I note Constantin’s post, first, was extraordinary uncharitable and inflammatory (e.g. the title for the section discussing Wiblin’s remark “Keeping promises as a symptom of Autism”, among many others); second, these errors were part of a deliberate strategy to ‘inflame people against EA’; third, this strategy is hypocritical given the authors (professed) objections to any hint of ‘exploitative communication’. Any of these in isolation is regrettable. In concert they are contemptible.
{ETA: Although in a followup post Constantin states her previous comments which were suggestive of bad faith were an “emotional outburst”, it did not reflect her actual intentions either at the time of writing or subsequently.}
My view is that, akin to Hostadter’s law, virtues of integrity are undervalued even when people try to account for undervaluing them: for this reason I advocate all-but lexical priority to candour, integrity, etc. over immediate benefits. The degree of priority these things should be accorded seems a topic on which reasonable people can disagree: I recommend Elmore’sremarks as a persuasive defence of according these virtues a lower weight.
‘Lower’, however, still means ‘quite a lot’: if I read Elmore correctly, her view is not one can sacrifice scrupulously honest communication for any non-trivial benefit, but that these norms should on occasion be relaxed if necessary to realise substantial gains. The great majority of EAs seem to view these things as extremely important, and the direction of travel appears to me that ‘more respected’ EAs tend to accord these even greater importance (see MacAskill; c.f. Tsipursky).
My impression is that EAs—both individually and corporately—do fairly well in practice as well as principle. As Naik notes, many orgs engage in acts of honesty and accountability supererogatory to secure funding. When they do err, they tend to be robustly challenged (often by other EAs), publicly admit their mistake, and change practice (all of the examples Constantin cites were challenged at the time, I also think of Harris’s concerns with promotion of EA global, GPP’s mistaken use of a Stern report statistic, and now ACE). Similar sentiments apply to an individual level: my ‘anecdata’ is almost the opposite of Fluttershy’s extremely bad experience: I sincerely believe (and even more sincerely hope) that mine is closer to the norm than theirs.
In absolute terms, I don’t think EA in toto has a ‘lying problem’ (or a ‘being misleading’, ‘not being scrupulously honest’ problem). It seems to do quite well at this, and the rate and severity of the mistakes I see don’t cause great alarm (although it can and should do better). Although relative terms are less relevant, I think it does better than virtually any other group I can think of.
I offer some further remarks on issues raised by some of the examples given which do not fit into the ‘lying problem’ theme:
1) Ironic, perhaps, that the best evidence for Todd’s remark on the ‘costs of criticism’ arise from the aftermath of a post which (in part) unjustly excoriates him for that particular remark. My impression is that bad criticism is on average much more costly than bad praise, and some asymmetry in how these are treated seem reasonable.
I do not know whether journalistic ‘best practice’ around ‘right of reply’ extends to providing the criticism in full to its subject—regardless, it seems good practice to adopt for the reasons Todd explains. I have done this with my co-contributors re. Intentional Insights, and I have run a (yet to be published) piece about MIRI by MIRI as it had some critical elements to it. Naturally, if a critic does not do this for whatever reason, it does not mean their criticism should be ignored (I have yet to see a case of criticism ‘shunned’ for these reasons) but I think this is a norm worth encouraging.
2) Nonetheless, it may not have been advisable for the head of one ‘part’ of CEA to bring this up in context of criticism addressed to another part of CEA. Issues around appropriate disclosure have been mentioned before. In addition, remarks by ‘EA public figures’ may be taken as indicative of the view of their organisations or EA in toto even if explicitly disclaimed as ‘personal opinion only’. A regrettable corollary (as Gordon-Brown notes) is a chilling effect on ‘EA public figures’ refraining from making unguarded remarks publicly. The costs of not doing so may be worse: if EA grows further, we may collectively regret ‘providing more ammunition’ to external critics to misuse.
3) Given the social costs towards an individual critic, there may be benefit (by organisations or, better, an independent ‘Grand Inquisition’ collaboration) to canvass these anonymously. The commonest shared could then be explored further: this would be valuable whether they point to a common misconception or a common fault. In the meanwhile, anyone is welcome to disclose criticisms or concerns to me in confidence.
4) Certain practices could be more widely adopted by EA orgs—beyond recording predictions, a prominent ‘mistakes’ page (per Givewell) would be desirable, likewise scrupulous declaration of relevant conflicts of interests.
5) (I owe this to Carl Shulman). Donors could also pitch in by carefully evaluating empirical or normative claims made by particular EA organisations: Plant, Dickens, and Hoffman would all be laudable examples, and I hope both to contribute some of my own work to this genre and to encourage others to do likewise.
