# cousin_it comments on Ingredients of Timeless Decision Theory

• This is bet­ter than noth­ing, thanks and up­vote. Now let’s be­gin trans­lat­ing this stuff. AFAICT, a “de­ci­sion the­ory” is sup­posed to have two parts:

1) A blah blah ver­bal al­gorithm for trans­lat­ing real-world prob­lem de­scrip­tions into a cer­tain kind of for­mal struc­ture.

2) A math­e­mat­i­cal al­gorithm that ac­cepts that for­mal struc­ture and out­puts a de­ci­sion.

I don’t fully un­der­stand what for­mal struc­ture you’re propos­ing (a Pearl-style causal graph with ad­di­tional “log­i­cal” ar­rows? why would this always be acyclic?), and can’t un­der­stand the al­gorithm un­til the struc­ture is clear enough.

• why would this always be acyclic?

If the ar­rows are ma­te­rial im­pli­ca­tions, then A → B → C → A col­lapses via iff to a sin­gle node. Can you give an ex­am­ple of cyclic log­i­cal un­cer­tainty?

• I was think­ing of some case where the cy­cle con­tains both phys­i­cal and log­i­cal ar­rows. Log­i­cal ar­rows can point back­wards in time, so this doesn’t seem to be im­pos­si­ble in prin­ci­ple. Sorry, can’t give a spe­cific ex­am­ple be­cause I don’t fully un­der­stand what you mean by “log­i­cal un­cer­tainty”.

• My read­ing is that log­i­cal nodes can point to phys­i­cal nodes, but not vice versa. (Also that it doesn’t make sense to say an ar­row from a log­i­cal node “points back­wards in time”. Log­i­cal nodes are time­less.)

• Phys­i­cal ar­rows shouldn’t point to log­i­cal nodes, though… right?