1.NeurIPS 2018 Tutorial on Counterfactual Inference05:08
2.Introduction of Speaker: Susan Athey05:13
3.Counterfactual Inference06:38
4.Artificial Intelligence/Machine Learning Desired Properties for Applications10:53
5.Artificial Intelligence and Counterfactual Estimation25:56
6.Counterfactual Inference Approaches29:52
7.Program Evaluation, Treatment Effect Estimation31:08
8.Structural Estimation, Generative Models & Counterfactuals40:22
9.Causal Discovery, Learning the Causal Graph45:56
10.Recently, Literatures Have Started Coming Together47:40
11.Preview of Themes48:11
12.Estimating ATE under Unconfoundedness55:28
13.Idea56:13
14.Example: Effect of an Online Ad57:54
15.Setup58:25
16.Intuition for Most Popular Methods1:00:33
17.BREAK1:05:33
18.Intuition for Most Popular Methods (Continuation)1:16:01
19.Setup1:17:34
20.Using Supervised ML to Estimate ATE Under Unconfoundedness - Method I: Propensit...1:20:44
21.Method II: Regression adjustment1:21:53
22.Method III: Estimate CATE and take averages1:25:06
23.Method IV: Double robust/double machine learning1:26:20
24.Method V: Residual Balancing1:31:00
25.What if unconfoundedness fails?1:31:37
26.Local Average Treatment Effects1:34:35
27.IV Approaches: Including Covariates1:35:58
28.User Model of Clicks:
Results from Historical Experiments
(Athey, 2010)1:37:37
29.IV: Heterogeneous Treatment Effects1:39:58
30.Instrumental Variables (IV)1:40:30
31.Search Ads Application of Deep IV: Relative Click Rate1:45:32
32.Generalized Random Forests1:47:55
33.Local Linear Forests 1:52:05
34.Comparing Regression Forests to Local Linear Forest: 1:53:17
35.Structural Models and Discrete Choice Models1:54:26
36.Combining Discrete Choice Models with
Modern Machine Learning....1:57:29
37.The Nested Logit Factorization Model1:57:59
38.Goodness of Fit (Tuned for CF) 1:59:36
39.Personalized Pricing2:01:24
40.Q & A2:03:03