Please view the main text area of the page by skipping the main menu.

Japan local gov'ts mull using AI to rate seriousness of bullying cases

An artificial intelligence system designed to assess the seriousness of school bullying cases is pictured in Otsu, Shiga Prefecture, on Sept. 25, 2020. (Kyodo)

TOKYO (Kyodo) -- Nearly 30 local governments across Japan are planning to or interested in introducing an artificial intelligence system designed to assess the seriousness of school bullying cases in hopes of better responding to them, a source close to the matter said Thursday.

    Otsu city government, which came under fire for the way it handled a high-profile bullying case in 2011, has teamed up with information technology services provider Hitachi Systems Ltd., to develop the AI system, which predicts how serious a case of bullying has the potential to become based on an analysis of past cases.

    School bullying has long been a concern in Japan, with the education ministry data showing that elementary, junior and senior high as well as special-needs schools nationwide reported 612,496 cases in the year through March, up 68,563 from a year earlier.

    When a new case of bullying is reported, information on the incident, such as time, place and perpetrator, is fed into the system, which then searches its database to come up with an estimate of how serious the case is, expressed as a percentage. In all, about 50 pieces of data are used for analysis.

    If the result shows that there is a 70 percent or higher risk of the situation worsening, teachers and officials are recommended to intervene urgently and appropriately, according to the developers.

    On top of 20 local governments who have already expressed an interest in the system, seven others are considering introducing it, the source said. The seven are Obihiro in Hokkaido, Saitama in Saitama Prefecture, Matsudo in Chiba Prefecture, Yokkaichi in Mie Prefecture, Takarazuka in Hyogo Prefecture, Tottori in Tottori Prefecture and Nobeoka in Miyazaki Prefecture.

    In the past, the Otsu city education board was not able to properly treat some bullying cases as its officials had to check one by one the hundreds of reports sent to them via email every month.

    But using the AI system, the local government is able to quickly identify the cases with a high risk of becoming serious to encourage teachers, in particular those who do not have much experience, to respond properly.

    For example, a case reported in September involving a female first-grader in elementary school who was ignored by her friends appeared to be a typical case of its kind.

    But after the system evaluated its seriousness at 75 percent, the Otsu city education board decided it needed to be dealt with.

    The city education board was criticized for having failed to find a connection between a bullying case involving a 13-year-old junior high school student and his suicide in 2011. The high-profile case led Japan to enact a law the same year obliging schools to set guidelines to prevent bullying.

    Among high-risk cases are those that happen at a place and time not under the watch of teachers, and cases of bullying that take place on social media, according to the developers of the new system. It is also difficult to identify high-risk cases when the perpetrators involved are both male and female, they said.

    "We can (now) make an objective judgement and find cases which used to be overlooked," thanks to the AI system, said Jiro Hamazaki, an Otsu city education board official.

    Madoka Hiwatashi, a professor at Hyogo University of Teacher Education versed in educational administration, also cited the merits of utilizing the system and not relying only on a teacher's experience.

    "As there is always something in common in bullying cases, (the system) is a useful way not to overlook signs of bullying as it looks at the data objectively," Hiwatashi said.

    Also in The Mainichi

    The Mainichi on social media