<p>In recent years, the transition of public opinion from diversity to polarization — and the resulting social tensions — has attracted growing scholarly interest. Building on social judgment theory in social psychology, this study extends the classic threshold model by proposing a dual-threshold opinion dynamics framework. The model simultaneously incorporates assimilative and repulsive social influences from neighboring opinions within a network, where the strength of each influence is determined by the proportion of neighbors exerting that influence. Agent-based simulations reveal that the interplay between assimilative and repulsive forces generates complex — and at times counterintuitive — dynamics. Notably, increasing assimilative influence does not always reduce opinion divergence; on the contrary, when repulsive influence is weak, it can intensify polarization. Moreover, reducing the density of interpersonal connections can mitigate polarization under strong repulsive influence but fails to promote consensus. Interestingly, when the repulsion threshold is lowand the assimilation threshold is high, an initially polarized opinion distribution proves more effective in curbing bi-polarization than a uniformly random distribution. Although these effects are confined to specific regions of the parameter space, simulation results confirm that the observed patterns are robust across more realistic network structures.</p>