A Behavioral Perspective in State-of-the-art Big Data Methods: Theoretical Background, Empirical Questions, and Implications for Future Research
From smart phones to self-driving cars, big data technology increasingly impacts our lives in ways once thought unimaginable. If advancements in big data technology herald the new digital revolution, as scholars, we must ask how these methods will impact our research and what would be novel questions that we could address in our respective fields. In this talk, we aim to take a deep dive into the recent developments and challenges in big data methods, such as image and facial recognition. We would explore how these algorithms can assist researchers in exploring patterns in unstructured data, such as photographs and videos, and capture behavioral insights from these sources. At the same time, we highlight a critical shortcoming in these methods—shortcut learning—which researchers should be aware of and avoid. We discuss some recent research in these domains, as well as implications for future research. We hope that this talk will help scholars interested in exploring big data research to gain a birds’ eye view of the field, understand where the field came from and where it is going, as well as inspire new research ideas using big data methods.
Dawei Wang is a PhD candidate in Management & Organizations at Northwestern Kellogg School of Management (expected to graduate in June 2022) and a visiting PhD student at the Amaral Lab at Northwestern's McCormick School of Engineering. Dawei's research focuses on human decision-making and artificial intelligence. Specifically, he aims to explore the limits and bounds of different intelligent systems, such as human, machine and collective, as well as theorize the benefits and disadvantages associated with them. Dawei's research is published in top journals such as Journal of Personality and Social Psychology and Psychological Science. He will be joining HKU Business School in July 2022.
Theoretical Background (20 minutes)
- A decision-making perspective from the Carnegie School
- A brief history of artificial intelligence research, from 1950s to 2020s
- Recent developments in deep learning: image and facial recognition
- Rising challenges: shortcut learning in deep learning including NLP
Empirical Questions (50 minutes)
- Does shortcut learning exist in behavioral big-data research?
- Given shortcut learning, how do we still adopt deep learning?
- We can increase the interpretability of models using attribution models
- We can rely on social psychological methods: conduct experiments!
- We can triangulate results using different methods
Implications for future research (10 minutes)
- More interpretable research methods and models
- A greater focus on the quality of data and limitations
- A general awareness of the limits of machine intelligence
Questions and answers (10 minutes)