Abductive reasoning—the search for plausible explanations—has long been central to human inquiry, from forensics to medicine and scientific discovery. Yet formal approaches in AI have largely reduced abduction to eliminative search: hypotheses are treated as mutually exclusive, evaluated against consistency constraints or probability updates, and pruned until a single “best” explanation remains. This reductionist framing fails on two critical fronts. First, it overlooks how human reasoners naturally sustain multiple explanatory lines in suspension, navigate contradictions, and generate novel syntheses. Second, when applied to complex investigations in legal or scientific domains, it forces destructive competition between hypotheses that later prove compatible or even synergistic, as demonstrated by historical cases in physics, astronomy, and geology. This paper introduces quantum abduction, a non-classical paradigm that models hypotheses in superposition, allowing them to interfere constructively or destructively, and collapses only when coherence with evidence is reached. Grounded in quantum cognition and implemented with modern NLP embeddings and generative AI, the framework supports dynamic synthesis rather than premature elimination. For immediate decisions, it models expert cognitive processes; for extended investigations, it transforms competition into “co-opetition” where competing hypotheses strengthen each other. Case studies span historical mysteries (Ludwig II of Bavaria, the “Monster of Florence”), literary demonstrations (Murder on the Orient Express), medical diagnosis, and scientific theory change. Across these domains, quantum abduction proves more faithful to the constructive and multifaceted nature of human reasoning, while offering a pathway toward expressive and transparent AI reasoning systems.
Remo Pareschi (Thu,) studied this question.