Inferential risk in Nigerian public health communication: coded dataset and codebook

This dataset tests the hypothesis that Nigerian public health institutions manage anticipated interpretive trouble prospectively rather than only reactively — extending Elder and Haugh's (2023) interactional concept of unwanted inference to written, non-sequential institutional communication, where audiences give no observable uptake for institutions to repair. The data was gathered by close, iterative coding of five public-facing texts from three Nigerian federal health agencies (a NACA HIV/AIDS FAQ, an NPHCDA COVID-19 vaccine FAQ, and NCDC advisories on Ebola virus disease, Human Metapneumovirus and Lassa fever), following a six-stage sequence: identifying each available inferable, its institutional status, the inferential risk attached to it, the management strategy used, the preferred inference established, and the behavioural action that preferred inference supports.

The data shows that of 68 inferables coded across 28 inferential episodes, 28 were institutionally preferred, 12 institutionally neutral, and 28 institutionally unwanted — meaning institutions do not treat all available meanings as problematic, but selectively intervene on a specific subset. The 28 unwanted inferables cluster into six recurring risk types (delayed or unsafe action, 8 episodes; stigma/exclusion and distrust/refusal, 6 each; panic/alarm, 5; complacency, 4; distorted responsibility, 1), each with the specific episode IDs that instantiate it.