I’ve got a seemingly obvious flaw to point out; in fact, it appears so obvious to me that I would be surprised if it hadn’t been addressed in the original post or one of the subsequent comments and I simply skipped over it. Nonetheless, it may be of use.
I feel that the whole experiment is rather undermined by selection bias. I think its a fair assumption that you would want this method tried elsewhere were the experiment successful, you would want “Dragon Houses” to pop up anywhere where there is a sufficient rationalist community. However, it would be an error to think that the dragon house model would work elsewhere once you aren’t picking out the people most suitable for living there, and have to expand to more general types. Again, I feel it is a fair assumption that some people will simply be a lot more suited to authoritarian communities, such as the Army, than others. If you can pre-approve for authoritarian types, and eject those who don’t fit as you identify them, then it seems far more likely that the community will survive, but it could still be inferior to another model that does not select so heavily.
Is this merely a proof of concept? i.e., you will run the dragon house for a short period of time under perfect conditions to ensure that it is not a complete disaster, and does not result in the ‘toxic cult’ dangers that others have outlined, before expanding the experiment? In which case the selection bias would be removed in the second run and you could ascertain the general effectiveness of the model.
Apologies for the poor construction of the above, I struggled somewhat to put it into words but I hope you can comprehend regardless
I’m not, in fact, looking to build an exportable model. I’d like to export pieces of a model—specific activities that worked, norms that are high-impact, cool insights and so forth. But I don’t particularly want or expect other Dragon Army houses to pop up elsewhere.
It’s not a proof-of-concept so much as something that I, myself, want to have access to/experience/be a part of. If I can make it work even just this one time, in this one place, that’s enough.
I’ve got a seemingly obvious flaw to point out; in fact, it appears so obvious to me that I would be surprised if it hadn’t been addressed in the original post or one of the subsequent comments and I simply skipped over it. Nonetheless, it may be of use.
I feel that the whole experiment is rather undermined by selection bias. I think its a fair assumption that you would want this method tried elsewhere were the experiment successful, you would want “Dragon Houses” to pop up anywhere where there is a sufficient rationalist community. However, it would be an error to think that the dragon house model would work elsewhere once you aren’t picking out the people most suitable for living there, and have to expand to more general types. Again, I feel it is a fair assumption that some people will simply be a lot more suited to authoritarian communities, such as the Army, than others. If you can pre-approve for authoritarian types, and eject those who don’t fit as you identify them, then it seems far more likely that the community will survive, but it could still be inferior to another model that does not select so heavily.
Is this merely a proof of concept? i.e., you will run the dragon house for a short period of time under perfect conditions to ensure that it is not a complete disaster, and does not result in the ‘toxic cult’ dangers that others have outlined, before expanding the experiment? In which case the selection bias would be removed in the second run and you could ascertain the general effectiveness of the model.
Apologies for the poor construction of the above, I struggled somewhat to put it into words but I hope you can comprehend regardless
No, this is a valid point.
I’m not, in fact, looking to build an exportable model. I’d like to export pieces of a model—specific activities that worked, norms that are high-impact, cool insights and so forth. But I don’t particularly want or expect other Dragon Army houses to pop up elsewhere.
It’s not a proof-of-concept so much as something that I, myself, want to have access to/experience/be a part of. If I can make it work even just this one time, in this one place, that’s enough.