Moderating online content is essential to ensure that digital communication is safe and enjoyable. Although automation is necessary to manage the substantial volume of content on large digital platforms, human expertise is deemed indispensable in decision-making processes, as well as in the design, training, and implementation of automated solutions. The “human-in-the-loop” (Hilo) concept is a key approach to integrating humans into (semi-)automated decision-making systems. However, its definition and implementation vary considerably across academic disciplines. Furthermore, the factors influencing the quality of such hybrid decision-making processes are not sufficiently well understood. This integrative literature review builds on previous conceptual work and identifies 14 factors that influence meaningful human involvement in (semi-)automated decision-making in content moderation. The findings provide an empirically grounded basis for evaluating existing and planned Hilo implementations. They will also support the design of socio-technical systems in which humans improve automated decision-making and will inform discussions on effective Hilo requirements. To this end, the review offers actionable insights to help platforms and policymakers enhance human–machine collaboration and decision-making quality.