I note Constantin’s post, first, was extraordinary uncharitable and inflammatory (e.g. the title for the section discussing Wiblin’s remark “Keeping promises as a symptom of Autism”, among many others); second, these errors were part of a deliberate strategy to ‘inflame people against EA’; third, this strategy is hypocritical given the authors (professed) objections to any hint of ‘exploitative communication’. Any of these in isolation is regrettable. In concert they are contemptible.
{ETA: Although in a followup post Constantin states her previous comments which were suggestive of bad faith were an “emotional outburst”, it did not reflect her actual intentions either at the time of writing or subsequently.}
My view is that, akin to Hostadter’s law, virtues of integrity are undervalued even when people try to account for undervaluing them: for this reason I advocate all-but lexical priority to candour, integrity, etc. over immediate benefits. The degree of priority these things should be accorded seems a topic on which reasonable people can disagree: I recommend Elmore’s remarks as a persuasive defence of according these virtues a lower weight.
‘Lower’, however, still means ‘quite a lot’: if I read Elmore correctly, her view is not one can sacrifice scrupulously honest communication for any non-trivial benefit, but that these norms should on occasion be relaxed if necessary to realise substantial gains. The great majority of EAs seem to view these things as extremely important, and the direction of travel appears to me that ‘more respected’ EAs tend to accord these even greater importance (see MacAskill; c.f. Tsipursky).
My impression is that EAs—both individually and corporately—do fairly well in practice as well as principle. As Naik notes, many orgs engage in acts of honesty and accountability supererogatory to secure funding. When they do err, they tend to be robustly challenged (often by other EAs), publicly admit their mistake, and change practice (all of the examples Constantin cites were challenged at the time, I also think of Harris’s concerns with promotion of EA global, GPP’s mistaken use of a Stern report statistic, and now ACE). Similar sentiments apply to an individual level: my ‘anecdata’ is almost the opposite of Fluttershy’s extremely bad experience: I sincerely believe (and even more sincerely hope) that mine is closer to the norm than theirs.
In absolute terms, I don’t think EA in toto has a ‘lying problem’ (or a ‘being misleading’, ‘not being scrupulously honest’ problem). It seems to do quite well at this, and the rate and severity of the mistakes I see don’t cause great alarm (although it can and should do better). Although relative terms are less relevant, I think it does better than virtually any other group I can think of.
I offer some further remarks on issues raised by some of the examples given which do not fit into the ‘lying problem’ theme:
1) Ironic, perhaps, that the best evidence for Todd’s remark on the ‘costs of criticism’ arise from the aftermath of a post which (in part) unjustly excoriates him for that particular remark. My impression is that bad criticism is on average much more costly than bad praise, and some asymmetry in how these are treated seem reasonable.
I do not know whether journalistic ‘best practice’ around ‘right of reply’ extends to providing the criticism in full to its subject—regardless, it seems good practice to adopt for the reasons Todd explains. I have done this with my co-contributors re. Intentional Insights, and I have run a (yet to be published) piece about MIRI by MIRI as it had some critical elements to it. Naturally, if a critic does not do this for whatever reason, it does not mean their criticism should be ignored (I have yet to see a case of criticism ‘shunned’ for these reasons) but I think this is a norm worth encouraging.
2) Nonetheless, it may not have been advisable for the head of one ‘part’ of CEA to bring this up in context of criticism addressed to another part of CEA. Issues around appropriate disclosure have been mentioned before. In addition, remarks by ‘EA public figures’ may be taken as indicative of the view of their organisations or EA in toto even if explicitly disclaimed as ‘personal opinion only’. A regrettable corollary (as Gordon-Brown notes) is a chilling effect on ‘EA public figures’ refraining from making unguarded remarks publicly. The costs of not doing so may be worse: if EA grows further, we may collectively regret ‘providing more ammunition’ to external critics to misuse.
3) Given the social costs towards an individual critic, there may be benefit (by organisations or, better, an independent ‘Grand Inquisition’ collaboration) to canvass these anonymously. The commonest shared could then be explored further: this would be valuable whether they point to a common misconception or a common fault. In the meanwhile, anyone is welcome to disclose criticisms or concerns to me in confidence.
4) Certain practices could be more widely adopted by EA orgs—beyond recording predictions, a prominent ‘mistakes’ page (per Givewell) would be desirable, likewise scrupulous declaration of relevant conflicts of interests.
5) (I owe this to Carl Shulman). Donors could also pitch in by carefully evaluating empirical or normative claims made by particular EA organisations: Plant, Dickens, and Hoffman would all be laudable examples, and I hope both to contribute some of my own work to this genre and to encourage others to do likewise.