Moral Preferences and Political Behavior: A Deep Learning Model of Iran's 2017 Presidential Election on Twitter

Human decision-making, which is influenced by internal and contextual factors, plays a fundamental role in shaping social actions and political behaviors. Understanding the moral preferences underlying these decisions is essential to explaining the mechanisms that govern social affairs. This article, adopting an interdisciplinary approach, uses Twitter to examine moral preferences and their relationship to individuals’ political orientation. The framework of moral foundations was analyzed and validated using deep learning models on a large dataset of tweets. The research findings highlight concepts such as groupism, attention seeking, distrust, value orientation, responsibility, and cognition as key elements in shaping political behavior. These concepts exhibit distinct patterns across different groups.



The findings show that the positive poles of group preferences and attention, along with preferences related to value, responsibility, and trust, play the most dominant role in organizing political discourse on Persian Twitter, and that users’ behavior is influenced by distinct patterns of moral preference rather than being solely subject to formal party labels. Competing political groups prioritize a different set of moral values, and these values ​​are reproduced through intergroup interactions, which are instrumental in the formation of collective identities. The ethics of responsibility and trust, along with group dynamics that tend to conform, have underpinned the stability of political camps. Reformists and fundamentalists, despite discursive conflict, have a similar “moral signature,” while dissenters display a distinct profile based on highlighting power, distrust, and opposition to stagnation, and nonpolitical users focus most on close relationships and micro-identities.