Prototype home sensors could flag loneliness in older adults

Loneliness is hard to see from the outside. A clinic visit lasts minutes, and older adults often don't mention feeling alone, so the mental and physical health risks that come with it are easy to miss. Researchers from Virginia Commonwealth University (VCU) and Emory University want the home itself to raise the alarm: a prototype called HomePal that watches everyday life inside the house and reports what it sees to a caregiving team.

HomePal is a passive home sensing system, meaning it asks nothing of the resident: no app to open, no button to press. It combines commercially available and custom sensors that measure movement, daily activities, sleep and phone use. Motion and door sensors detect movement inside the home and may flag social withdrawal. Smart plugs on the television estimate screen time, a often solitary activity. Sleep sensors track rest, since poor sleep can point to symptoms of anxiety or depression. Smart speakers are used to detect loneliness-related vocal patterns, and phone call patterns measure social connectedness.

The team deployed the prototype for three months in the homes of 70 adults aged 65 and older living independently in metro Atlanta. The sensor data is being compared with biweekly loneliness assessments led by Chung, a former VCU School of Nursing professor. An AI model is being used to determine which sensed behaviors are most strongly associated with loneliness, and how those patterns match the assessment results. The researchers say accuracy has yet to be checked against the traditional survey approach, and no such comparison is reported yet. Kamel, the computer science Ph.D. student who leads the study, is careful about what it claims: “We are not going to say we are diagnosing loneliness directly.”

What it could change is a daily loneliness score or trend that family members or medical caregivers could use to guide care, and that flags worrying shifts over time. The researchers ultimately hope to build a commercial product that could be installed in any home, much like a security system. No date for that is given. For now, the users who stand to gain are the people the study follows and the teams that look after them, and the study's own goal is more modest: to find out which patterns “may deserve a follow-up.”

The privacy question is built into the design. HomePal could be modified to use only sensors that operate and store data on local networks, and the researchers say residents would decide which data is shared with caregivers. Until accuracy is measured on real households over a longer period, HomePal remains a promising prototype rather than a tool anyone can rely on.