Understanding User Engagement
Introduction
An ASA Marketing Section Webinar, 26 Nov 2024
INTEREST CATEGORY: MARKETING RESEARCH
POSTING TYPE: Events
Posted by: Shibo Li
American Statistical Association (ASA) Marketing Section Webinar on November 26, 2024
The next speaker in our ASA Marketing Section webinar series in Fall 2024 is who is an Assistant Professor of Business at Columbia Business School. She uses machine learning, econometric, and experimental methods to study how emotions impact consumer behavior, taking advantage of the rich unstructured data (text, images, video, music) that is increasingly available. A distinguishing feature of her interests involves going beyond ML’s use in prediction to study how to incorporate domain-specific theoretic and managerial knowledge into ML systems and make them more interpretable. She also has a broader interest in questions at the interface of marketing and society (e.g., fairness). Her work has received recognition, including the American Statistical Association Dissertation Proposal Award and the MSI Alden G. Clayton Dissertation Proposal Award. Hortense obtained her Ph.D. in Quantitative Marketing from Yale.
The time and date of the webinar are 9pm-10pm Eastern time, Tuesday November 26, 2024, with the webinar link: . You are welcome to see the webinar details and bookmark the . Below is the title and abstract of the paper she will be presenting:
Title: A Generative Approach for Modeling Expectations and Uncertainty in Stories to Understand User Engagement
Abstract: Understanding when and why consumers engage with narrative content is crucial for content creators and platforms, yet modeling engagement has been challenging due to the unstructured nature of stories and limited data relative to potential feature dimensionality. Drawing on economic theories of expectations and uncertainty, we introduce a novel framework that uses large language models (LLMs) to simulate readers’ beliefs about how stories might unfold. Our method generates multiple potential continuations for each story and extracts features related to expectations, uncertainty, and surprise using established content analysis techniques. While previous work has focused on features from consumed content, our approach captures forward-looking beliefs that may influence engagement decisions. Applying our method to around 35,000 book chapters from Wattpad, we demonstrate that our framework complements existing feature engineering techniques by amplifying their marginal predictive power by 20-116%. The results reveal that different types of engagement—continuing to read, commenting, and voting—are driven by distinct combinations of current content features and anticipated story elements. Our framework provides a novel way to study and explore how readers’ forward-looking beliefs shape their engagement with narrative media, with implications for marketing strategy in content-focused industries.
We look forward to seeing you in the webinar.