Addiction as Frozen Reconsolidation on the Positive Valence Axis: Rock Bottom as Competitive Co-Encoding in a Persistent Embodied Agent

Al-Kaddah (2026) identifies trauma therapy, cognitive decline, and developmental learning as domains that fall naturally within the SGI framework without requiring new mechanisms. This paper identifies one domain he did not name: addiction. Addiction is not a separate phenomenon requiring separate architecture. It is the symmetric counterpart to trauma, operating on the positive valence axis using the same intensity encoding, the same slow decay rules, and the same reconsolidation dynamics described in Riggleman (2026d) and Riggleman (2026l). Trauma is a high-intensity negative event the agent avoids reconsolidating. Addiction is a high-intensity positive event the agent cannot stop retrieving. Both are produced by the same distance-from-sigma metric. Both resist therapeutic intervention through the same decay-resistance mechanism. This paper formalizes addiction as a positive attractor in three-dimensional affective space, defines dependence as sigma drift toward the addictive state, and identifies rock bottom as a competitive co-encoding event of sufficient negative intensity to attach a counterweight to the addictive memory that fires on retrieval. Relapse is the context-dependent failure of that co-encoding to activate. The geometry predicts all four without new mechanisms. The encoding dynamics independently derived here converge with Sutton and Barto's (1998) temporal difference learning formalism and with Schultz, Dayan, and Montague's (1997) dopamine prediction error work, following the same pattern of independent convergence documented for Russell (1980) in Riggleman (2026m).