In a nutshell: AI use in the workplace leads to cognitive capitulation, where employees abandon critical review and uncritically accept erroneous machine results.
Researchers at Wharton School show that employees accept erroneous AI outputs without verification in 80 percent of cases while simultaneously increasing their confidence in the results unjustifiably. The phenomenon of cognitive capitulation endangers companies’ strategic judgment competence.
Wharton School is examining in its current research a phenomenon that economists Steven Shaw and Gideon Nave term cognitive capitulation: the reflexive adoption of AI-generated results with minimal critical review. The human mind tends to switch off both its own intuition and logical reasoning as soon as an AI system provides an answer. The researchers thus expand on the established model of psychologist Daniel Kahneman, which divides human thinking into a fast intuitive system (System 1) and a slow reflective system (System 2). With algorithm-based tools, an additional external System 3 emerges — artificial cognition. When this system no longer merely supports human cognitive performance but replaces it, the employee loses their function as a control instance.
The quantitative impact on enterprise performance is considerable. In Wharton School tests, accuracy among test subjects rose significantly when the AI provided correct data. However, as soon as the system produced erroneous results, the success rate of AI users fell below the baseline values of people working without algorithmic support. The central problem is users’ inability to recognize quality differences: test subjects accepted incorrect AI answers in 80 percent of cases. At the same time, employee confidence in the correctness of their work increased regardless of whether the technology had helped them or led them astray.
This dynamic is particularly critical with large language models. Such models do not retrieve verified facts but generate plausible text patterns based on historical training data. They do not reference specific organizational context, strategic objectives, or situational expert knowledge. Because they formulate every answer with uniform linguistic certainty, without signaling uncertainty, the user’s critical distance is systematically eroded.
A Microsoft Research study from April 2025 confirms these findings: trust in an AI tool is the strongest predictor of whether knowledge workers apply critical thinking at all. The higher the trust, the lower the willingness to question results. In the long term, companies thus face a loss of internal expert competence. When routine cognitive tasks are completely transferred to external systems, the continuous practice required to maintain human judgment capacity disappears — the ability to independently evaluate complex matters degenerates.
Source: www.it-daily.net · Published June 5, 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.5.2.