We offer an extensive analysis of AI-augmented overconfidence, which we term “Borrowed Brilliance,” arguing that fluent AI systems act as a “digital sidekick” to the Dunning-Kruger effect. Using real-world examples, we explain how AI’s confident tone functions as a User Interface (UI) feature that inflates human users’ perceived competence, leading to critical errors, such as a lawyer citing non-existent cases.
The core of the analysis identifies specific human cognitive traps, such as the Illusion of Authority and Automation Over-Reliance, which are amplified by corresponding machine pathologies like Hallucinatory Confabulation and Echo Drift, creating a dangerous self-perpetuating cycle. Finally, we propose five mitigations—including Explain-Back Prompts and Provenance/Confidence Bands—designed to introduce necessary friction and transparency to re-calibrate user confidence with actual competence, thereby addressing harms at the personal, organizational, and societal levels.
