Anthropomorphism evokes social norms that make interrupting or correcting feel rude

What it is

The assistant is speaking in a stream; the user already sees it is about to book the wrong flight. The stop control is there; the hand hovers — like interrupting someone who is talking. Anthropomorphism evokes social norms that make people unwilling to interrupt or correct. The loss happens in the seconds where control is held by politeness.

This is not capability expectation, and not first person implying responsibility. It is a politeness cost.

Why it happens

Nass’s social actor wakes “don’t interrupt, don’t correct to their face, wait until they finish” as a bundle. Streaming output, anthropomorphic tone, a “thinking” animation, all play a person speaking. Correction becomes contradicting someone’s face; stop becomes rudeness. The same stop in a tool interface carries no politeness burden.

The norms also bias toward letting the system finish its sentence. Before it finishes, the wrong flight may already be written, already sent. When the politeness window and the system’s irreversible window overlap, the norm is a harm path. Users who are more vulnerable, more schooled in “don’t be rude,” hold the window longer.

Studying it

The same about-to-err streaming process: one group hears anthropomorphic speech, one sees a neutral progress bar. Measure time to press stop or correct, and whether they press. Independent variables: tone, whether “speaking” is shown, label on the stop control (stop / interrupt it). Dependent variables: time to first intervention, intervention rate, later “felt rude” ratings.

Time to intervene is the core. Only reporting whether they finally got it right will miss acts already sent in those seconds.

Where it stops holding

Professional operators who cut processes daily will have the norms trained down. In text chat, “interrupt” is socially weaker than in voice or an anthropomorphic video. Making stop a system-facing “cancel task” loosens the norm. Claims to understand, remember, or be sorry are an inner-state issue that can strengthen the norm, but are not the norm itself.

Applying it

• Labels on stop and correct use verbs toward the system: “cancel generation,” “edit this sentence,” not “sorry to interrupt.”
• During a stream, put stop in a place more salient than the text; once an error can be judged, allow a change before the sentence ends.
• High-stakes writes must not wait for the anthropomorphic voice to finish. Speech can break; writes must be stoppable.
• Check: in a replay of an about-to-err, does the anthropomorphic group press stop clearly later, and do they mention “felt bad.” Later, and the norm is already holding control.

Related

• Same group: L5.12.1 Name, avatar, and tone raise expectations of both ability and reliability · L5.12.3 Claims to understand, remember, or be sorry imply inner states the system does not have · L5.12.4 Anthropomorphism raises early acceptance and amplifies disappointment at failure · L5.12.5 Vulnerable users more readily take anthropomorphic language at face value, which creates an extra duty of care
• Nearby: L5.06 Risks of Anthropomorphism · L4.05 Interruptibility and Rollback · L3.11 Streaming the Generation Process
• Search terms: social norms · interrupting an agent · politeness cost