Winson Peng, Ph.D.

Winson Peng, Ph.D.

Professor of Communication

Posts by Collection

portfolio

publications

Can Large Language Models Estimate Public Opinion about Global Warming? An Empirical Assessment of Algorithmic Fidelity and Bias

Published in PLOS Climate, 2024

Recommended citation: Lee, S., Peng, T. Q., Goldberg, M. H., Rosenthal, S. A., Kotcher, J. E., Maibach, E. W., & Leiserowitz, A. (2024). Can large language models estimate public opinion about global warming? An empirical assessment of algorithmic fidelity and bias. PLOS Climate, 3(8), e0000429. https://www.doi.org/10.1371/journal.pclm.0000429

talks

Public agenda dynamics: Competition, cooperation, and more!

Published:

Agenda-setting has been extensively investigated over the past 50 years as a formal theory of media effects. Of the various research questions about agenda setting, the competitive nature of the interissue relationship on public agenda is particularly intriguing but has not been settled. In the past decade, we have conducted a series of empirical studies (Xu et al., 2013; Sun et al., 2014; Peng et al., 2017; Peng & Zhu, 2022) to understand public agenda dynamics in both online and offline settings. I will first present two studies based on large-scale social media data: one re-tested the issues competition theory and the other proposed and tested an issues coopetition model. Then I will share our latest effort that proposed an ecological perspective to explicate interissue relationships on public agenda. With the Gallup Most Important Problem (MIP) polls in the United States from 1958 to 2020, we empirically show that the issue ecosystem of the American public is essentially competitive and that the balance of competition and cooperation has remained unchanged over time.

When communication research meets visual analytics: An integration of arts and sciences

Published:

Big data is “of the people, by the people, and for the people”. But data could not speak for itself. The interdisciplinary collaboration between computer scientists and social scientists helps restore silent data into dynamic interaction between social topics. This talk will introduce two interdisciplinary studies (Xu et al., 2013; Sun et al., 2014) which aimed to examine how social issues compete and cooperate with one another for public attention on social media. Building upon classical agenda-setting theory in communication research, the two studies proposed two visual analytical systems that can facilitate panoramic and in-depth analysis of topic interaction on social media. Reflecting on this collaboration, I will discuss how social scientists and computer scientists can work together to create a 1+1>2 effect.

Navigating the Interdisciplinary Frontier: Computational Social Science and Public Agenda Dynamics

Published:

Computational social science (CSS) has spearheaded a transformative era in research, permeating every facet of the social sciences. CSS stands as an intersection, seamlessly merging computational algorithms and statistical methodologies from computer science and statistics with the rich tapestry of theoretical concepts and frameworks drawn from diverse social science disciplines. In this talk, I embark on a decade-long research journey that delves into the intricate realm of public agenda dynamics—a seemingly simple yet profoundly intricate facet of public opinion research. Our expedition begins with two comprehensive studies rooted in expansive social media datasets: one reevaluating the conventional issues competition theory and another introducing and assessing an innovative issues coopetition model. Furthermore, I unveil our most recent endeavor—an innovative ecological perspective that unravels the intricate web of interissue relationships on the public agenda. This perspective was empirically tested using Gallup Most Important Problem (MIP) polls spanning from 1958 to 2020 in the United States. Through this lens, I illuminate how CSS reinvigorates our understanding of existing and emerging questions within the realm of social science. Lastly, I explore the dynamic synergy of interdisciplinary integration within CSS, highlighting its potential to redefine the landscape of social science research and education.

Catalyzing Public Opinion Research with Large Language Models: Promise or Peril?

Published:

Large language models (LLMs) have garnered substantial attention across various academic disciplines. In this presentation, I will discuss the potential applications of LLMs in advancing public opinion research, both as tools and as dynamic contributors to the research process. I will introduce several ongoing research initiatives that leverage LLMs to elicit human responses to survey inquiries, particularly within the domains of climate change, cultural values, and political behaviors. Through a synthesis of empirical findings from external studies and our own investigations, we will explore the opportunities and challenges associated with the use of LLMs in public opinion research. This talk aims to prompt contemplation of LLMs’ potential as not only language models but also active participants in the intricate interplay of data and human understanding.

teaching

Computational Social Science: Principles and Applications

Graduate course, Michigan State University, Department of Communication, 2022

Computational thinking and methods have been widely discussed and adopted by social scientists in various subject areas (e.g., anthropology, communication, political science, public health, and sociology). This course is about how computational social science (CSS), as an emerging paradigm of research, changes the way in which social scientists empirically observe and understand human society. The course is composed of three modules:

  • The first module focuses on the fundamental principles in CSS, including research design, implementation, data collection and management, and data analysis.
  • The second module focuses on the conceptualization and modeling of three types of data preeminent in CSS research: text, time, and structure.
  • The third module concentrates on the application of computational methods in some prominent research domains, such as health communication, political communication, and user analytics.

Communication Research Design II

Graduate course, Michigan State University, Department of Communication, 2023

This course is the second in the Ph.D. statistics sequence at the Department of Communication, MSU. It has a prerequisite of COM 901 or its equivalent. It is a four-credit class. The course aims to provide PhD students with a working knowledge of the assumptions, concepts, and theories underlying the most frequently used multivariate analysis techniques in quantitative social and behavioral sciences. These techniques include, but are not limited to, multiple regression, exploratory and confirmatory factor analysis, path analysis, structural equation modelling (SEM), multilevel analysis, and time series analysis. The selection of specific topics may be tailored to students’ research needs. The focus will be on practical issues such as selecting the appropriate analysis, preparing data for analysis in the popular statistical tools (e.g., SPSS, AMOS, or R), interpreting output, and presenting results of a complex nature. This course is a mixture of lectures, discussions, and hands-on practices.