Â鶹´å

How Emotions and Moral Values Shape the Stability of our Attitude

Why do some attitudes hold steady while others shift readily? Psychologists call this difference attitude stability: how consistently someone holds an attitude over time. Some attitudes are fleeting and update when new information is available. Others are more enduring, guiding our actions, resisting persuasion, and influencing how we interpret new information. Understanding what makes certain attitudes so stable helps explain why some public debates (e.g., about vaccines, climate change, immigration) can stay polarized, no matter the evidence.

A common explanation is morality. When we see an issue as a matter of right and wrong, our stance often hardens. Moral convictions can make attitudes feel non-negotiable. Yet emotions are deeply intertwined with moral thinking. Could emotions actually change how strongly we hold our beliefs rooted in morality?

To explore this question, our study analyzed more than 15 million tweets from 1.7 million people about COVID-19 vaccines, posted between September 2020 and August 2021. This period captured key moments in the pandemic, including vaccine rollouts, public debates, and major policy shifts. Instead of relying on self-reported attitudes, we looked at the language people used online when expressing their attitudes towards COVID-19 vaccines. We analyzed which moral values (e.g., care, fairness, or liberty) and emotions (e.g., anger, sadness, joy) appeared in each tweet and then tracked how consistent each person’s stance on vaccines (i.e., pro vs anti) remained over time. To measure attitude stability, we calculated how often the vaccine stance stayed the same across consecutive tweets by the same user (values near 1 = more stable; near 0 = less stable).

Key findings

Not all moral values make beliefs stronger. Overall, we found that moral language was linked to less stable views: people who used moral framing were actually more likely to shift their attitudes over time. There were two exceptions: tweets framed around care (protecting others) and loyalty (standing with one’s group) were associated with more stable positions towards COVID-19 vaccines. In contrast, fairness, authority, purity, and liberty were tied to more flip-flopping. This finding challenges a long-held idea from survey research that moralized beliefs are always the most enduring. Our results suggest it depends which moral value people draw on and how they perceive its association to their attitude.

1. Emotions change how morality affects stability. Emotions can amplify, diminish, and have limited effect at all on the relationship between morality and stability. In general, emotions tend to make attitudes more stable, regardless of the moral value involved. In particular, anger tended to stabilize attitudes regardless of the co-occurring moral value. It strengthened the stabilizing effects of care and loyalty, while dampening the effects of fairness, authority, purity, and liberty. Other emotions had more nuanced effects. For example, sadness increased stability when it accompanied care-based moral language but decreased stability when paired with liberty-focused messages. The same emotion could either cement or soften people’s beliefs depending on the moral frame they used.

This figure displays the interaction effect for each combination of moral value and emotion. A value greater than 1 (red) indicates that combination of moral value and emotion is associated with more stability, while a value less than 1 (blue) indicates that combination is associated with less stability.

3. Observational social media studies complement surveys. Most research on attitudes happens in labs or surveys, where people answer questions about how they feel. We took a different approach: looking at organically occurring language on social media, where people express their views publicly and often over time. This allows us to study attitude stability under the social and environmental constraints present in online environments.

 Takeaways

This work has important implications for researchers who study beliefs, policymakers who shape public communication, and people who design and moderate online platforms - including the following:

•  Emotion and morality are deeply intertwined. Emotions don’t just add intensity to moral beliefs, they can change how those beliefs behave. These results show that morality and emotion should not be studied in isolation.
• Moral appeals in public communication do not necessarily produce more stable or unified beliefs. In some cases, invoking values like fairness or liberty may make public attitudes less predictable when emotions run high. By understanding how emotions shape moral reasoning, communicators can frame messages that promote reflection and understanding rather than division.
• Platforms like X (formerly Twitter) provide the stage where these emotional and moral conversations unfold. The combination of moral language and strong emotions - especially anger - can make beliefs more entrenched and amplify division. Design choices, such as how engagement metrics reward emotionally charged content, may unintentionally reinforce this dynamic. Insights from this work could inform more thoughtful approaches to moderation, recommendation systems, and digital dialogue.

Read the paper here: Phillips, S. C., Ng, L. H. X., Zhou, W., & Carley, K. M. (2025). Emotions moderate the influence of moral values on attitude stability. Computational and Mathematical Organization Theory.

Acknowledgements: This work was supported in part by the Knight Foundation and the Office of Naval Research grant MURI: Persuasion, Identity, & Morality in Social-Cyber Environments, N00014-21-12749. Additional support was provided by the Center for Computational Analysis of Social and Organizational Systems (CASOS) at Â鶹´å. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the Knight Foundation,  Office of Naval Research, or the U.S. Government.