I’m interested in this area; and the broader notion of machines “understanding” people.
As an experiment, I’m trialling periodic prompts of a PHQ-9 to capture aspects of my mood. I will likely expand this to a few other openly licenced, validated methodologies with relatively short self assessment. It is probably a bit too clinical feeling for day to day use.
So far, there’s been a number of projects that try to capture more about people’s states, but focus a lot on EEG, eye tracking, etc - tech that is a bit intrusive outside of a lab setting; and most importantly treat the research subjects as only subjects (no agency or little agency discussed).
For actually coupling current emotional-state/mood/subjective experience to actionable things; I am planning on modelling more immediate term states + the triggers for those states a bit.
Think discomfort due to room being outside of standard temperatures for person X, suggestion actions in room to resolve it; modelling circadian rhythm / sleep tracking to guess person Y may be tired; or even a neglected TODO list (average age of items trending up for multiple days in a row?) while someone is at home as an indicator they don’t feel motivated.
Most of the actions you can do with home assistant to actually “fix” these things are really just notifications and requests for human engagement. IE automated medication reminders.