only one that's been promoted to me via over multiple mediums. So I assumed that it must be the most official source of introduction, remaining alone in that spot over multiple years, seeing it bundled with EA VP also seemed like an endorsement. I also made the subconscious assumption that since there's plenty of alternative high quality EA writing out there, as well as resources put into writing, that the Handbook as a compilation is probably designed to be the most representative collection of EA meta, otherwise it wouldn't still be promoted the way it has been to me. I've had almost no interaction with the EA Forum before reading the Handbook, so very limited prior context to gauge how "meta" the Handbook is among EA communities, or how meta any of its individual articles are. (Which now someone has helpfully provided a bunch of reading material that is also fundamental but while having quite different perspectives.)

How can we improve Infohazard Governance in EA Biosecurity?

CEvans9mo6

Thanks for writing this, I found it helpful for understanding the biosecurity space better!

I wanted to ask if you had advice for handling the issue around difficulties for biosecurity in cause prioritisation as a community builder.

I think it is easy to build an intuitive case for biohazards not being very important or an existential risk, and this is often done by my group members (even good fits for biosecurity like biologists and engineers), who then dismiss the area in favour of other things. They (and me) do not have access to the threat models which p... (read more)

Tessa9mo17

This is more a response to "it is easy to build an intuitive case for biohazards not being very important or an existential risk", rather than your proposals...

My feeling is that it is fairly difficult to make the case that biological hazards present an existential as opposed to catastrophic risk and that this matters for some EA types selecting their career paths, but it doesn't matter as much in the grand scale of advocacy? The set of philosophical assumptions under which "not an existential risk" can be rounded to "not very important" seems common in th... (read more)

NickLaing

9mo

Nice comment, to respond to your options 1. Deference doesn't seem ideal, seems against the norms of the EA community 2. Like you say seems very feesable. I would be surprised if there wasn't something like this already? And even you could make the point that the threat models used aren't even the highest risk - others that you don't talk about could be even worse. 3. Obviously not ideal

Why isn't there a charity evaluator for longtermist projects?

CEvans9mo1

What's the kind of information you mean by semi-objective? Something comparable to this for instance? Nuclear Threat Initiative’s Global Biological Policy and Programs (founderspledge.com) (particularly the "why we recommend them" section)

Why isn't there a charity evaluator for longtermist projects?

CEvans9mo1

I think it could be bad if it relies too much on a particular worldview for its conclusions, which causes people to unnecessarily anchor on it. Seems like it could also be bad from a certain perspective if you think that it could lead to preferred treatment for longtermist causes which are easier to evaluate (eg. climate change relative to AI safety).

Future technological progress does NOT correlate with methods that involve less suffering

CEvans9mo1

Nice post - I think I agree that Ben's argument isn't particularly sound.

Are you thinking about this primarily in terms of actions that autonomous advanced AI systems will take for the sake of optimisation? If not, I imagine you could look at this with a different lense and consider one historical perspective which says something like "One large driver of humanity's moral circle expansion/moral improvement has been technological progress which has reduced resource competition and allowed groups to expand concern for others' suffering without undermin... (read more)

Jim Buhler

9mo

Thanks! Hum... not sure. I feel like my claims are very weak and true even in future worlds without autonomous advanced AIs. Agreed but this is more similar to argument (A) fleshed out in this footnote, which is not the one I'm assailing in this post.

Why isn't there a charity evaluator for longtermist projects?

CEvans9mo2

From a brief glance, it does appear that Founders Pledge's work is far more analogous to typical longtermist EA grantmaking than Givewell. Ie. it relies primarily on heuristics like organiser track record and higher-level reasoning about plans.

Tom Barnes

9mo

I think this is mostly correct, with the caveat that we don't exclusively rely on qualitative factors and subjective judgement alone. The way I'd frame it is more as a spectrum between [Heuristics] <------> [GiveWell-style cost-effectiveness modelling] I think I'd place FP's longtermist evaluation methodology somewhere between those two poles, with flexibility based on what's feasible in each cause

Why isn't there a charity evaluator for longtermist projects?

CEvans9mo1

Thanks for the comment Jeff! I admit that I didn't have biosecurity consciously in mind where I think perhaps you have an unusually clear paradigm compared to other longtermist work (eg. AI alignment/governance, space governance etc), and my statement was likely too strong besides.

