Weak lensing (WL) effects arise from the gravitational light deflection by large-scale structures, and thus are powerful tools to probe the nature of the dark universe. Their signals are dominantly from relatively small-scale nonlinear structures, and therefore possess significant non-Gaussianity. Statistical analyses of WL effects thus need to be multi approaches. In this presentation, motivated by the recent machine learning results, I will talk about the use of WL steepness as a cosmological probe by presenting our theoretical model and the comparisons with numerical simulations. The applicability of WL steepness statistics will also be discussed.