.png)
POST-DOC
PHILIP PARNAMETS - NYU Social Psychology Postdoctoral Fellow
Philip completed his PhD in cognitive science at Lund University and has spent two years as a postdoc with the Emotion Lab at Karolinska Institutet. His research is grounded in a broad interest in the cognitive and computational mechanisms underlying preference change, decision making and learning, especially in the moral domain. In the Social Identity and Morality lab his work focuses on dynamic models of social learning about moral agents and of moral choices generally. Philip spends his spare time creating, listening or dancing to electronic music.
At NYU’s Center for Conflict and Cooperation, we conduct cutting-edge research to understand what divides and what unites us. By collaborating with scientists worldwide, as well as policymakers, organizations, and civil society, we educate the public and translate our findings into real-world solutions for issues like misinformation, polarization, intergroup hate, extremism, climate change, and technology.
The Psychology of Virality. [pdf]
-
Social Media Amplifies Negative and Moral Content. Negative, high-arousal, and moralized information tends to go viral online. However, this content is often not widely liked, leading to a phenomenon called ‘the paradox of virality'.
-
Information Spreads Like a Virus. The ‘information as a virus’ metaphor is often used to describe the process of information spread—e.g., some types of information are more contagious than others, and ‘superspreaders’ contribute to most (mis)information spread.
-
Social Media's Design Shapes How Information Spreads. Social media platforms have various structural features, such as their algorithms, design features (reshare, like/dislike buttons, etc.), and social norms. These features interact with underlying psychological processes (e.g., preferential attention to negativity, social motivations) to drive the spread of information online.
Updating the identity-based model of belief: From false belief to the spread of misinformation. [pdf]
-
Accuracy and Identity Goals Shape Beliefs. People's political beliefs and information processing are shaped by a push and pull between two competing forces: the desire to form accurate beliefs (accuracy goals) and the need to belong and uphold a positive image of the ingroup (identity goals).
-
Identity Can Override Accuracy. When people prioritize social ties and belongingness to their party over accuracy, they align their beliefs with other party members rather than facts. People might believe or share false information simply because it reinforces their positive beliefs about their party, or they want to signal their party loyalty.
-
Exposure and Selection Influence Beliefs. Partisan differences also arise from differences in information diet. People actively seek out and prioritize content that reinforces a positive group image.
Addressing climate change with behavioral science: A global intervention tournament in 63 countries. [pdf]
-
Different Interventions Have Small But Significant Effects Globally. We compared 11 interventions on various climate mitigation outcomes from 63 countries. Overall, the different interventions had small but significant effects globally.
-
No Intervention Offers Silver Bullet Solution. Different strategies worked for different outcomes (e.g., belief in climate change), but no interventions increased effortful behavior towards fighting climate change via the tree planting.
-
The Effectiveness Of Intervention Varies Across Audiences. The effects of each intervention differed depending on people’s initial climate beliefs.
GPT is an effective tool for multilingual psychological text analysis [pdf]
-
GPT Accurately Identifies Psychological Constructs Across Languages. Across 15 datasets and 12 languages, GPT models agreed well with humans in detecting psychological constructs like sentiment, emotions, offensiveness, and moral foundations.
-
GPT is a Simple, Training-Free Method for Text Analysis. Text analysis through GPT was possible using simple prompts, little coding, and no additional training data, making it a practical method for researchers analyzing psychological text data.
-
GPT’s Multilingual Capabilities Extend to Lesser-Spoken Languages. GPT- 3.5 Turbo, GPT-4, and GPT-4 Turbo performed well across widely spoken languages (English, Arabic, Indonesian) and lesser-spoken African languages (e.g., Amharic, Yoruba, Kinyarwanda, Tsonga), with notable accuracy gains for lesser-spoken languages across successive models.
Polarization
Intergroup conflict, political violence, outgroup hate, and ideological extremism.
Morality
Prosocial behavior, altruism, value systems, and moral outrage.
Misinformation
Belief in and spread of misinformation and conspiracy theories.
Bias | Discrimination
Mitigating explicit and implicit bias, discrimination, and empathy gaps.
Social Media | Technology
The role of social media and technologies like generative A.I. in misinformation and polarization.
Cooperation | Cohesion
Cooperation, civic discourse, social bonds, and positive collective action.
Are a few people ruining the internet for the rest of us?
Jay Van Bavel
July 2025, The Guardian
Beyond doom and gloom, here's how to stimulate climate action.
Madalina Vlasceanu & Jay Van Bavel
February 2024, Scientific American
Why we click on stuff we know we won't like.
Steve Rathje & Jay Van Bavel
December 2023, The Boston Globe
... and get monthly updates about the center's latest work and events!