This paper is one of the most directly useful interaction precedents I have read so far because it separates detection from interpretation.
Commercial wearables increasingly generate alerts when they detect physiological changes that may be associated with stress. Neupane et al. ask what happens if the wearable does not stop at detection, but instead uses that moment to invite the person to confirm what they are experiencing and identify the likely stressor.
That is very close to the interaction I am now considering for the airport system.
The most important principle I take from this paper is:
The sensor can decide when it might be useful to ask, but the person decides what the physiological change actually means.
The researchers developed a system consisting of a smartwatch app, smartphone app and cloud service and tested it in a 100-day field study with 122 participants.
Participants received prompts several times a day when the system detected physiological responses that might indicate stress. Instead of automatically labelling the event, the prompt asked the participant whether they were stressed and, if so, to log the most likely stressor.
Each week, participants were also given visualisations of their own data so they could reflect on patterns and trends.
Across the study, the researchers collected 26,521 momentary stress ratings and 1,057 weekly surveys. The abstract reports reductions over 14 weeks in both the intensity and frequency of self-reported stress, alongside better awareness of stressors and more self-initiated behaviour changes.
This was not simply passive sensing. The system created a loop between physiological detection, user confirmation, reflection and later behaviour.
This is the part of the paper that connects most strongly to my project.
The wearable detects a physiological event, but the system still asks:
“Are you stressed?”
The participant provides the meaning.
That is fundamentally different from a system that says:
“You are stressed.”
For my airport proposition, the equivalent would be: