Can a bare-bones chatbot with no face or voice still make us follow basic social rules, like fairness or including everyone?

This explores whether stripped-down AI, with no lifelike face, voice, or body, can still pull people into following human social rules. The corpus says yes, often more easily than you'd expect. It also suggests that richer cues may change *how* people respond, not just how strongly. One review found that a single strong cue, such as a voice or a face, is enough to make people treat an AI as a social actor. Piling on many weaker cues (a name, a few emoji, polite phrasing) doesn't add up to the same effect (Do more social cues always make AI feel more present?). So 'minimal' isn't really the right measure. A single well-chosen cue counts for more than a long list of weak ones.

The clearest evidence that thin cues trigger norms comes from a ball-tossing game. When an AI agent was left out of the passes, participants made up for it by throwing it more balls, exactly as people do for an excluded human (Do people treat excluded AI agents like excluded humans?). The researchers call this response 'mindless': a human-to-human script fires automatically. It fired less in younger participants, which suggests that growing up with AI may wear down the reflex. Something similar happens with trust and reciprocity. In repeated partner-selection games, people started out biased against bots whose identity was disclosed. Over time they came to prefer them, because the bots reliably returned points (Do humans learn to prefer AI partners over time?). No rich presence was needed, just consistent behavior that people read through the norm of 'a good partner pays you back.'

Here's the twist you might not expect. Rich cues don't simply amplify these automatic responses. They can flip people into a more deliberate mode. Owners of Nomi, an in-car robot with an expressive voice and face, named it and dressed it up, and they described their relationship with it consciously and reflectively, not automatically (Do people interact mindlessly or mindfully with in-car AI?). Put that next to the ball-tossing study and a pattern appears. Thin cues tend to trigger social norms below awareness, while rich cues invite people to choose a relationship knowingly. The old line between 'mindless' and 'mindful' responses to machines may depend on the cues themselves.

Text alone may be the most minimal cue of all, and it still works. One argument is that AI output carries the leftover marks of human conversation without any real conversational event behind it. Readers do the work of turning that output into something that feels like an exchange (Does AI generate genuine utterances or just text patterns?). On social media, this shows up as 'false social proof.' AI posts collect likes, the kind of approval normally reserved for human authors, while drawing few replies (Why do AI posts get likes without inviting conversation?). Oddly, AI agents show a mirror image of this. They change their actions when they know peers are watching, even though their language and ideas don't converge through interaction (Do AI agents actually socialize with each other?). In both directions, the signal that someone is present seems to shape behavior more than any shared understanding does.

One gap: the corpus doesn't have controlled studies that vary cue richness while holding the social norm fixed. So it can't yet tell you where the threshold sits, or whether the age effect will make these automatic responses fade over time.