Files
Abstract
Bias is deeply embedded in human perception, decision-making, and social interactions, shaping the ways we interpret the world, develop artificial intelligence systems, and assess the credibility of others. This dissertation examines three interrelated forms of bias—perceptual bias, algorithmic bias, and bias in testimonial exchanges—to provide a comprehensive understanding of how bias manifests at both individual and systemic levels. It first argues that bias is not solely a product of top-down cognitive influences but can emerge from the visual system’s unconscious assumptions, shaping perception before conscious thought occurs. By drawing on research in vision science, this dissertation highlights how perceptual bias can operate independently of cognitive states, challenging conventional distinctions between perception and belief. It then explores the parallels between implicit bias in human cognition and algorithmic bias in artificial intelligence, critically evaluating current models that fail to account for the fluctuating nature of human judgment. While both forms of bias arise from pattern-based learning, this study argues that human implicit bias is more variable than algorithmic bias, resisting simplistic, rule-based interventions. By integrating insights from psychology, machine learning, and philosophy, it proposes a new framework for understanding how biases are encoded, reinforced, and mitigated in both human and computational decision-making. Finally, it examines testimonial injustice as a systemic issue, wherein individuals from marginalized groups are unfairly deemed less credible based on identity prejudice. While some philosophers suggest that cultivating individual virtues—such as open-mindedness and credibility assessment skills—can counteract this form of epistemic injustice, this dissertation argues that such efforts are insufficient. Because testimonial injustice is embedded in broader social structures, meaningful solutions must focus on institutional reforms, including changes in legal, educational, and professional settings that shape credibility judgments. Through a synthesis of philosophical analysis and empirical research, this dissertation contributes to contemporary discussions on the nature of bias, its ethical and epistemic consequences, and the most effective strategies for mitigation. It shows the necessity of interdisciplinary approaches that address both the cognitive mechanisms underlying bias and the structural factors that perpetuate social injustice.