Model-free and model-based cognition in deontological and consequentialist reasoning

Link post

Mo­ral dilem­mas en­gen­der con­flicts be­tween two tra­di­tions: con­se­quen­tial­ism, which eval­u­ates ac­tions based on their out­comes, and de­on­tol­ogy, which eval­u­ates ac­tions them­selves. Th­ese strik­ingly re­sem­ble two dis­tinct de­ci­sion-mak­ing ar­chi­tec­tures: a model-based sys­tem that se­lects ac­tions based on in­fer­ences about their con­se­quences; and a model-free sys­tem that se­lects ac­tions based on their re­in­force­ment his­tory. Here, I con­sider how these sys­tems, along with a Pavlo­vian sys­tem that re­sponds re­flex­ively to re­wards and pun­ish­ments, can illu­mi­nate puz­zles in moral psy­chol­ogy.

I find this (short) pa­per ap­peal­ing be­cause it:

  • Makes con­nec­tions across domains

  • Unifies com­pet­ing ex­pla­na­tions within an over­ar­ch­ing framework

  • Offers an el­e­gant ex­pla­na­tion of oth­er­wise con­fus­ing phe­nom­ena e.g. the means/​side-effect distinction

That said, it does still seem a bit spec­u­la­tive.