Hey Scott, thanks for all the effort you’re putting into democratising causal-inference. I’m equally enthusiastic about it becoming mainstream. My belief is that we need more folk that can derive causal inference from observational data. I belong to the chemical industry, and a company that is financially investing in data-science. While that’s a positive sign we are limited by leaders that can imagine the use-cases that could attract data-science application. A lot of the growth in application is happening bottom-up and horizontally, just like you’ve cited. I can’t imagine these folk investing in more infrastructure that could enable A/B experiments. Our industry lacks the velocity in feedback loops to take advantage of A/B experiment infrastructure that tech firm deploy. But I sincerely believe that if we were to apply systematic causal inference based on observational data we’d be able to appeal to the imagination of the senior stakeholders and hopefully accelerate the maturity of causal applications. I’ve been set-studying to hone my skills in causal inference but could have done with some formal coaching from practitioners like you. A model that allows expertise to be scaled not just through education but an apprenticeship model could turbo charge the causal inference field.
I hope your course will have a little bit of DAGs and SCMs in it at least, even if it mostly thinks of the world in terms of Rubin's "giant pandas dataframe with some NAs".
SCMs are going to be very intuitive to people who come from CS backgrounds and write code all day, and I think that's still where most data scientists come from. Likewise DAGs, I have the feeling they are still these weird objects to economists, but to anyone with an undergrad CS degree, they're a big part of every intro to algorithms course.
The Ongoing Role of Machine Learning Engineers and Data Scientists in Industry's Lucas Critique
Hey Scott, thanks for all the effort you’re putting into democratising causal-inference. I’m equally enthusiastic about it becoming mainstream. My belief is that we need more folk that can derive causal inference from observational data. I belong to the chemical industry, and a company that is financially investing in data-science. While that’s a positive sign we are limited by leaders that can imagine the use-cases that could attract data-science application. A lot of the growth in application is happening bottom-up and horizontally, just like you’ve cited. I can’t imagine these folk investing in more infrastructure that could enable A/B experiments. Our industry lacks the velocity in feedback loops to take advantage of A/B experiment infrastructure that tech firm deploy. But I sincerely believe that if we were to apply systematic causal inference based on observational data we’d be able to appeal to the imagination of the senior stakeholders and hopefully accelerate the maturity of causal applications. I’ve been set-studying to hone my skills in causal inference but could have done with some formal coaching from practitioners like you. A model that allows expertise to be scaled not just through education but an apprenticeship model could turbo charge the causal inference field.
I hope your course will have a little bit of DAGs and SCMs in it at least, even if it mostly thinks of the world in terms of Rubin's "giant pandas dataframe with some NAs".
SCMs are going to be very intuitive to people who come from CS backgrounds and write code all day, and I think that's still where most data scientists come from. Likewise DAGs, I have the feeling they are still these weird objects to economists, but to anyone with an undergrad CS degree, they're a big part of every intro to algorithms course.