Jahnavi Enaganti
Pronouns: she/her
Research Mentor(s): Lu Wang
Research Mentor School/College/Department: Computer Science and Engineering / Engineering
Program: UROP
Session: Session 7 (4:40pm – 5:30pm)
Authors: Lu Wang, Jahnavi Enaganti, Frederick Zhang
Abstract
This study was conducted to train a machine learning model to recognize the implicit ideological bias of different news sources based on the language they use, with a special focus on their moral events. Moral events, as put forward in the Moral Foundations Theory by Haidt et al. (2007), include events that encompass the five moral foundations of care/harm, fairness/cheating, loyalty/betrayal, authority/subversion, and sanctity/degradation. By recording the use of different moral foundations in news articles, we hope to recognize their ideological leaning. Therefore, the dataset of news articles was annotated to include the moral foundations present as well as the ideological leaning of the article. Furthermore, several annotators were employed to prevent any internal biases from impacting the dataset. By creating a high-quality dataset, this study aims to train an efficient and effective machine learning model that can accurately discern the ideological bias and political leaning of a news article from its contents. From the current dataset, the machine learning model flan-t5 was able to extract the moral agent 66.79% of the time, the moral agent 52.03% of the time, and the morality embodied 32.49% of the time, extracting the care/harm morality with the greatest success. This research study would allow individuals to consume media more critically and make them less prone to manipulation by the news they consume. Moreover, it would also prevent news articles from creating biases in their readers and allow people to make more informed decisions about the world around them. Haidt, J., Graham, J. When Morality Opposes Justice: Conservatives Have Moral Intuitions that Liberals may not Recognize. Soc Just Res 20, 98–116 (2007). https://doi.org/10.1007/s11211-007-0034-z
