TLDR:

I believe that distribution of current EA efforts is sub-optimal and should focus on generalist resilience efforts to a far greater extent. (anywhere between 5-50% of total efforts by upscaling current efforts x1000-x100.000)

 

Premises:

1 EA is built on the principles of evidence and reason when coming to conclusions about importance of cause areas and projects for the long-term future of humanity

2 rationality/EA tools such as forecasting and cause prioritization are good tools to come to the classic EA conclusions about 'High Impact Problems': the global problems and events that are (1) most probable to (2) cause the biggest magnitude of suffering/deaths/DALY/whatever out of all known problems.

3 Practically all EA efforts are currently built upon these 'classic EA conclusions'.

Disclaimer:

4 I generally agree with these conclusion given the expected impact, neglectedness and tractability of these well-formulated cause areas.

I generally agree that it is a good thing that EA aims to increase the amount of attention and work done on these specific issues with the goal of upscaling countermeasures by several orders of magnitude .

However:

5 I find it likely that there is a set of global problems / events causing expected suffering (based on 1 probability & 2 magnitude) similar to currently known High Impact Problems, that are currently unknown.

6 I find it likely that some of these unknown High Impact Problems can be 'discovered', arguably that's what happened with AI safety. Let's say discovery means that a problem becomes obvious and looming enough for someone to write something like a cause area profile on it.

7 I find it likely that some of these unknown High Impact Problems can not be 'discovered' until it is 'too late' for EA to start countering them - even by attempts of systematically finding any such possible blind spots or black swans. Let's say too late means that it is too late for preventive countermeasures leading to either: 
1 extinction or 2 mitigation that can not avert a large portion of the harm done

(more or less probable examples of such events/problems could be e.g. discovery of a 'black ball technology', an unforeseen strongly destabilizing political event, discovery of malevolent multidimensional lifeforms, EA becoming the focus of a targeted attack, let your imagination run free for more examples)

8  In these cases, generalist resilience building efforts are the only way to effectively prevent or mitigate unknown High Impact Problems occurring.

(examples of such efforts could be e.g. research/advocacy for increased societal resilience, research/advocacy for improved institutional decision making or university/career advice focusing more on general leverage rather than specialist cause area specific courses (think ai safety researcher)

Controversial conclusions:

9 Tools like forecasting or cause prioritization and their 'classic EA conclusions' do not give a comprehensive picture of all High Impact Problems and likely never will. 

10 Practically all EA efforts are currently built upon these 'classic EA conclusions'* and therefore likely sub-optimally distributed.

11 EA efforts should focus to a greater extent on general countermeasures to all possible High Impact Problems (known/unknown) 

12 My naive intuition for a desirable 'greater extent' would be:
Focusing a significant part of total EA efforts on generalist resilience building, say anywhere between 5-50%, maybe more, by upscaling efforts by x1000-x100.000, maybe more.

 

Caveats and further thoughts:

Arguably, generalist resilience efforts are not neglected as these topics are traditional politicians', economists', etc. daily bread-and-butter. 

It is likely hard, maybe impossible to find the ideal split of resource allocation to specialist vs generalist efforts as it could be impossible to quantify the expected negative impact of black swan type unknown High Impact Problems.

I realize that my intuitions about ideal allocation implicitly assume a certain prior probability of unknown High Impact Events occurring and that intuition about this probability might differ from person to person.
The classic book 'Black Swan' by Nassim Taleb is stuffed with historic examples of unknown High Impact Events and attempts to quantify risks overlooking them; My intuitions have probably largely been influenced by the author. 

Arguably, many generalist efforts generalize well to specific classic cause areas which would further increase the ideal proportion of resources going into generalist resilience building. 

 

__________________________________________________

*Keep in mind that even grants whose purpose is to identify thus far unknown High Impact Problems will only come up with new 'classic EA conclusions' and not give a comprehensive picture of all High Impact Problems due to 7.

15

6 comments, sorted by Click to highlight new comments since: Today at 9:57 PM
New Comment

Thought-provoking! I think one organization that illustrates the split you're talking about well is the Centre for Long Term Resilience. This is an EA-aligned advocacy group focused on increasing awareness of and capacity to respond to extreme risks in the UK government. What's interesting with regard to your post is that they've divided their work into three divisions. The first two, AI and biosecurity, are issue-specific. But the third division focuses on generalized "risk management," or as they put it, "the process of both transforming risk governance, and of identifying, assessing and mitigating all extreme risks."

I think I agree with you that the community could benefit from more investment in cross-issue and exploratory work of this type. The rest of the world is already too siloed and EA set itself apart in the first place with its commitment to cause neutrality. It's important to retain that openness and integrative approach even as we get deeper into implementation work on high-impact causes.

Thanks for your comment!
Fascinating, haven't heard of them thus far, I will look into their work!
Yes, I agree. I believe that past successes with cause prioritization and forecasting could lead to some overconfidence in regard to competence at discovering unknown cause areas: It will be important to keep up the cause-neutrality, especially in regards to the unknown!

Thank you for your post. I agree that "general resilience-building" would in any case be valuable.  I want to draw parallels to biology. In biology, they call the capacity of a species to weather crises "robustness". I feel that "robustness" is an apt term to describe what your post is aiming at. Another useful concept is the idea of "evolutionary capacitance" (i.e., being able to diversify swiftly after a catastrophe, which then generates many "templates" on which natural selection can then act on). While not wanting to get lost in semantics,  we can use these terms to peruse the literature in order to look at strategies and principles for how nature was able to achieve resilience over millions of years. We could then  derive practical principles&strategies&tactics&tools for increasing the resilience of humanity

I agree that resilience is overall relatively neglected in the GCR community. Another organization working on this is the Alliance to Feed the Earth in Disasters (ALLFED). Also, you would probably be interested in this paper: Defence in Depth Against Human Extinction: Prevention, Response, Resilience, and Why They All Matter.

Thanks for the write up. 

To add some more suggestions to how generalist resilience may come about, check this out by 80k: https://80000hours.org/problem-profiles/ 

I'm currently unsure of what to think about your claim that generalist resilience needs to be upscaled by that large factor. This is mainly due to being unsure about what proportion of efforts is currently being directed there. Do you know anything about the distribution of efforts already at generalist resilience?

Thanks for your comment! Yes I am not sure about this factor myself, it is not based on a comprehensive review of all EA efforts on global resilience relative to total efforts. 

Checking OpenPhil's Grant Database just now was not as conclusive as I'd hoped as grant categories such as 'Global Catastrophic Risks' or 'Global Health & Development' leave open whether these are spent on generalist resilience or specialist efforts.