The integration of artificial intelligence (AI) in cybersecurity promises enhanced threat detection and response capabilities, yet its adoption is hindered by human factors, particularly cognitive biases and trust issues. This study investigates how cognitive biases, such as automation bias (47%) and confirmation bias (37%), influence security analysts’ trust in AI-driven tools, drawing on Kahneman's dual-process theory. Through qualitative interviews with 19 cybersecurity professionals and a comparative analysis of AI solutions from Microsoft, CrowdStrike, Darktrace, and IBM, we identify key barriers to adoption, including explainability gaps and high false positive rates. Findings reveal that 65% of analysts express skepticism toward AI alerts, favoring hybrid human-AI models (79%) over full automation. We propose strategies like explainable AI (XAI), bias-awareness training, and adaptive trust calibration to mitigate biases and foster trust. These insights highlight the need for user-centric AI designs that balance technical performance with human cognitive realities in cybersecurity operations.
Hagen et al. (2025) studied this question.