Beyond Persuasion: The TUNNEL Framework and MESH Mechanisms as Original Contributions to Cognitive Security

Existing cognitive security and influence operations literature addresses artificial intelligence primarily as a tool of persuasion: a means of changing beliefs, opinions, or discrete behaviours. This paper proposes that this framing understates the depth of the threat. What AI-enabled influence operations increasingly target, this paper argues, is not beliefs but identity itself, drawing on the belongingness hypothesis (Baumeister and Leary, 1995), social identity theory (Tajfel and Turner, 1979), identity fusion research (Swann et al., 2009, 2012; Swann, Klein and Gomez, 2024), and evidence on AI companions (Allen, 2025) and Bayesian persuasion (Kamenica and Gentzkow, 2011; Cheng and You, 2025). The distinction matters because identity, once fused with a group or ideology, generates further beliefs and behaviours indefinitely: correcting one belief does nothing to prevent the next, because the next is produced by the same fused identity that produced the last. Surveillance capitalism theory (Zuboff, 2019) supplies the infrastructural context for why this is now possible at scale. The paper makes what are, to the author’s knowledge, three original theoretical contributions. First, the TUNNEL framework, a six-stage dual-use analytical model mapping to Targeting, Understanding the environment, Narrowing attention, Network simulation, Emotional exploitation, and Literalism lock-in. Second, the four MESH mechanisms (Monotropic Capture Mechanism, Social Simulation Mechanism, Emotional Dysregulation Exploitation Mechanism, and Epistemic Literalism Mechanism), which specify how AI does this work at TUNNEL stages three to six, drawing on monotropism theory (Murray, Lesser and Lawson, 2005), algorithmic radicalisation evidence (Ribeiro et al., 2020), emotional vulnerability research (Psychological Defence Research Institute, 2024), and dual-process theory (Tversky and Kahneman, 1974; Kahneman, 2011), extended by a fifth, AI-specific meta-mechanism, the Adaptive Amplification Mechanism (AAM): the capacity of AI systems to continuously test, learn from, and optimise influence against an individual in real time, at a scale and persistence no human recruiter could match (Section 4.5b). Third, the AI-Mediated Identity Fusion Pipeline, describing how the TUNNEL stages combine to produce identity-level change, including memetic amplification as a transmission layer through which personalised narratives become socially replicable, remixable and identity-reinforcing. Why this is different, and not simply a more severe form of behavioural manipulation, is addressed directly: identity manipulation is treated here as a fifth and distinct layer of cognitive attack, above information, narrative, emotional, and behavioural manipulation (Wardle and Derakhshan, 2017), because existing governance frameworks are built to correct beliefs and interrupt behaviours, not to protect the process by which identity itself is formed. The concept of Identity Security is proposed as the organising framework for that governance response. Keywords: TUNNEL framework; MESH mechanisms; MEE mechanisms; Adaptive Amplification Mechanism; cognitive security; identity fusion; AI influence operations; radicalisation theories; System 1 thinking; heuristic bias; neurodivergent vulnerability; four-domain vulnerability model; monotropism; policing AI governance; surveillance capitalism; Bayesian persuasion; memetic warfare; memetic amplification; generative memesis; Identity Security