Literature Review on Domestic Robotics and Japan's Hikikomori
An introductory course during my bachelor's gave us a broad theme — domestic robotics — and left the framing open. It was a short, time-limited assignment, meant mainly to teach the PRISMA review workflow and the fundamentals of scientific writing. Looking back with more experience, I notice several shortcomings in the paper — the topic however remains important, and this paper still works as a broad overview and entry point into it.
With my minor in human-machine interaction, I wanted to anchor the research question to a concrete social problem rather than study some technical limitations. I chose hikikomori: people in Japan who withdraw from society, often for years, with estimates exceeding one million.
The specific concern is the 8050 problem. Around 613,000 hikikomori aged 40–64 are typically supported by parents now in their 80s. When those parents pass away, individuals who may not have left home in decades can be left unable to care for themselves, sometimes ending in kodokushi, a lonely death often discovered only weeks later.
My research focus was whether domestic robotics — a field Japan leads — could be applied to support hikikomori and help prevent these lonely deaths?
I ran a systematic literature review following PRISMA, searching PubMed and ACM and narrowing 278 records to 27 relevant studies. Direct research on robotics for hikikomori was scarce, so I mapped the question onto a larger, better-studied problem: elderly care robotics in Japan, on the assumption that two isolation-prone populations share relevant conditions such as anxiety, depression, and autism.
Rather than a fix, the result was a set of candidates worth transferring. Beyond conversational and pet robots like Paro, several systems addressed problems hikikomori also face: Action Reproducer targets the reduced mobility of prolonged withdrawal, and Tabby offers low-pressure, ambient presence without demanding interaction. Combined with reports that people may self-disclose more readily to a robot than a clinician, these suggest plausible but untested relevance. Therefore I image future research in this direction to be promising.