However, I think there is a clear difference between what you describe and the types of feedback in eg. global health. In your case, you are acting with multiple layers of proxies for what you care about, which is very different to measuring the number of lives saved by AMF... (read more)

Why isn't there a charity evaluator for longtermist projects?

CEvans9mo2

Thanks for the detailed response! Your examples were helpful to illustrate your general thinking, and I did update slightly towards thinking some version of this could work, but I am still getting stuck on a few points:

Re. the GHD comparison: firstly to clarify, I meant "quality of reasoning" primarily in terms of the stated theory of change rather than a much more difficult to assess general statement. I would expect the quality of reasoning around a ToC to quite strongly correlate with expected impact. Of course this might not always cash out in actual i... (read more)

Why isn't there a charity evaluator for longtermist projects?

CEvans9mo2

Whats the version/route to value of this that you are excited about? I feel quite skeptical anything like this could work (see my answer on this post) but would be eager for people to change my mind.

Thomas Larsen9mo17

I agree that the lack of feedback loops and complicated nature of all of the things is a core challenge in making this useful.

I think you really can't do better than trying to evaluate people's track records and the quality of their higher level reasoning, which is essentially the meaning of grantmakers' statements like "just trust us".

I do have this sense that we can do better than illegible "just trust us". For example, in the GHD regime, it seems to me like the quality of the reasoning associated with people championing different interventions cou... (read more)

Why isn't there a charity evaluator for longtermist projects?

Answer by CEvansJul 29, 202324

I am surprised no one has directly made the obvious point of there being no concrete feedback loops in longtermist work, which means that it would be very messy to try and compare. While some people have tried to get at the cost effectiveness of X-risk reduction, it is essentially impossible to be objective in evaluating how much a given charity has actually reduced X-risk. Perhaps there is something about creating clear proxies which allows for better comparison, but I am guessing that there would still be major disagreements over what could be best... (read more)

Jeff Kaufman

9mo

This seems much too strong. Sure, "successfully avert human extinction" doesn't work as a feedback loop, but projects have earlier steps. And the areas in which I expect technical work on existential risk reduction to be most successful are ones where those loops are solid, and are well connected to reducing risk. For example, I work in biosecurity at the NAO, with an overall goal of some thing like "identify future pandemics earlier". Some concrete questions that would give good feedback loops: * If some fraction of people have a given virus, how much do we expect to see in various kinds of sequencing data? * How well can we identify existing pathogens in sequencing data? * Can we identify novel pathogens? If we don't use pathogen specific data can we successfully re-identify known pathogens? * What are the best methods for preparing samples for sequencing to get a high concentration of human viruses relative to other things? Similarly, consider the kinds of questions Max discusses in his recent far-UVC post.

(Even) More Early-Career EAs Should Try AI Safety Technical Research

CEvans9mo3

I think this post is interesting, while being quite unsure what my actual take is on the correctness of this updated version. I think I am worried about community epistemics in this world where we encourage people to defer on what the most important thing is.

It seems like there are a bunch of other plausible candidates for where the best marginal value add is even if you buy AI X- risk arguments eg. S risks, animal welfare, digital sentience, space governance etc. I am excited about most young EAs thinking about these issues for themselves.

How much do you... (read more)

tlevin

9mo

Yep, all sounds right to me re: not deferring too much and thinking through cause prioritization yourself, and then also that the portfolio is too broad, though these are kind of in tension. To answer your question, I'm not sure I update that much on having changed my mind, since I think if people did listen to me and do AISTR this would have been a better use of time even for a governance career than basically anything besides AI governance work (and of course there's a distribution within each of those categories for how useful a given project is; lots of technical projects would've been more useful than the median governance project).

AI Safety Europe Retreat 2023 Retrospective

CEvans1y4

This seems cool, thanks for running it!

What was the primary route to value of this retreat in your opinion? I'd be curious to know whether it was mainly about providing community and thus making participants more motivated, or if there were concrete new collaborations or significant individual updates derived from interactions at this retreat.

