Narrative Frames: A New Approach to Analysing Metaphors in AI Ethics and Policy Discourse

Metaphors fundamentally shape how we reason about complex issues like artificial intelligence, yet current approaches to metaphor analysis in political discourse suffer from inconsistent definitions and methodologies. This paper introduces Narrative Frames, a novel categorisation system that addresses these limitations by providing a standardised framework for identifying and analysing metaphors in AI policy debates. Building on Lakoff and Johnson's conceptual metaphor theory, we derive 49 distinct narrative frames through a two-stage process: inductively coding 685 metaphors from the MetaNet database, then cross-referencing findings with 82 critical metaphor analysis studies. This methodology grounds the typology in both empirical data and established theoretical concepts while resolving definitional ambiguities that have hindered cross-study comparison. The Narrative Frames system offers researchers, journalists, and policymakers a shared vocabulary for analysing how metaphors shape public perception and policy priorities in AI governance. By revealing both the frames present and notably absent in discourse, this approach enables more transparent analysis of underlying assumptions and power dynamics. We discuss limitations and propose future applications, including computational scaling using large language models.