Navigating Meaning Spaces: A Contextualist Perspective on Conceptual Engineering

Abstract

Conceptual Engineering (CE), the deliberate revision of linguistic meanings to address defects', has emerged as a philosophical project, yet existing projects reveal a fundamental methodological gap: the absence of systematic linguistic analysis. This gap runs deeper than mere theoretical oversight. CE projects reveal a critical need for methods and frameworks from linguistic analysis to ground their interventions in empirical reality. The central research question asks: What insights can linguistics provide for CE? To answer this, the thesis adopts both top-down' theoretical and `bottom-up' empirical approaches to examine meaning construction and change. I first demonstrate that CE requires fundamental reconceptualisation through three interconnected arguments about the collapse of the descriptive/prescriptive divide, the failures of uniform ameliorative approaches, and the limitations of purely lexical-based engineering. This repositions CE as a linguistically grounded enterprise requiring integrated theories of word meaning and attention to propositional content rather than isolated words. The theoretical contribution develops the Partially Inflated Balloon (PIB) Theory of word meaning, which extends existing dual-content theories by adding linguistic-specific structure (L-structure) to referential (E-structure) and cognitive (C-structure) components. I then develop a taxonomy of word types, based on lexical orientation, abstractness and functional differentiation, arguing that CE goals and methods must vary across these categories. Then, motivated by contextualist observations that meaning underdetermination at the word level requires pragmatic enrichment propositionally, I identify dynamic functional propositions as promising for CE applications. The empirical contribution employs BERT word embeddings from the British National Corpus alongside questionnaire data to analyse how meanings emerge from usage patterns and speaker intuitions, uncovering cultural biases, semantic deference patterns, and distributional properties invisible to purely theoretical approaches. I then examine how individual words contribute to conveyed meaning in naturally occurring discourse, using questionnaire data to analyse participant interpretations of utterances and reveal that contextual factors and inferential processes shape meaning beyond individual word analysis. The thesis concludes that successful CE requires methodological integration combining linguistic theory's explanatory power with the descriptive richness of empirical data through a `positively eclectic' approach that combines word embeddings, questionnaires, and theoretical analysis.