Magdalena Wache

I think the value comes from: * connections made * participants made 6.7 new connections on average (Question in the feedback form was "How many additional people working on AI safety would you feel comfortable asking for a small professional favor after this retreat? E.g. having a call to discuss research ideas, giving feedback on a draft, or other tasks that take around 30 minutes..."). * 18 people mentioned their plans to reach out to people from the retreat for feedback on posts or career plans, for research discussions or for collaborations. * motivation boost * projects/collaborations started * 9 people mentioned projects (Writing posts, doing experiments, and field building projects) they would start as a result of the retreat. * participants getting a better map of the space * which people exist, what kinds of things they are good at, what they have worked on, what you can ask them about. * What projects and what research exist. * A general sense of "getting more pieces of the puzzle". * I think this is hard to measure, but really valuable. I find it hard to say which of these is most important, and they are also highly entangled with each other

What AI Safety Materials Do ML Researchers Find Compelling?

CEvans1y4

Do you plan on doing any research into the cruxes of disagreement with ML researchers?

I realise that there is some information on this within the qualitative data you collected (which I will admit to not reading all 60 pages of), but it surprises me that this isn't more of a focus. From my incredibly quick scan (so apologies for any inaccurate conclusions) of the qualitative data, it seems like many of the ML researchers were familiar with basic thinking about safety but seemed to not buy it for reasons that didn't look ... (read more)

Vael Gates

I'm not going to comment too much here, but if you haven't seen my talk (“Researcher Perceptions of Current and Future AI” (first 48m; skip the Q&A) (Transcript)), I'd recommend it! Specifically, you want the timechunk 23m-48m in that talk, when I'm talking about the results of interviewing ~100 researchers about AI safety arguments. We're going to publish much more on this interview data within the next month or so, but the major results are there, which describes some AI researchers cruxes.

Sharmake

I don't see this as suspicious, because I suspect different goals are driving EAs compared to AI researchers. I'm not surprised by the fact that they disagree, since even if AI risk is high, if you have a selfish worldview, it's probably still rational to work on AI research.

University & City Group Outreach Testing - A Callout to Group Leaders

CEvans1y4

Hi Kynan thanks for writing this post.

It is great to see other people looking into more rigorous community building work! I really like the objective and methodology you set out, and do think that there are currently huge inefficiencies and loss in how information is currently transferred between groups.

I think one thing I am worried about with doing this on a large scale is the loss of qualitative nuance behind quantitative data. It seems difficult to really develop good models of why things work and what the key factors to consider are, without act... (read more)

david_reinstein

Thanks for being in touch (and I enjoyed our conversation). One thing to note is that some of the trials we are considering could be considered trials in 'what general paths and approaches to recommend', rather than narrowly-defined specific scripts. E.g., "reach out to a broad group of students" vs "focus on a small number of likely high-potential students." This could be operationalized, e.g., through which 'paths to involvement' (through fellowship completion or through attending meetings and events), or through 'which courses/majors to reach out to'. However, every university group could still have the flexibility to adopt the recommended guidelines in a manner that aligns with their unique culture and surroundings. We could then try to focus on some generally agreed 'aggregate outcome measures'. This could then be considered a test of 'which recommended general approach works better' (or, if we can have subgroup analysis, 'which works better where'.

Kynan Behan

Agreed on all points. An important consideration is heavily involving group leaders and organisers in this process to preserve the qualitative aspects of 'what works' in outreach. Keeping those involved with implementing the methods engaged throughout the research process is vital for ensuring these methods transfer into the real world. Whilst some of the nuances are inevitably going to be lost through large-scale testing, we can counteract this by knowing where to allow room for flexibility and where rigidity is worthwhile. I'll be in touch, thanks!

Needed: Input on testing fit for your career

CEvans3y3

Thanks for your comment. For your first point, I definitely agree in an ideal world that benchmarks for improvement would be useful but I would be hesitant for a few reasons.

Firstly, you face quite a risk of putting people off a certain career when really you don't have the certainty to give that advice (especially when I am not a specialist in the field), and that could be really damaging and maybe not that useful. Secondly, these things are generally really context specific for how good X amount of progress is in Y amount of time. Eg. for your exam... (read more)

Towards a longtermist framework for evaluating democracy-related interventions

CEvans3y13

I think this is really cool and a great way of breaking things down - thanks for writing this up!