This is a good article and a valuable discussion to have. I have a couple of nitpicks on the discussion of theoretical frameworks that tend to be ignored by EA. You mentioned the following examples:
Sociological theory: potentially relevant to understanding causes of global poverty, how group dynamics operates and how social change occurs.
Ethnography: potentially highly useful in understanding causes of poverty, efficacy of interventions, how people make dietary choices regarding meat eating, the development of cultural norms in government or research organisations surrounding safety of new technologies, and other such questions, yet I have never heard of an EA organisation conducting this sort of analysis.
Phenomenology and existentialism: potentially relevant to determining the value of different types of life and what sort of society we should focus on creating.
Historical case studies: there is some use of these in the study of existential risk, mostly relating to nuclear war, but mostly this method is ignored as a potential source of information about social movements, improving society, and assessing the risk of catastrophic risks.
Regression analysis: potentially highly useful for analysing effective causes in global development, methods of political reform, or even the ability to influence AI or nuclear policy formation, but largely neglected in favour of either experiments or abstract theorising.
I have never seen any sociological analysis in EA and agree that it’s been (almost?) completely ignored.
Ethnographies have been almost completely absent from EA from all of its history, with the exception of a recent small increase in interest. A recent upvoted, positively received, and discussed post on this forum expressed interest in an ethnography of EA. A comment on that post mentioned a couple of ethnographies of EAs, including this one of London EA.
Phenomenology and existentialism: I’m not sure what this means. EA has spent a fair amount of time thinking about the value of different types of biological and synthetic lifeforms (e.g. wildlife suffering, suffering in AI). The second example seems a bit underdefined to me. I’m not familiar with phenomenology and existentialism and might be misunderstanding this section.
For historical case studies, I think “mostly ignored” is misleading. A more accurate description might be that they’re underutilized relative to their full potential and relative to other frameworks, but they’re taken seriously when they come up.
As you mention, historical case studies have been used in x-risk analysis. The Open Philanthropy Project has also commissioned and written about several historical case studies , going back to its GiveWell Labs days. Their page on the History of Philanthropy says:
We’ve found surprisingly little existing literature on the history of philanthropy. In particular, we’ve found few in-depth case studies examining questions like what role philanthropists, compared with other actors, played in bringing important changes to pass. To help fill that gap, we are commissioning case studies on past philanthropic success stories, with a focus on cases that seem — at first glance — to be strong examples of philanthropy having a major impact on society.
They go on to list 10+ case studies (only one focusing on x-risks), including some from a $165k grant to the Urban Institute specifically for the purpose of producing more case studies. Of course $165k is a small amount for OpenPhil, but it seems to me, for a few reasons, that they take this work seriously.
The Sentience Institute has published 6 reports, 3 of which are historical case studies. Historical case studies relating to nuclear war, like the Petrov and Arkhipov incidents, have been widely discussed in EA as well. The Future of Life Institute has published some material relating to this.
Regression analysis: I find this example puzzling. Regressions are widely used in development economics, which has heavily influenced EA thinking on global health. EAs who are professional economists or otherwise have reason to use regressions do so when appropriate (e.g. Rachel Glennester and J-PAL, some of Eva Vivalt’s work, etc). GiveWell’s recommendation of deworming charities is largely dependent on the regression estimates of a couple of studies.
More generally, regressions are a subset of statistical analysis techniques. I’m not sure if EA can be credibly accused of ignoring statistical analysis. I also don’t think the other examples you gave of uses of regressions (political reform, AI and nuclear policy) are a great fit for regression analysis or statistical analysis in general.
This is a good article and a valuable discussion to have. I have a couple of nitpicks on the discussion of theoretical frameworks that tend to be ignored by EA. You mentioned the following examples:
I have never seen any sociological analysis in EA and agree that it’s been (almost?) completely ignored.
Ethnographies have been almost completely absent from EA from all of its history, with the exception of a recent small increase in interest. A recent upvoted, positively received, and discussed post on this forum expressed interest in an ethnography of EA. A comment on that post mentioned a couple of ethnographies of EAs, including this one of London EA.
Phenomenology and existentialism: I’m not sure what this means. EA has spent a fair amount of time thinking about the value of different types of biological and synthetic lifeforms (e.g. wildlife suffering, suffering in AI). The second example seems a bit underdefined to me. I’m not familiar with phenomenology and existentialism and might be misunderstanding this section.
For historical case studies, I think “mostly ignored” is misleading. A more accurate description might be that they’re underutilized relative to their full potential and relative to other frameworks, but they’re taken seriously when they come up.
As you mention, historical case studies have been used in x-risk analysis. The Open Philanthropy Project has also commissioned and written about several historical case studies , going back to its GiveWell Labs days. Their page on the History of Philanthropy says:
They go on to list 10+ case studies (only one focusing on x-risks), including some from a $165k grant to the Urban Institute specifically for the purpose of producing more case studies. Of course $165k is a small amount for OpenPhil, but it seems to me, for a few reasons, that they take this work seriously.
The Sentience Institute has published 6 reports, 3 of which are historical case studies. Historical case studies relating to nuclear war, like the Petrov and Arkhipov incidents, have been widely discussed in EA as well. The Future of Life Institute has published some material relating to this.
Regression analysis: I find this example puzzling. Regressions are widely used in development economics, which has heavily influenced EA thinking on global health. EAs who are professional economists or otherwise have reason to use regressions do so when appropriate (e.g. Rachel Glennester and J-PAL, some of Eva Vivalt’s work, etc). GiveWell’s recommendation of deworming charities is largely dependent on the regression estimates of a couple of studies.
More generally, regressions are a subset of statistical analysis techniques. I’m not sure if EA can be credibly accused of ignoring statistical analysis. I also don’t think the other examples you gave of uses of regressions (political reform, AI and nuclear policy) are a great fit for regression analysis or statistical analysis in general.