Modeling Ambiguity in AI-Enhanced Learning: A Neutrosophic Approach to Stance Detection and Causal Evaluation

Oscar José Alejo Machado^1,*, Adriana María Estupiñán Sera¹, Maikel Y. Leyva Vazquez²,
Florentin Smarandache³

¹Instituto Superior Tecnológico de Investigación Científica e Innovación (ISTICI), Quito, Ecuador

²Bernardo O’Higgins University, Institutional Research Center, Chile

³Emeritus Professor of Mathematics at the University of New Mexico, Gallup, New Mexico, USA

Emails: oscar.alejo@istici.edu.ec; Sera.Adriana@gamil.com; Vazquez.Maikel@gamil.com; smarand@unm.edu  

Abstract

This work presents a neutrosophic stance detection model to bridge computational assessment and logic of indeterminacy in artificially intelligent (AI)-mediated learning and its outcomes. Utilizing the BART-large-MNLI model, a causal assessment was made of five hypotheses stemming from AI-supported learning between teacher-student relationships. These stances were then transformed into refined neutrosophic values (truth (T), partial support (), indeterminacy (I), partial opposition () and falsity (F)). Ultimately, findings suggest that partial support is the most prevalent stance applied to any of the hypotheses, revealing that AI is, largely, a boon to education. However, this valence is tempered by indeterminacy among axes as well as stance magnitude. The largest partial support in rank order came from personalized education and access to AI tutors, while the most importance was given to opposition of relying on AI as support and replacement AI learning. Such findings confirm neutrosophic stance analysis and causal graph modeling as increasingly successful for applying measurable patterns to epistemically ambiguous fields. The neutrosophic causal graph integrates the above findings with a visualization of proposed dynamics between each vertex based on both quantitative patterns and epistemic uncertainty trends. The current research holds implications for educational theory, policy and instructional design integrity in 21st century learning. Uncertainty became a tangible concept; instead of devaluing AI in the classroom, it must be present as an enhancing supplemental tool, never replacement, for ethical considerations and equitable access. The potential for neutrosophic to transform apparent truths that are at times contradictory is confirmed through the human-machine interactive learning process, with subsequent suggestions for future research into AI-mediated education's causal relationships and decision-making potential.

Keywords: Neutrosophic; Stance detection; Causal analysis; AI in education; Uncertainty modeling