When you say 'thank you' to a chatbot without thinking, is that different from actually deciding how to treat it?

This asks whether people react to AI agents in two different ways: a fast, automatic social reflex and a slower, deliberate one. It also asks whether those two reactions pull apart. The corpus says yes, and it adds something less expected: researchers have often blended the two into one idea, and the line between them may be wearing away. Do people respond to machines mindlessly or reflectively? separates two traditions that usually travel together under the 'Computers Are Social Actors' banner. The Media Equation holds that minimal cues set off mindless 'as-if' social behavior: you say 'thank you' to the voice assistant without deciding to. Media Evocation holds that machines make us think precisely because they don't fit neatly into the categories of person or thing. The first is a reflex. The second is a puzzle people actively work on.

The automatic side shows up clearly in experiments. In Do people treat excluded AI agents like excluded humans?, people passed a virtual ball more often to an AI agent that others were leaving out. That is the same compensating move people make for an excluded human, and nobody stopped to ask whether a bot can feel left out. The study also found that younger participants did this less. That hints that the reflex isn't fixed: familiarity with AI may weaken how readily human social scripts transfer to machines. Do more social cues always make AI feel more present? adds that a single strong cue, such as a voice or a face, is enough to make an AI feel like a social presence. A pile of weaker cues is not.

The reflective side appears in a field study. Do people interact mindlessly or mindfully with in-car AI? interviewed owners of an in-car AI companion. They named it, dressed it up, and talked about it as having social agency. They did this consciously, not by accident. The same high-quality voice and facial cues that should trigger mindless responses instead invited deliberate relationship-building. So the two responses don't just diverge. Strong cues can push people out of autopilot and into reflection. Over time, reflection can also override an initial bias: in Do humans learn to prefer AI partners over time?, people first avoided partners labeled as bots, then learned to prefer them once the bots proved more reliable.

A related angle: Does AI generate genuine utterances or just text patterns? argues that AI text has the surface markers of communication but no one actually 'saying' it, so the human quietly does the interpretive work of turning it into a conversation. That labor can be automatic or deliberate, which is the same split again. Do humans and LLMs differ fundamentally or just superficially? offers a parallel lens. Seen from outside, humans and LLMs are completely different kinds of thing. From inside a conversation, they draw on the same shared language, so the difference feels small. These views line up roughly with reflective and automatic responses.

There is a design payoff. If automatic responses are the default, designers can deliberately prompt reflection. Do reflection questions help people make better decisions with AI? found that AI assistants that asked people reflective questions alongside giving advice led to better decisions than assistants that only advised. The gap between reflex and reflection isn't only something to observe. It's something interfaces can steer.