Do people treat AI chatbots with brand-new habits, or just run the same manners they'd use on a human, without thinking?

This asks how researchers can tell whether people treat AI with interaction habits built specifically for machines, or are simply running their usual human social rules on autopilot. The surprise is that both explanations involve mindlessness. The question isn't whether people think carefully about AI. They mostly don't. What matters is which automatic script fires. The original 'Computers Are Social Actors' (CASA) research said people reflexively reuse human-to-human politeness, reciprocity and flattery norms with machines. The extended version argues that people also build a second, machine-specific script through repeated use, and that this script runs just as automatically Do humans apply human-human scripts to AI interactions?.

The clearest sign is change over time. If people were only reusing human rules, their behavior toward an AI should look the same on day one and on day fifty, because the human script is already fully formed. Longitudinal studies instead find that responses shift steadily with repeated AI interaction, which suggests a new script is being learned rather than an old one borrowed. A second sign is divergence: if a habit doesn't match any human-to-human norm, borrowing can't explain it. Over time, those two signs together separate 'treating it like a person' from 'treating it like the kind of thing it is.'

User perception research points the same way. When people describe how they mentally model a dialogue agent, perceived competence explains about half the variation in their impressions. Human-likeness is a separate and smaller factor, and communicative flexibility is a third How do users mentally model dialogue agent partners?. If people were simply applying human social rules, human-likeness should dominate. Because competence comes first, people seem to judge agents partly against standards they don't use for other people, and that is what a media-specific script looks like.

A lateral angle makes the distinction murkier. Shanahan argues that dialogue agents are best understood as characters played by the model. Folk psychology applies to the character, not to the system underneath Should we treat dialogue agents as role-playing characters?. So when people respond socially to an AI, they may be responding correctly to a performed persona, which is neither mindless rule-following nor a new machine script. Conversation-analysis work also shows that human dialogue structures, such as pausing to ask a clarifying question, can be deliberately designed into agents When should AI agents ask users instead of just searching?. When an agent is built to behave like a human conversation partner, applying human rules to it is no longer a mistake.

The library is thin here. Only one note addresses the media-agent-script versus CASA distinction directly. None describes a specific experimental design for separating the two beyond the longitudinal-change evidence. Treat the competence-first and role-play angles as nearby supporting material, not settled answers.