AI in Electoral Prediction: Ethical and Political Analysis

Abstract

This work is the analysis of artificial intelligence (AI) use to predict elections, the focus of which is on the technical capabilities and the ethical and political implications of AI forecasts.  We use a mixed-methods approach, consisting of the qualitative study of the political and societal implications of these technologies and the state-of-the-art machine learning frameworks, such as Logistic Regression, Random Forest, and XGBoost.  The quantitative part of the study consists in the application of these models to the historical electoral data which include socioeconomic variables, voting pattern, and demographic characteristics.  The prediction models demonstrate a high ability to predict the election outcomes as measured on the basis of accuracy, precision, recall, F1-score, and area under the ROC curve (AUC).