COGNITIVE BIAS DETECTION THROUGH NATURAL LANGUAGE PROCESSING: A COMPUTATIONAL FRAMEWORK FOR ORGANIZATIONAL DECISION-MAKING

Abstract

Cognitive biases distort human judgment and undermine rational decision-making in organizations. Recent advancements in Natural Language Processing (NLP) now make it possible to computationally detect linguistic signals associated with these biases, offering a transformative pathway for data-driven governance. This study examines how machine learning, semantic embeddings, and transformer-based models can identify patterns linked to biases such as confirmation bias, anchoring, optimism bias, loss aversion, and overconfidence within managerial communication. Using a cross-sectional dataset of organizational emails, meeting transcripts, and corporate reports from multinational firms, the analysis evaluates linguistic indicators, contextual dependency patterns, and decision outcomes. Findings reveal a strong correlation between bias-associated language and suboptimal strategic decisions, demonstrating that NLP-based detection significantly enhances risk mitigation and decision transparency. The study also highlights limitations related to contextual ambiguity, domain adaptation, and privacy constraints. Results indicate that integrating computational bias-detection systems with organizational workflows can serve as a critical safeguard for rational decision behavior, strengthening corporate governance and reducing cognitive risk.