The comprehension of the nonlinear effects provided by mixed alkali effect (MAE) in oxide glasses is useful to optimize glass compositions to achieve specific properties that depend on the mobility of ions, such as the chemical durability, glass transition temperature, viscosity and ionic conductivity. Although molecular dynamics (MD) simulations have already been applied to investigate the MAE on silicates, less effort has been devoted to study such phenomenon in mixed alkali aluminosilicate glasses where alkali cations can act both as modifiers, forming non-bridging oxygens and percolation channels, and as charge compensator of the AlO4- units present in the network. Moreover, the ionic conductivity has not been computed yet; thus, the accuracy of the atomistic simulations in reproducing the MAE on the property is still open to question. In this work, we have validated five major interatomic potentials for the classical MD simulations by modelling the structure, density, glass transition temperature and ionic conductivity for three aluminosilicate glasses, (25 - x)Na2O - x(K2O) - 10(Al2O3) - 65(SiO2) (x = 0, 12.5, 25). It was observed that only the core-shell (CS) polarizable force field well reproduces the experimentally measured MAE on Tg and the ionic conductivity as well as the higher conductivity of single sodium aluminosilicate glass at low temperature and the higher conductivity of single potassium aluminosilicate glass at high temperature. The MAE is related to the suppression of jump events of the alkaline ions between dissimilar sites in the percolation channels consisting of both sodium and potassium ions as in the case of alkaline silicates. The superior reproducibility of the CS potential is originated from the larger and the flexible ring structures due to the smaller Si-O-Si inter-tetrahedra angle, creating appropriate percolation channels for ion conductivity. We also report detailed assessments for using the potential models including the CS potential for investigating MAE on aluminosilicates.