Keynote: The Mathematics of Causal Inference: with Reflections on Machine Learning

The development of graphical models and the logic of counterfactuals have had a marked effect on the way scientists treat problems involving cause-effect relationships. Practical problems requiring causal information, which long were regarded as either metaphysical or unmanageable can now be solved using elementary mathematics. Moreover, problems that were thought to be purely statistical, are beginning to benefit from analyzing their causal roots.

[@sphereron]

Slides: simons.berkeley.edu/sites/default/files/docs/422/pearljudea.pdf

[@EustaquioSantimano]

Thank you. It would help if we saw more of the slides when Judea is talking.

[@saiananth8751]

This lecture here is a bloody masterpiece

[@JohnWilliamsFromBluff]

I have a postgraduate degree in statistics, and the more I learn about this material the more I'm impressed with Judea Pearl and his colleagues. Prof. Pearl is also a very impressive human being, even if you disregard his acheivements described here. Look him up: he's awesome.

[@prub4146]

Causal Inference in Statistics: A Primer brought me here

[@kevalan1042]

360p? WHY?

[@mire1ac]

This comment probably made me laugh a bit too much.

[@JohnWilliamsFromBluff]

Because it's from Microsoft, the company that treats the end-user with disdain, Like most companies, to be fair; they're just more blatant.

[@galenseilis5971]

So that someone can come along to practice using super-resolution imaging. ;)

[@sleepingevenbetter]

Who ever put this together, the differences between the slides in some cases are (or at least appear to be) subtle... next time please show the transitions.

[@TeaParty1776]

See _Leap Of Logic_ by physicist David Harriman for induction.

[@jaydevbaba]

You do not need models for causality. If he mentions models like the economics example he gives. Models are helpful in predicting things and not that great on inference. I would recommend looking into Rubin's talk for that aspect. Interesting talk though but that what he proposes is not so strong like what Rubin proposes.. so ..i think under many practical constraints one can use Rubin's approach , if you are using prediction to imply causality it would be like a simulation and emulation approach, you really need to be sure of what you have simulated and emulated. All said and done, am glad he gave this talk and people are concerned about causality.

[@khwajawisal1220]

I hope you understand what you just said.