Artificial Intelligence and Machine Learning in Economics
– Updated: Aug. 6th, 2019
Tech. (Technology) companies are hiring more and more economists. In a job-hire talk that I was present, representatives from Amazon claimed to almost double their economists force, every year. What can then economists contribute to the CS advances of artificial intelligence and machine learning? A lot! But don’t take my word for it. Read this paper about what economist and economics can contribute to tech companies from two of the most high-profile economists heavily involved in the tech world, Susan Athey and Michale Luca. Check out a more condensed version of it in Harvard Business Review’s post.
As economists, If it’s one thing that we know and care about with regards to AI and ML is that more data availability in conjunction with AI and ML techniques have greatly reduced the cost of prediction. Prediction of what? Well, more traditional prediction problems involve demand estimation, cost projections with increased scale, etc. However the reduction in the cost of prediction has had another important consequence: A lot of problems that we did not used to think as prediction problems can be reformulated as prediction problems and solved cheaply using the aforementioned techniques. Example? Self-driving cars. Read more about it in this article and this website by Ajay Agrawal. Also check out the Creative Destruction Lab also founded by Prof. Agrawal.
The Brookings institution has some really fascinating work in this area. Read one of their articles “How humans respond to robots: Building public policy through good design“. There is a conference on the “Economics of AI“. The book of the 2017 conference is incredibly fascinating. Finally, a good place to keep up to date with the workings of economists in the industry is the National Association for Business Economics. Also check out their annual meetings and conferences that consistently attract some of the leading economists working both in academia and in the industry.
Artificial Intelligence in Society
In a time where the world observes the emergence of artificial intelligence with both excitement and uncertainty about how AI will shape the future of our societies, a new wave of research in AI is aiming to structure the AI knowledge and put it to use in tackling social issues. This article nicely summarizes the concept of “AI for social good”. For real-life societal applications of AI and ML do check out the USC Center for Artificial Intelligence in Society (CAIS), inspired and co-founded by the incredible Milind Tambe and the work of him and his team teamcore at USC Viterbi. Their incredible work ranges from applying insights from machine learning and game theory (security games) to tackle wildlife problems (for example helping rangers predict the most likely locations of elephant poachers at the Murchinson Falls National Park inUganda) to combining AI and game theory to help with security and public safety (for example in random checks at LAX and public transit in LA). Also check out Kevin Leyton-Brown, Professor of Computer Science at UBC and his team whom I almost joined for a PhD in Computer Science a few years ago. Kevin chairs the UBC’s new Center for Artificial Intelligence Decision-Making and Action.
AI for social good has attracted the attention of the largest organizations like Google and Microsoft. Check out Google’s AI for using AI to solve global issues and Google’s AI Education on using AI for social good. Recently, Google has announced the opening of another research center in Bangalore, India, dedicated to applying AI insights to solve problems in healthcare, agriculture, education and more. For Microsoft, the dedicated page for the application of AI to society is here. Read McKinsey and Company ‘s article on the ways in which AI is used to tackle most challenging social problems (This article summarizes this discussion paper). I am extremely grateful to Jonathan Woetzel of McKinsey and Company for the generous fellowship I am receiving from them as reward for winning the “3rd year paper award” at the department of economics at USC. Another great report touching on urban sustainability, health and public welfare, supported by the Computing Community Consortium is here. What is more, read this website for applications of AI in development. What is more, there’s a dedicated workshop for papers directly related on applications of AI to social problems.
Fairness in Machine Learning
Fairness in Machine Learning is a growing area of research and of massive policy debates. The gist of it is simple. While ML algorithms are great for out-of-sample prediction, they train on historical datasets generated by human decisions, which.. as history teaches us, are oftentimes biased in favor or against of a social group. For the sake of argument, let’s take a company that wants to implement a ML software to suggest hiring and promotion decisions inside the firm. The algorithm’s job is then to predict which candidate is more likely to succeed in a given position. To do that, the algorithm would train on historical decisions of the firm. These include the candidates’ credentials and characteristics (x variables) as well as how well they performed in a given position which could be proxied by whether they got hired or promoted (y variable). Suppose now that past hiring managers had a bias for male candidates. Thus male candidates had a higher chance of being hired and promoted for the same credentials. That would then lead the algorithm to predict that a male candidate is more likely to succeed and thus suggest the hiring or promotion to that candidate.
The problem is that simply removing sensitive variables like race and gender from a dataset is not enough to make the bias go away since most of these variables are highly correlated with other characteristics. I have had the chance to work with some incredible people on a related project where we proposed a system in which the algorithm got penalized by the prediction of those sensitive variables. We proposed a setup in which a policymaker could choose the level of fairness desired (which would obviously tradeoff with prediction accuracy) and the algorithm would make predictions given that level.
If you’re interested for a light reading I suggest this easy-to-read book “Weapons of math destruction”. For more in-depth reading I recommend going through this archived course from UC Berkeley’s Moritz Hardt who is at the forefront of research in this area. His book (coauthored with Solon Barocas and Arvind Narayanan) is also a great resource. It is available online here.
Artificial Intelligence and Behavioral Economics
The more artificial intelligence integrates into everyday life and the more people interact with softwares powered by AI and machine learning, the more behavioral economics can contribute to the development and application of these systems. While research in computer science and information systems is largely focused on improving the efficiency and accuracy of the algorithms, behavioral economics can examine the psychological side of the human-machine interaction. Read this article and these slides by Colin Camerer where he discusses two potential avenues for interaction between artificial intelligence and behavioral economics.
Machine Learning and Behavioral Economics
A big part of behavioral economics as discussed in the section Behavioral Economics in Policy above is identifying and applying subtle behavioral interventions (“nudges”) to help people make better decisions. In this aspect, machine learning algorithms of classification can help behavioral interventions through personalization. ML algorithms can help categorize heterogeneous individuals that are likely to benefit from different behavioral nudges, thus helping maximizing the overall policy impact. “Normally, an experimenter would look for the most effective intervention and apply it to everyone. Why should one size fit all?“. Read more on these posts: post1, post2, post3, and in this paper.
Machine learning can also inform behavioral game theory through providing better ways of modeling of strategic interactions as well as help consolidate the findings of enormous data from experimental economics. Check out the video lecture below of one of my favorite computer scientists Kevin Leyton-Brown discussing his work on combining machine learning models to predict human behavior in various economic experiments by using a large pooled dataset of numerous experimental results:
Feedback Systems, Review Aggregators & Recommender Systems.
While the above may, on first sight, sound purely like topics that Computer Scientists study, it’s not hard to see the connection with economics. Feedback systems collect information about products, sellers, consumers, locations, workers and employers in order to facilitate trust and help interested parties discover the other party’s quality. They do that through voluntary feedback contributions from parties that have previously interacted with one another. As such, voluntary provision of feedback in online marketplaces is a crucial component that facilitates transactions through adding the missing trust between strangers that have never interacted before. In addition, these feedbacks help the platforms (and as such other consumers) to differentiate between competing products. As a behavioral economics working on information economics, this area is of great interest to me and the area of my job market paper.
I am a firm believer in the power of voluntary contributions of feedback. Most of my reviewing activity is concentrated on Amazon.com (where I have contributed over 80 reviews and earned almost 400 “helpful votes”) and Google.com (where I am a level-6 “local guide”) Check out my reviewer profiles here-Amazon, and here-Google.
Get in touch
Please do get in touch if you have any questions or concerns. I’m glad to discuss new ideas and suggestions for research or other work. Shoot me an email, text me on skype or write your message in the box below. I’ll get back to you as soon as time permits.