This study investigates how human mobility has changed in the long-term in response to the COVID-19-related information in Japan. We use publicly available data from Google on human mobility in retail & recreation and residential spent time. These variables can be explained using daily data on the number of infected cases, whether the state of the emergency is declared or not, and the cumulated number of the vaccinated person. In the regression analysis, we use the 'interactive effects model' to control complicated unobservable factors that vary across time and cross-sectional dimensions. Our regression results find that people feared an unknown virus in the 1st wave, but the habituation trend for human mobility is noticed for the repeated similar infection information. However, from a different kind of information about the spread of new variants, people's habituation comes to a halt to some extent. Further, the spatial interaction of infection information is observed. We also show that people reacted appropriately to infection information even without a state of emergency declaration. Also, vaccination promotion encourages people to go out with security. When implementing policies to control human mobility, it is essential to consider the timing, and the degree of information penetration, carefully.