Intent is only one layer.
A system may try to infer what someone wants, but human situations also carry emotion, hesitation, social expectations, history and competing goals. A useful interpretation may require understanding that a person is uncertain, conflicted or simply does not want the system to act.
Nuance often lives between signals.
The words may say one thing while tone, timing and behavior suggest another. A pause can matter. So can distance, repetition, unusual behavior or a change from an individual's baseline. No single signal has a universal meaning.
Emotion should not be treated as a barcode.
Human states are contextual and culturally mediated. Systems that map a facial movement or vocal feature directly to a definitive emotional label can easily overclaim. Sensitivity requires calibrated uncertainty and, often, restraint.
From prompts to situated interaction.
Today's AI often waits for a user to translate a need into text or speech. Human sensing points toward interfaces that can use surrounding context to reduce that translation burden — while making inference visible and preserving user control.
The goal is not mind-reading.
The more credible ambition is situational sensitivity: enough understanding to ask better questions, adapt appropriately, and know when not to infer.