On Oct 17, 2022, the second “Topical Social Sciences Seminar” of Fudan Institute for Advanced Study in Social Sciences (Fudan IAS) was held in Fudan University. The speaker of this session was Dr. Wen Qu, assistant professor and research fellow at Fudan IAS and doctor of quantitative psychology from the University of Notre Dame. Her topic was Sentiment Analysis in the Social Sciences. Prof. Sujian Guo, Dean and Distinguished Professor of Fudan IAS, chaired the seminar. All researchers and post-doctoral fellows of Fudan IAS attended the seminar.

Dr. Qu first introduced the expertise and interdisciplinary features of her research area as well as the latest developments, methodological innovations and applications of the quantitative psychology based on new technological tools. She pointed out that sentiment analysis (SA), as a method of text analysis, has been increasingly emphasized and applied in social sciences, behavioral sciences and pedagogy in recent years. Then, she introduced in detail the application of SA in social sciences by taking teaching assessment as an example. Taking more than 10 million teaching assessment data from the last two decades as the research object, she built an aspect-based SA model for the textual comment data, and created a sentiment lexicon as well as an aspect expression table for teaching assessment, and also generated a labeled dataset for it, which provided a training set for the machine learning methods in this field. She compared the trends of rating assessment data and textual sentiment data, and found that the rating data and textual data revealed different trends. While rating data shows a downward trending, textual sentiment data is more positive. The common view is that there is a gender bias in teaching assessment: male teachers will be more likely to receive better assessment. By using mass data of teaching assessment, she explored whether there is an underlying gender bias in different types of data. Finally, she stated the results of the study: while the numerical data on overall ratings as well as course difficulty verifies the general view, there is no significant difference between assessment of male and female teachers under SA, and female teachers instead receive more positive feedback in some assessment aspects. Also, the word embedding distance analysis provides a different approach to the textual data as well as an analytical perspective. This also confirms that writing reviews is less susceptible to implicit bias than rating, and that textual data can provide more dimensions and perhaps fairer assessment than numerical data.

After the Dr. Qu’s speech, the participants discussed and exchanged their views on the concepts, theories, approaches, data and other issues related to this topic.