@jillnephew 2. I worry that the focus on mechanism is a kind of map-territory conflation. A mechanism is our working model of a system, but the map is not the territory - for example, different causal mechanisms can have the same statistical fingerprint h
@ThosVarley @_fernando_rosas @erikphoel That's a key step which distinguishes the interventional/causal approach, since it removes correlations between the source and target past. Whether that's good or bad compared to TE depends on what you want it to mea
@AMTraderX @rasmansa This gets into some deep questions about information and causality - for ex. just MI is probably too limited and we should be looked at something like the transfer entropy. But even that isn't a true measure of causation. See: https://
@olivercliff @okularab @PhysRevResearch @novelli_leo @bendfulcher @jmacshine Yep, we defs don't like the use of causality for that. Sticking with it as an established name but noting the distinction in the paper - pointing to our earlier work differentiati