Viral reach or behavioral change? A dissemination–conversion tradeoff in social campaigns for stigmatized conditions
Joint work with Alexandra Avdeenko (World Bank), Luc Behaghel (Paris School of Economics, INRAE), Andreas Ette (Federal Institute for Population Research), Yagan Hazard (Collegio Alberto), Alexander Moldavski (ZI Mannheim), Andreas Meyer-Lindenberg (ZI Mannheim), Nicolas Rüsch (Ulm University), and C. Katharina Spiess (Federal Institute for Population Research).
Vulnerable populations with stigmatized conditions rarely seek help despite pressing need. Social campaigns aim to bridge this gap, yet their design involves a potential tradeoff: light-touch messages that diffuse easily may lack the depth to motivate behavioral change, while more complex and persuasive messages may resist peer-to-peer sharing. We formalize this tradeoff and test it in a randomized experiment promoting mental health care uptake among 1,993 Ukrainian refugees in Germany using a purpose-built bot on Telegram. We varied the messenger (celebrity endorsers vs. a peer patient), the message format (video or plain text), and whether peer-to-peer sharing was actively encouraged. Help-seeking was measured by calls to a mental health hotline.
The patient video induced 7.2% of recipients to call the hotline — nearly double the rate of the plain text (4.4%, P=0.03) and the celebrity video (4.8%, P=0.06) — yet recipients were half as likely to share it (6.8%) compared to the plain text (12.4%, P<0.001). A structural model allows us to quantify the dissemination–conversion tradeoff. No single approach significantly outperforms the others overall, but the patient video tends to disproportionately induce calls by users most in need–those with moderate to high mental health symptom burden. Across all experimental conditions, dissemination remained limited: the social multiplier reached at most 0.5, with the highest level observed in the plain-text condition. Actively encouraging recipients to spread the message did not increase sharing but reduced help-seeking among users with minimal to mild symptom burden.
These findings extend prior survey and lab evidence to a real-world field setting, confirming that users with significant mental health symptom severity react most favorably to help-seeking messages conveyed through patient testimonials. They also underscore the limits of horizontal peer-to-peer dissemination: active encouragement to share videos, or the use of a simplified text message, fell far short of inducing viral spread. Our fieldwork further suggests a possible way forward: relying on trusted intermediaries–Telegram group administrators–to facilitate the dissemination of sensitive, complex message content.
