The role of Generative AI in localizing development communication campaigns: Efficiency vs. ethical risk

The fundamental success of Development Communication (DevComm) rests on the ethical imperatives of co- creation, trust, and deep cultural participation to foster sustainable social change in the Global South (Melkote & Steeves, 2015). The complexity of cross-cultural message adaptation has historically constrained DevComm initiatives; a constraint seemingly resolved by the arrival of Generative Artificial Intelligence (GenAI) and Large Language Models (LLMs). These tools promise extraordinary efficiency and scalability in message localization, yet this pursuit of speed introduces a critical ethical dilemma, as the technology functions as an Algorithmic Mediator built upon globally unrepresentative, often Western-centric, datasets. This theoretical paper argues that this Mediator risks eroding the co-creation mandate by systematically propagating algorithmic bias, resulting in miscontextualized, culturally inauthentic, and potentially harmful messages. Applying the Participatory Communication Model and critiquing global AI ethics frameworks, the study develops a Trust Erosion Model that charts the causal path from unchecked algorithmic efficiency to catastrophic campaign failure. We conclude by proposing a comprehensive Participatory Governance Framework that places the local human validator at the heart of the localization process, ensuring that the technology serves, rather than subsumes, the ethical demands of genuine, participatory development.