Ineffective Response to COVID-19 and Risk Compensation

UPDATE: This post was writ­ten and re­flects an ear­lier set of my be­liefs. I have up­dated sig­nifi­cantly in a num­ber of ways since it was posted, based on both ex­ter­nal events and re­search, and no longer en­dorse it.

Epistemic sta­tus: I have a men­tal model that I think sep­a­rates my view of re­sponse ac­tivi­ties from that of the ma­jor­ity of what I see on Less­wrong and as­so­ci­ated places. If it is in­cor­rect, I’d be happy to up­date, but I think this is an area I have con­sid­ered more than most other posters. I want to write a short post ex­plain­ing this to al­low oth­ers to up­date, and see­ing if some­one has an ar­gu­ment that changes my mind.

Put sim­ply, my claim is that bring­ing at­ten­tion to likely in­effec­tive per­sonal meth­ods for re­duc­ing risk is not net neu­tral with a large up­side if they work, it is in­stead likely to be on net fairly harm­ful, albeit with a large up­side if they work.

Argument

First, we have in­cred­ibly effec­tive and vastly un­der­uti­lized ways to pre­vent spread of COVID-19, namely hand­wash­ing and not touch­ing your face. Given that, if I pro­pose an in­ter­ven­tion like mak­ing home­made masks from fabric which re­duced hand­wash­ing com­pli­ance by 1% (per­haps due to dis­tract­ing peo­ple or mak­ing them think hand­wash­ing is less crit­i­cal,) it would need to be as­ton­ish­ingly effec­tive to be net pos­i­tive. And most such ap­proaches be­ing dis­cussed are, as far as I can tell, nowhere near that level of effec­tive­ness.

Se­cond, most read­ers of Less­wrong and effec­tive al­tru­ism blogs and face­book groups aren’t hard­core ra­tio­nal­ists, and even hard­core ra­tio­nal­ists aren’t im­mune to Akra­sia. On top of that, peo­ple like Scott Alexan­der have huge read­er­ships and some­times link peo­ple to Less­wrong. Many peo­ple read­ing posts here aren’t wash­ing their hands enough as it is, and aren’t go­ing to ra­tio­nally eval­u­ate the rel­a­tive effec­tive­ness of hand­wash­ing ver­sus other in­ter­ven­tions.

Third, ev­i­dence ex­ists that risk-com­pen­sa­tion is a mean­ingful is­sue. Ac­tions that make peo­ple feel safer usu­ally lead to less at­ten­tion paid to more an­noy­ing /​ more in­tru­sive mea­sures. (There is ev­i­dence, such as Vrolix’s pa­per*, that risk com­pen­sa­tion re­duces the size of the pos­i­tive im­pact, but does not make in­ter­ven­tions net nega­tive. This is con­di­tioned on the im­pact be­ing sig­nifi­cant and pos­i­tive, how­ever, and seems not to ap­ply to spec­u­la­tive in­ter­ven­tions like those be­ing pro­posed.

This is not an ar­gu­ment that we should not look into bet­ter op­tions for re­sponse. It’s an ar­gu­ment that we should be more care­ful in vet­ting them be­fore en­courag­ing peo­ple to do them just in case they work.

*) Vrolix, Klara (2006). “Be­havi­oural Adap­ta­tion, Risk Com­pen­sa­tion, Risk Homeosta­sis and Mo­ral Hazard in Traf­fic Safety” )