Synergies Between Economics and Computer Science

Having taken courses and worked with both fellow economists and computer scientists alike at Stanford, several differences emerge (specifically from the lens of academia).

Computer science focuses more on solving problems.

Computer scientists approach research and discovery much more from the perspective of: “will it work”?

Deriving asymptotic results and fitting into more elegant theories are often viewed as second-order goals, rather than the primary task of more efficiently and reliably predicting, for example, the number of clicks on a website.

It’s easy to see these points lived out in the patterns of the faculty and graduate students.

For example, Andrew Ng, one of the top computer scientists in the department, is now Chief Scientist at Baidu, China’s Amazon-equivalent. Put simply, many faculty either create start ups, take a sabbatical with a company, or simply partner with companies on an ongoing basis.

Economics focuses more on developing theory and creating structure in the midst of chaos.

While there will always be plenty of people who throw rocks at economics because it hasn’t solved all our problems yet, economics is as much of an art as a science.

Rightly or wrongly, there’s a big premium associated with writing research papers that develop organizing theories or provide microeconomic evidence behind a theory.

Find an algorithm that works well at predicting X? That’s nice, but you’re probably not going to find an economics journal outlet to publish it.

These differences both have their costs and benefits, but what’s exciting is how the two disciplines complement and reinforce each other.

What can we learn from their differences?

Among other reasons, one of the luxuries of academia is the opportunity to become an entrepreneur of ideas. You can push yourself as hard as you want — to learn, to serve, and to produce.

I’ve worked hard to invest in learning from both computer science and economics, and these dual disciplines induce several exciting areas of synergy.

While I’ll be writing about these differences more in the future, one that I want to focus on right now is the concept of causality in economics.

Economics tends to approach questions from the perspective of: “how is X related to Y?” Computer science, in contrast, tends to approach questions from the perspective of: “what function f(.) best describes the mapping between X and Y?”

Understanding the distinction between causality and correlation has fundamental implications.

While that’s an oversimplification, it’s a useful one since the concept of causality is so fundamental to the work we do. Regardless of your work, every organization prioritizes understanding how a given decision will affect its attainment of a particular goal.

Economics provides a class of tools for recovering these causal effects, which is especially tough since interactions and mutually-reinforcing phenomena permeate the environment around us.

Whenever we see movement in some measurement, whether it’s profits or student test scores, we should train ourselves to ask questions about what’s driving that variation and how it might be linked with the outcomes we care about.

If we’re interested in how changes in incentive pay contracts affect employee performance, then the job of a data scientist or economist is to figure out how to isolate movements in incentive compensation that are not correlated with unobserved determinants of employee performance. Otherwise, we fall prey to the prototypical “correlation does not equal causation” claim.

Feel free to check out my academic webpage at www.christosmakridis.com.