Seminar (Dept. of Organization and Human Resources)

Theme：Using Machine Learning to Advance the Science of Culture: Identifying Novel Cultural Dimensions, Quantifying Cultural Distance, and Predicting Innovation

Speaker：Krishna Savani (Nanyang Technological University, Singapore)

Time：2021-05-12 16:00

Language：English

Zoom Meeting： 

https://zoom.com.cn/j/61660406165?pwd=ZXU4dTdLZTgyM25KcVhGK1NSVzcxdz09

Meeting ID：616 6040 6165

Password：086255

ABSTRACT：

Consider these questions: (1) Which attitudes, values, and beliefs best explain cross-national variation in important outcomes, such as innovation? (2) Which attitudes, values, and beliefs best differentiate the world’s cultures in general? (3) If all countries were to be differentiated based on a single latent dimension, what would that dimension be? In this research, we used machine learning methods to answer these longstanding questions.

In the first project, we trained a deep learning model to predict 86 countries’ innovation scores from 314,989 residents’ responses to 680 questions included in the World Values Survey. Even in a subset of the data to which the model was never exposed, the correlation between the model’s predicted and actual innovation scores was r = .96. Follow-up analyses revealed novel predictors of country-level innovation, including feeling pride in the country’s long history, being opinionated, and perceiving homosexuality as justifiable. These findings help identify novel antecedents of innovation, which can be tested using experiments.

In the second project, based on individuals’ responses to 594 items measuring attitudes, values, and beliefs, a deep learning model identified which of 98 countries they belonged to with 90% accuracy. The model then identified attitudes, values, and beliefs that were most helpful in predicting people’s county of origin. We used these to create an AI-based Culture scale, which featured a number of themes that are not covered by existing cross-cultural scales (e.g., religion, politics, and society). This scale outperformed existing scales in explaining cross-cultural differences in behavior.

In the third project, we used an autoencoder neural network to compress 594 items measuring attitudes, values, and beliefs into a single dimension. The model could identify people’s country of origin with 44% accuracy based on that dimension. The dimension was defined predominantly by political orientation, religiosity, and conscientiousness. Countries’ location on this AI-based Culture Dimension explained national levels of innovation and Covid-19 severity better than existing culture dimensions.

Overall, this research highlights how machine learning methods can be used to generate novel theoretical insights in the social and organizational sciences.

SHORT BIOGRAPHY：

Krishna Savani is the Provost’s Chair in Business, Director of the Culture Science Innovations Center, and Associate Professor of Leadership, Management, and Organization at Nanyang Business School, Nanyang Technological University, Singapore. He obtained his PhD in Psychology from Stanford University and has done a Postdoctoral Fellowship at Columbia Business School. He has conducted extensive research on culture, norms, choice, decision making, lay theories, and policy positions. His research has been published in multiple academic journals, including Journal of Personality and Social Psychology, Psychological Science, Journal of Applied Psychology, and Organizational Behavior and Human Decision Processes. Dr. Savani has been recognized as a Rising Star by the Association for Psychological Science in 2015, and was featured in Poets and Quants’ “Top 40 Business Professors Under 40” in 2018. He is currently an Associate Editor at the Journal of Personality and Social Psychology, and serves as an Editorial Board member at Organizational Behavior and Human Decision Processes.