Paper Type
full
Description
Previous empirical studies reveal that the personalities of senior executives may influence corporate decisions. Grounded in the big-five personality model established in the field of psychology and guided by the design science methodology, the work reported in this paper examines the predictive power of executives' personalities on corporate behaviors. Our main contributions are twofold. First, we design a novel deep learning-based artifact, namely pAttCLSTM which analyzes the senior executives’ personalities on a large scale through social media data. In our experiment, pAttCLSTM outperforms other state-of-the-art detectors over two benchmark datasets. Second, empowered by pAttCLSTM, we empirically verify the predictive power of executives' personalities on firms' policy variables by using a dataset including Tweets of 507 executives worked in S&P 1500 companies. To our best knowledge, this is the first successful design of an automatic detector for senior executives’ personalities, and hence to examine their impact on firms’ behaviors.
Recommended Citation
Yang, Kai and Lau, Raymond, "Detecting Senior Executives’ Personalities for Predicting Corporate Behaviors: An Attention-based Deep Learning Approach" (2019). ICIS 2019 Proceedings. 13.
https://aisel.aisnet.org/icis2019/data_science/data_science/13
Detecting Senior Executives’ Personalities for Predicting Corporate Behaviors: An Attention-based Deep Learning Approach
Previous empirical studies reveal that the personalities of senior executives may influence corporate decisions. Grounded in the big-five personality model established in the field of psychology and guided by the design science methodology, the work reported in this paper examines the predictive power of executives' personalities on corporate behaviors. Our main contributions are twofold. First, we design a novel deep learning-based artifact, namely pAttCLSTM which analyzes the senior executives’ personalities on a large scale through social media data. In our experiment, pAttCLSTM outperforms other state-of-the-art detectors over two benchmark datasets. Second, empowered by pAttCLSTM, we empirically verify the predictive power of executives' personalities on firms' policy variables by using a dataset including Tweets of 507 executives worked in S&P 1500 companies. To our best knowledge, this is the first successful design of an automatic detector for senior executives’ personalities, and hence to examine their impact on firms’ behaviors.