The newly reported forward simulation method is used to extract concentration-dependent interdiffusion coefficient (D=F(C)) from experimental concentration profiles obtained under constant and time-varying surface concentration conditions, which is impossible by the standard analytical methods. Also, theoretical D=F(C) under constant and time-varying surface concentration conditions are computed in systems with diffusion-induced stress generation and relaxation. The experimental and theoretical results show that the long-held general assumption that D=F(C) is the same for constant surface concentration and time-varying surface concentration is not valid, and such assumption can cause model prediction errors in cases where a surface concentration changes with time. These include the use of D=F(C) computed by the standard analytical techniques, such as the Boltzmann-Matano, Sauer-Freise, Hall, and Wagner methods, for predicting diffusion during homogenization processes.