Synthetic Dialogue in the Grey Zone

Abstract

Large language models (LLMs) are increasingly mediating public conversations through chatbots, search assistants, and content generation tools. In the context of grey zone competition, which involves hostile activity below the threshold of open conflict, LLM systems can be repurposed to amplify cognitive warfare by influencing individual and collective cognition to shape attitudes and behaviors. This article presents a philosophical-theological framework for evaluating “synthetic dialogue”, understood as machine-generated conversational interaction designed to shape beliefs, trust, and social bonds beliefs, trust, and social bonds. Drawing on the NATO-affiliated definition of cognitive warfare, the concept of information disorder, and dialogical philosophy, the paper argues that LLM-enabled persuasion and AI-driven deception (including deepfakes and plausible yet false claims) can erode the conditions of genuine dialogue by undermining epistemic trust and instrumentalizing individuals as targets. The theological-ethical analysis engages with recent Vatican documents on AI, truth, and peace, translating them into normative criteria for evaluating synthetic dialogue in grey zone contexts: truthfulness, transparency, responsibility, proportionality of influence, and respect for human dignity. Methodologically, the study combines conceptual analysis with an empirical probe that evaluates LLM outputs on curated philosophical and theological questions, as well as public-interest scenarios anchored in primary sources. The result is a draft evaluative model and a set of practical recommendations for researchers, platform designers, and religious communities seeking resilience against cognitive attacks.