Artificial Intelligence in Statistics

Abstract and Poster Explanation

This poster details the relationship between artificial intelligence, including machine learning, and advanced predictive analytics. Artificial intelligence methods (here abbreviated as “AI”) assist statisticians and researchers in data cleaning, data manipulation, pattern recognition, outlier detection, predictive analytics, and data analysis involving multiple sources of disparate or hidden data. AI can also support advanced data analytics, data visualization, and the interpretation of complex data.

This poster proposes AI methods as a helpful and innovative addition to the toolkit of statisticians and researchers. AI methods can support advanced predictive analytics and can also be applied to large and complex datasets. In addition, AI methods can help discover previously hidden relationships in data. This feature is especially valuable in contexts where complex data is rapidly accumulating, rapidly changing, or both.

This poster also warns that AI cannot and should not completely replace human statistical thinking. AI methods, especially those based on machine learning, may produce highly simplified outcomes that emerge from complex processes. These outcomes can contain bias and may also result from overfitting. In some cases, they may reflect “black-box” procedures that are difficult to explain. Therefore, AI and machine learning methods need to be used under rigorous human supervision.

The AI statistical workflow presented in the poster condenses the process into data cleaning, data exploration, data modeling, data validation, data interpretation, and data reporting. Statisticians must confirm that machine learning and predictive analytics results lead to valuable and rational conclusions.

The poster argues that AI methods can strengthen statistical practice, but only when used in controlled and supervised ways. AI improves the efficiency of complex data analysis and predictive modeling, whereas human statistical thinking provides the rational limits, structure, and meaningful synthesis needed to interpret the results.