Negotiating emotional support & privacy with humanoid care robots.
A research project asking how patients and family caregivers weigh the comfort a humanoid assistive robot offers against what it sees, hears, and records, in hospitals, rehabilitation centers, and long-term elderly care.
How do patients and family caregivers perceive and negotiate the trade-offs between emotional support and privacy when interacting with humanoid assistive robots in care settings?
The robot that comforts you is also the robot that watches you.
Humanoid robots are entering care facilities as companions and helpers, but they also function like mobile sensor networks that reach into the most intimate spaces of a patient's life. The same expressive presence that reduces loneliness raises a hard question: what is being recorded, who can access it, and how do people stay in control? We mapped the HCI and HRI literature across three intersecting themes to find out.
Expressiveness helps, until it overpromises.
Robots that show emotion through facial expressions, vocal tone, gestures, or empathic dialogue measurably increase comfort, engagement, and even self-disclosure: caregivers share more with a robot they perceive as non-judgmental and consistently attentive. But when a robot's emotional display suggests more understanding than the system can deliver, people describe it as inauthentic or "pretending," and trust drops. Emotional design has to stay honest about what the robot actually is.
"Is every moment of my life being recorded?"
In care settings, privacy is not just data confidentiality; it is physical, emotional, and relational. Most users don't know what a robot records, for how long, or who sees it, and that ambiguity makes people avoid or switch off the device. Studies show users feel most secure when a robot makes its sensor state visibly legible, closing its eyes or turning its back to say "I am not watching." Transparent, customizable data collection isn't a nice-to-have; it's the condition for social acceptance.
Real-time feedback about sensor activation, a robot that visibly closes its eyes, helps users calibrate trust and keep a sense of control.
Trust is calibrated, not automatic.
People tend to over-trust human-like robots even when their capabilities are limited, and in care, that automation bias is risky, because small failures erode confidence fast. Trust holds when the robot states its limits plainly ("I didn't understand, I'm calling your nurse"), keeps the human-robot role division clear, and gives users real control over data and interaction frequency. Boundaries are what let a robot support care without pretending to replace it.
- 01Align emotional design with real capability. Expressiveness builds comfort only when it doesn't overpromise.
- 02Make sensing legible. Users trust what they can see: visible on/off states beat privacy policies.
- 03Give patients dynamic consent. Control over what is stored, and when the robot disengages, preserves autonomy.
- 04Complement care, don't substitute it. Robots should lighten the load on nurses, not overshadow human connection.
Most HRI research to date comes from lab studies or short-term prototype testing, focused on technical performance or a single user group. Far less is known about how robots integrate into the daily routines of real care facilities: how patients, families, and staff negotiate the "rules" for a robot over time. Our research question targets that gap directly: studying these trade-offs in situ, and identifying which design features shape how people manage the tension between emotional support and privacy.