Who Gets Addicted to What, and Why Cognitive-Architecture Determinants of Differential Addiction Vulnerability and the Algorithmic Optimization of Compulsive Engagement An Application of Adaptive Compression Advantage Theory (ACAT)

The neuroscience of addiction has made substantial progress through the reward prediction error (RPE) framework (Schultz, Dayan, & Montague, 1997) and the incentive salience model (Berridge & Robinson, 2016). However, these models do not explain differential vulnerability: why specific individuals become addicted to specific substances or behaviors. This paper extends the Adaptive Compression Advantage Theory (ACAT; Murata, 2026) to derive four contributions beyond existing addiction models. First, we propose the α × θ_c vulnerability matrix, which predicts that an individual’s cognitive architecture—characterized by compression efficiency (α) and compression-activation threshold (θ_c)—determines which class of addictive input they are most vulnerable to. Second, we formalize the addictive optimization function of recommendation algorithms as max(PE/compression_cost), demonstrating that short-form video platforms represent a qualitatively new category of addictive technology that achieves maximal reward-circuit activation with zero cognitive growth. Third, we derive why psychedelics have the lowest addiction potential of any psychoactive class: they force genuine model destruction and reconstruction rather than delivering counterfeit reward signals. Fourth, we propose compression-activity therapy as a mechanistically grounded recovery approach, predicting that relapse rates will decrease as a function of access to genuine compression opportunities. Each contribution generates testable predictions that go beyond existing frameworks. Keywords: addiction, cognitive architecture, differential vulnerability, recommendation algorithms, compression, ACAT, psychedelics, ADHD, social media, TikTok, recovery, prediction error, dopamine, behavioral addiction