Reading intention, feeling and understanding into a system because it uses language the way people do. It is almost impossible to avoid, since fluent language is our strongest cue for a mind, and it distorts judgement in both directions: over-trusting the output and over-reacting to expressions of preference.
Technical view
Encouraged by design choices including first-person voice, names, stated preferences and expressions of uncertainty or care. Consequences include automation bias, inappropriate disclosure of personal information, and difficulty maintaining accurate mental models of capability. Countermeasures include interface framing, explicit capability statements and avoiding gratuitous personification, balanced against the fact that some anthropomorphic framing genuinely aids usability.
Related
Part of AI Terminology Index, section: Society and Impact.