Could we market the stock market 20 years out using machine learning to go straight from data to a prediction? As David Laibson said, one of the interesting aspects of the stock market is it combines billions of peoples opinions and forecasts to form a stock price. The price that is created at the end is approximately a random walk, and so it may be difficult to model the stock market. When analyzing so many different inputs, as in the example of the stock market, it becomes extremely necessary to understand how a machine learning model is drawing connections and making predictions of the data. The Professors talk about the need to understand the psychology of AI and why AI makes certain decisions. Whether algorithms reflect biases that we would ethically not approve of is something that is crucial to be understood. It takes computer and data scientists who are acutely aware of the data they are using, the coefficients they are using, and whether their input variables could be used as a proxy for unfair biases, such as screening by race, to create proper machine learning models. As they conclude, even with machine learning, we may not escape human error, as these models try to mimic human decision making.
top of page
bottom of page