Very interesting question, in some extent you are right there is less expert available worldwide to teach AI and ML with reference to economics and its application.

there has been work in the last few years to combine econometrics with machine learning (e.g. for causal inference), including my own research.

see this, for example:

1.Article Deep Learning for Causal Inference

2. https://github.com/susanathey/causalTree

for an R implementation of random forests for heterogeneous treatment effect estimation (Rubin's causal model)

Hi Indeed interesting question, I would at some point agree especially in our Pacific today there happens to be a gap till to date. This is with respect to lack of space for machine learning in Economics... And thus I would strongly edge this area to work in..

Thanks

Priynka Sharma

Adriano Marcos Rodrigues Figueiredo ,

While I would like to see my field expand, I wonder whether it would be wise to introduce economics students early on to (ML). Economics has a solid theoretical and mathematical basis that needs to be understood first before engaging in data analysis . Also, the field of ML if taught correctly demands additional mathematics (linear algebra, calculus of variations) and computational knowledge (statistical learning theory, computational theory). I would advise leaving those for Master's or higher.

Just a suggestion

I understand that undergraduate economics students have still faced difficulties studying ML.... I have seen my students having difficulties to understand regular econometrics, more and more difficulties if I use R (but I insist on it and moving to integrate with Python soon), and to move into ML we need more multivariate methods, which indeed they have ways to achieve it. But, as Arturo Geigel said, we need economic Theory **before** using methods. I insist on this. But while we discuss this, I see many other science fields running over economic issues with ML and with lack of economic theory. So, Economics' scientists must hurry up!

Adriano Marcos Rodrigues Figueiredo ,

Maybe there is a middle ground. Maybe it is not that much important on the tools but on teaching early on the context on which to put the research done in ML. To put this research within the context of economic theory. I think showing this type of analysis early on will incentivize students of economics into ML.

Another point worth discussing more, as you point out, is that of other science fields running economic issues with ML. This happens in my field so often that is alarming. I feel your pain and I feel it more since it is not just constrained to economics, it is in civil, mechanical, bio engineering fields, biology, chemistry, etc. Frankly this nonsense of the data scientist applying ML just because there is available data has to stop. Part of being a responsible data scientist is to acknowledge that data has to be put within the context of the field. To do this one has to understand the field, then the problem at hand.

I think that economist must be incetivized to explore ML but also that data scientists have to stop meddling into other fields they do not understand. It is doing more harm than good. ML is here to stay, but it has to be used in the proper context, that is, as a tool within the toolbox and subjected to the understanding and problem of the field.

Actually many aspects of ML are being taught in early academics. I believe, ML is just the name that has been given to a set of economic/statistical/mathematical theories such as covex problems, complex theory, optimization, data science related theories, and many more; combining which later we get ML. However, to ease the process I completely agree with your suggestions to include it in early academics particularly in developing nations.

See this: Machine Learning in Gravity Models: An Application to Agricultural Trade: https://www.nber.org/papers/w27151

And this: Preprint International trade and the gravity model: recent evidence i...

Machine learning has the potential to dramatically enlarge those data sets and allow economists to test their models faster than ever.