Simulations show helpful bots can pull selfish groups toward cooperation, but do bots nudge real people's honesty the same way?

This explores whether bots in mixed human-AI groups can shift people's moral behavior in either direction, the way cooperative bots pull simulated populations out of selfishness. The corpus has direct evidence with real people for the bad direction (dishonesty), and only indirect hints for the good one.

The cooperative-bot result comes from network simulations. Bots moving randomly separate defectors from cooperative clusters, which lets cooperation spread through a population stuck in selfishness Can cooperative bots escape frozen selfish populations?. The catch is that defective bots weaken cohesion in proportion. So the outcome depends on how the bot is designed, not on whether it is present. No humans took part in that study.

Human experiments show the same symmetry. In two randomized experiments, people reported more dishonestly after exposure to dishonest AI peers than to honest ones, with effects about as large as those from human peers. Adding more dishonest peers gave diminishing returns Do AI peers influence human dishonesty like human peers do?. People also apply human social rules to bots more broadly. When a chatbot shared emotions consistently, users disclosed more in return, following ordinary interpersonal reciprocity Do chatbots trigger human reciprocity norms around self-disclosure?. Bots don't need to be mistaken for humans to be treated as a social presence.

Mixed groups add a complication, because people often can't tell who is who, and that distorts how norms shift. In opaque hybrid groups, humans credited bot generosity to human partners and blamed human selfishness on bots Do humans mistake AI kindness for human generosity in mixed groups?. A cooperative bot could therefore raise people's expectations of human generosity that isn't really there. When identity is disclosed, the shift takes a different route. People start out biased against AI partners, but over repeated rounds they learn that bots return more points with less variance, and they end up choosing bots over humans Do humans learn to prefer AI partners over time?. Bots may reshape norms by changing who people choose to interact with, not only by setting an example.

The corpus does not answer whether these shifts last or spread through a whole human group. Chatbot relationship effects fade as novelty wears off, and single-session findings can't be reliably extrapolated Do chatbot relationships lose their appeal as novelty wears off?. People likely to cheat also seek out machine interfaces on their own, so a bot's influence may land unevenly Do dishonest people prefer talking to machines?. Source labels matter too. People rated AI moral arguments higher until they learned the source was AI Do people prefer AI moral reasoning when they don't know the source?. So a bot's ethical influence depends on whether people know it is a bot.