A Review of Hypothesis Testing and Type I and Type II Errors + The Negative Impacts of Implementing Expert Systems

This teaching demonstration will touch on one of the most fundamental aspects of statistical inference, the hypothesis test. It will focus on not just the mechanical process of conducting a hypothesis test, but the fundamental behavior of data that undergirds our ability to use hypothesis testing, presented at a level approachable by students in even the most introductory statistics courses. Alongside the primary statistical learning objectives of the talk, the lesson will seek to utilize the time to also help spread awareness for pressing social issues, attempting to promote a more equitable future through education.

The latter portion of the presentation will seek to highlight research goals, motivated by examples of harm by expert systems. As we enter the "Age of AI", more and more decision making will fall to expert systems designed to maximize accuracy, profits, or even more esoteric ambitions. The wealth of available data and computing has led to recent developments in generative models, which have in turn allowed us to view the bias inherent in such systems literally with our own eyes, or read it in plain language. While the creators of these models work hard to sanitize the output, less visible expert systems are driving decisions in healthcare, housing, and criminal justice, similarly trained on data that biases and real harm at all levels of society. Understanding the real world impact of implementing these models is key to minimizing that impact on human lives.