The first author has proposed to use Akaike Bayesian Information criterion (ABIC) to adjust relative weight between objective and subjective information in inverse analysis in order to overcome the problem of ill-posedness. The method has been applied to various civil engineering problems in the past and found to be very effective. However, the reason for this effectiveness was not necessarily clearly explained. In this study, an attempt is made to explain the behavior of ABIC from the

more »

... of information entropy. It is found that ABIC chooses the estimates of parameters that maximize the reduction of information entropy from the entropy given at the beginning of the analysis. This fact actually extends the use of ABIC to wider selection of the prior information, i. e. not only choice of prior variance but also alternative prior means can be examined using ABIC as a criterion. The findings are not explained theoretically, but also illustrated using a simple numerical example.