Automated question answering for unveiling leadership dynamics in U.S. Presidential speeches
Krzysztof I. Rybinski · Expert Systems with Applications · 2024
This study introduces an innovative methodology utilising deep learning and automated question-answering techniques to explore leadership dynamics. The research analyses 989 speeches from U.S. Presidents, ranging from George Washington to Joe Biden, applying a machine learning model to decipher the essence of effective leadership within the context of presidential rhetoric. This approach facilitates an interrogation of “What constitutes a great leader?” by extracting pertinent attributes from these presidential discourses. The study conducts a comprehensive statistical examination of the responses generated, unveiling the evolution of presidential perspectives on leadership across historical epochs. Furthermore, the research employs a regression analysis that extends over 120 years, integrating the attributes of outstanding leadership with the Economic Policy Uncertainty (EPU) index specific to the United States. The findings indicate that periods marked by heightened uncertainty necessitate leaders with a charismatic approach, whereas servant leadership is more effective during times of reduced uncertainty. An extensive validation and robustness assessment of the study affirm that the outcomes are steadfast, notwithstanding variations in key parameters of the deployed machine learning models. Moreover, these findings align coherently with the qualitative assessments of U.S. Presidential views on leadership undertaken in prior research, thus contributing novel insights into the intricate relationship between leadership qualities and historical and economic contexts.