Research shows that policy tools to reduce residential segregation are also effective to reduce economic inequality and, for decades, achieving the adequate level of 'social mix' has been an important policy goal in housing and community policy fields. Yet whether social mix creates social interactions, thus producing favourable social outcomes, is still disputed. This study aims to expand the empirical record on the modus operandi of social mix and its policy viability, and to provide useful information on relevant contextual factors to prescribe effective social mix policies to improve social interaction. We fit interactive linear models to data from Los Angeles County in California, USA, to test the hypothesis that social mix (measured via residential income mix) promotes social interactions (in the form of civic, daily-life, and political activities and group memberships), and to examine whether arts and recreation centres (i.e. the places where social interactions occur to produce social mix) moderate the effect of residential income mix on favourable social outcomes. Results show that while residential income mix is negatively associated with social interactions, recreation centres help income-mixed communities increase social interactions among their residents. We also find evidence that communities with higher income mix levels manifest higher social interaction as there are more arts and recreation centres in their communities. Our findings shed light on why social mix policy works in some communities but does not work in others, elucidating why previous research displays conflicting claims.