Abstract
Generative artificial intelligence (AI) has entered higher education faster than institutional policy can regulate it, a gap visible in Philippine humanities programs, where interpretive originality and authorial voice are central to disciplinary practice. Humanities programs in regional, resource-constrained Philippine settings remain largely absent from the AI-adoption literature, a gap this study addressed by examining AI adoption, perceived usefulness and ease of use, perceived risk, AI literacy, and behavioral intention among humanities faculty and undergraduate students at a state university in Zamboanga City, Philippines, where no institutional AI policy exists. A convergent mixed-methods design combined survey data from 200 students and 23 faculty with ten in-depth interviews; quantitative data were analyzed using descriptive statistics, Pearson correlation, multiple regression, and Welch's t-test, while qualitative data underwent reflexive thematic analysis, integrated through joint display analysis. Students used AI moderately and selectively, favoring comprehension and organization over original composition. Faculty reported higher AI familiarity than students but markedly lower classroom use (M = 3.59 versus M = 2.67), a pattern proposed here as epistemological friction: principled disciplinary resistance rather than technical inability. Both groups perceived high risk, yet risk predicted only the intention to limit AI use (r = .531, p < .001), not the intention to discontinue it (r = .047, ns), producing concealed adoption tentatively labeled a shadow AI ecosystem. Perceived usefulness (β = .427) and ease of use (β = .365) were the strongest predictors of continued use, while students significantly outperformed faculty in AI literacy (d = -0.943), reversing the assumed top-down transmission of competence. AI acceptance here was shaped as much by disciplinary values and unclear institutional guidance as by usefulness or ease of use, supporting tiered governance, faculty development addressing disciplinary and ethical concerns, and structured reverse mentoring to narrow the literacy gap and reduce concealed AI use.