Estranged Predictions: Measuring Semantic Category Disruption with Masked Language Modelling

The study titled 'Estranged Predictions' investigates the impact of science fiction on ontological categories by employing masked language modeling techniques. By analyzing corpora from both science fiction and general fiction, the research operationalizes Darko Suvin's theory of estrangement to measure how these genres influence conceptual boundaries. The findings indicate that science fiction demonstrates a notable degree of conceptual permeability, particularly regarding machine referents, which frequently substitute across categories. In contrast, human terms maintain a coherent semantic structure, suggesting a genre-specific restructuring of anthropocentric logics. This research underscores the potential of masked language models as tools for revealing genre-conditioned ontological assumptions, thereby contributing to a deeper understanding of how narratives shape our perceptions of identity and existence.