automat identif quasi experiment design discov causal knowledg david jensen andrew fast brian taylor marc maier research social behavior scienc routin reli quasi experiment design discov knowledg larg data base quasi experiment design qed exploit fortuit circumst experiment data identifi situat call natur experi provid equival experiment control random qed research domain divers sociolog medicin market draw reliabl infer causal depend experiment data identifi exploit qed remain painstak manual activ requir research scour databas appli substanti knowledg statist recent advanc express databas increas size complex provid condit automat identifi qed paper describ system discov knowledg appli quasi experiment design identifi automat demonstr qed identifi tradit databas schema identif requir small number extens schema knowledg quasi experiment design encod order logic theorem prove engin describ kei innov enabl system includ method automat construct experiment unit creat aggreg variabl unit show appli result design identifi import causal depend real domain provid exampl academ publish movi make market peer product system final discuss integr qed approach causal discoveri includ joint model direct experiment 