Plural Logics and Artificial Intelligence: A Neutrosophic Approach to Causal Analysis

This article proposes a framework of logical pluralism applied to causal analysis in artificial intelligence (AI) and social sciences. Grounded in the pluralist thesis that there is no single monolithic concept of causation, but rather a family of related concepts, we confront the Humean premise that “we do not observe causation’ with Anscombe’s counter-argument that we do, in fact, perceive causal actions such as pushing, striking, and cutting. Building on this foundation, we present an illustrative study comparing crisp-set Qualitative Comparative Analysis (csQCA), fuzzy-set QCA (fsQCA), and neutrosophic QCA (nQCA). From this comparison, we derive both the methodological and substantive implications. Methodologically, we find that: (i) strict dichotomization (csQCA) can underestimate causal relationships; (ii) graded membership (fsQCA) more effectively captures empirical strength and relevance; and (iii) neutrosophic decomposition (nQCA) explicitly accounts for truth, indeterminacy, and falsity, thereby offering a pluralistic diagnosis of the causal relationship. We further enrich this analysis by incorporating a pluralistic view of causality, encompassing mechanical production, counterfactual difference, capacities/dispositions, and mechanisms.