BTW, we explicitly mention the potential risks you refer to in the discussion section.

Open AI‘s GPT 4o

No, we use a standard llm but feed it with information for the user interactions. So, we cannot exclude certain bias, but haven’t observed it in our study.

Our AI communication is embedded in a gamified set-up via avatars speaking to users. We prompted it with specific evidence-based information, so the risk of misinformation spreading is very low.

We hope that the findings will help to improve and tailor climate-related communication for the public. Thanks for the financial support @undp.org & Tryg Fonden (5/5)

💡 Takeaway: One-size-fits-all climate communication doesn’t work. AI isn’t a magic bullet, but it can help reach people who would otherwise disengage from climate information altogether. If we want broader impact, we need smarter, audience-sensitive communication strategies. (4/5)

📊 Findings: Both communication formats can promote behavioral intentions and engagement. However, people who are less curious and less concerned about climate change are more likely to engage when information is delivered through interactive, AI-based formats (e.g., smartphone apps). (3/5)

🔍Aim: Across two preregistered studies, we examined how different communication formats—relying either on information-based or experience-based communication—affect climate-friendly behavior and engagement with climate information. (2/5)

Ich sehe das anders. Anwendungsnahe Forschung inkl. Feldforschung ist sehr populär in der Ökonomie. Und sie machen es oftmals besser als (viele) Psycholog*innen: gute kausale Identifikationsstrategien + formale Prozesstheorien. Wir müssen uns methodisch weiterentwickeln um hier dran zu bleiben.

Here is a summary report of the paper: csl.mpg.de/892188/how-p... (7/7)

Taken together, we present a new toolkit for measuring individual- and group-level social preferences, show that both are shaped by conflict experiences and perceptions, and demonstrate that they predict conflict engagement across diverse samples and group memberships. (6/7)

In a quasi-experimental Study 5 among US participants, we show that parochialism is elevated in high- relative to low-conflict group pairings, whereas altruism appears comparatively less depended on perceived conflict intensity. (5/7)

Study 4 employed a lab-in-the-field approach with members of the Nyangatom, a small-scale society in Ethiopia engaging in cross-border conflicts. We find that higher conflict experience is related to higher levels of altruism, particularly among participants with higher levels of parochialism. (4/7)

In Study 1, we validate our measurement toolkit, showing that altruism and parochialism are separate social preferences. In Studies 2-3, we find that individual-and group-level preferences independently predict participation in real-world conflict (football derby fans and political camps). (3/7)

Classic “parochial altruism” models assume two distinct preferences drive conflict participation:
𝗔𝗹𝘁𝗿𝘂𝗶𝘀𝗺: helping one’s ingroup at a personal cost,
𝗣𝗮𝗿𝗼𝗰𝗵𝗶𝗮𝗹𝗶𝘀𝗺: favoring ingroups over outgroups.
We develop a toolkit for measuring these individual- and group-level social preferences. (2/7)

Thank you all for your interest in our research. We hope this week's posts helped to raise some awareness for AMR and what we can do about it using social and behavioral science. #AMR #WorldAMRAwarenessWeek #WAAW2025 Join us in future research efforts: www.a-bc.network

The paper is available here: doi.org/10.1177/1359... (4)

However, when we induced empathy for future generations in the between-generations condition, antibiotic overuse in case of mild infections dropped. This suggests that how we frame the dilemma—and empathy-based interventions—can help promote more responsible antibiotic use. (3)