CRITICAL THINKING AND FORMAL LOGIC IN ARTIFICIAL INTELLIGENCE SYSTEMS: A PHILOSOPHICAL ANALYSIS

Authors

DOI:

https://doi.org/10.54891/2786-7013/2025-2-4

Keywords:

critical thinking, formal logic, neuro-symbolic AI, AI hallucination, AI explainability, AI epistemology.

Abstract

The article provides a comprehensive philosophical account of critical thinking and formal logic within contemporary artificial intelligence (AI) systems against the backdrop of the pervasive «hallucination» phenomenon–confidently produced yet factually incorrect outputs. It argues that purely statistical language models, lacking intrinsic verification mechanisms, cannot deliver epistemic reliability and therefore require «disciplining» via logical and metacognitive components. Methodologically, the study combines a history-of-philosophy perspective–from Aristotle, Descartes, Kant and Popper to current debates on the semantic limits of syntactic procedures (e.g., Searle’s «Chinese Room») – with a survey of technical strategies for embedding logic in AI. The paper traces the pendulum swing from symbolic (GOFAI) to neural approaches and shows an emerging consensus in favor of neuro-symbolic architectures that couple statistical power with the transparency of formal inference. It identifies three human critical-thinking components relevant for algorithms: a normative one (logical laws and rules), a descriptive one (awareness of common cognitive biases), and a prescriptive one (procedures for self-correction). Building on this mapping, it formulates principles of «logical discipline» for AI: explicit multi-step reasoning, internal self-verification and confidence estimation, tool-assisted fact-checking (calls to external modules and knowledge bases), causal modeling, and explainability. The analysis also clarifies key limitations: valid formal inference does not ensure truth under false premises; the absence of consciousness and intentionality precludes an «intrinsic» drive to truth; long reasoning chains may rationalize an initial error. The paper contends that a human-in-the-loop paradigm is currently optimal: AI acts as a rapid logical filter and hypothesis generator, while the human agent remains the final epistemic arbiter. Practical significance spans education, science, and managerial decision-making: introducing logical and metacognitive control loops can reduce factual errors and strengthen trust in AI. The contribution lies in an interdisciplinary mapping of critical-thinking components to implementable algorithmic mechanisms and in outlining a research program for «thinking» models that integrate logical rigor, causal reasoning, and transparent explanations – while remaining under human oversight.

Published

2025-12-18

Issue

Section

Dnipro Academy of Continuing Education Herald. Series: Philosophy, Pedagogy

How to Cite

NEVMERZHYTSKA, O., & KYIASHKO, S. (2025). CRITICAL THINKING AND FORMAL LOGIC IN ARTIFICIAL INTELLIGENCE SYSTEMS: A PHILOSOPHICAL ANALYSIS. Dnipro Academy of Continuing Education Herald. Series: Pedagogy, 9(2), 33-44. https://doi.org/10.54891/2786-7013/2025-2-4