3 ms·
This "new" method seems to be an old method, indistinguishable from Pearl's do-calculus. What is the innovation here, aside from a new software library? The qu
by uoaei 4y ago
This "new" method seems to be an old method, indistinguishable from Pearl's do-calculus.
What is the innovation here, aside from a new software library? The quantification of each candidate root cause's influence on the outcome? I am surprised the authors found nothing similar throughout the entire corpus of academic research.
I must say, Amazon's "science" blog is the most unimpressive of the big tech companies. It churns out PR like the rest, but the others at least have some substance behind them.
- tpoacher 4y agoIt's ironic that they say they find no good definition for the term, but then go on to cite the same people that use those exact terms to describe the exact same thing: https://www.semanticscholar.org/paper/Causality%2C-Responsibility%2C-and-Blame%3A-A-Approach-Halpern/212ee07c3b76f0fd0ec603584d9f6034cf77b6b9 https://www.semanticscholar.org/paper/Causality%2C-Responsib...
- svnt 4y agoIt would be more ironic if they made a mistake like this in a way that wasn’t quite so ego-reinforcing.
- just_boost_it 4y agoEvery one that I see is a pretty weak summary of a decades old idea, but presented like its new research.