Attention-Deficit/Hyperactivity Disorder (ADHD) is characterized by a paradoxical cognitive profile that includes both attentional inconsistency and the capacity for intense, sustained focus. This paper introduces a computational model that unifies the core attentional paradox of ADHD, reconciling these contradictory traits as emergent properties of a single, adaptive cognitive strategy driven by working memory (WM) limitations. The model posits that a constrained WM capacity makes parallel processing computationally prohibitive, necessitating a shift to a serial, Depth-First Search (DFS)-like strategy. This serial mode mechanistically accounts for hyperfocus. The model further incorporates a low activation threshold for shifting paths, a single mechanism that produces both a high susceptibility to distraction and a greater propensity for creative insight. This framework provides a unifying architecture for the ADHD cognitive style, framing its traits as the predictable outcomes of a fundamental trade-off. From a Reinforcement Learning (RL) perspective, this can be understood as a cognitive policy biased towards exploration over exploitation, resulting in a reduction in systematic task efficiency in exchange for an increase in opportunistic, creative discovery.
Jeremy Meredith (Sat,) studied this question.