Research by Lee et al. (2025), Bloom, and Gerich has demonstrated that as trust in AI increases, humans stop thinking for themselves, leading to a decline in critical thinking skills. This causes 'cognitive offloading,' where humans delegate cognitive tasks they should perform in their own minds to external tools and environments.

Finding 1: Generative AI Use Reduces Cognitive Effort Across All 6 Stages

Over 70% of respondents reported a decrease in cognitive effort not only in 'lower-order cognition' such as knowledge, comprehension, and application, but also in 'higher-order cognition' like analysis, synthesis, and evaluation. The most significant delegation of thinking is occurring at the core of business decision-making.

- In a study of 319 individuals and 936 business cases, over half responded that 'cognitive effort decreased' across all 6 stages of Bloom's taxonomy. Comprehension (79%) and synthesis (76%) were the highest, while evaluation (55%) was the lowest but still exceeded half.

Finding 2: The Dual Effect of Trust

The decline in thinking ability due to overconfidence in AI is widespread and powerful, but defenses provided by self-confidence are limited and weak. Individual confidence alone cannot resist cognitive offloading.

Finding 3: The Mediating Structure of Cognitive Offloading

Frequent use of AI does not directly rob individuals of their thinking skills; rather, thinking skills are eroded through the formation of a 'habit of having AI think for you' (cognitive offloading)—demonstrating this mechanism.

- The higher the frequency of AI use, the more cognitive offloading (delegation of thinking to AI) increases (r=+0.72), which in turn reduces critical thinking skills (r=−0.75). The direct effect is also r=−0.68. The 'habit of having AI think for you' functions as a mediating variable. Gerlich's (2025) mediation analysis elucidates the mechanism of why AI use leads to a decline in critical thinking skills.

The Expansion of Cognitive Offloading to Higher-Order Thinking

Originally, 'cognitive offloading' is a cognitive science term. Everyday examples include saving a number in a smartphone's contacts instead of memorizing it, or relying on a GPS instead of learning a route. This in itself is a natural behavior humans have engaged in since ancient times; taking notes and using calculators are also forms of cognitive offloading in a broad sense.

However, the particular problem this time is that with the advent of generative AI, the target of offloading has expanded to include higher-order thinking. Lee et al. (2025) empirically demonstrated exactly this phenomenon: because AI outputs appear fluent and authoritative, humans subconsciously judge that they 'don't have to think for themselves' and abandon critical thinking altogether. Moreover, this tendency is stronger in individuals with high trust in AI, creating a paradoxical structure where the more accustomed one becomes to AI, the duller one's thinking skills become. Even more dangerous is the fact, as shown in Finding 2, that even self-confident individuals have limited and weak defenses against excessive trust in AI and offloading.

Now that generative AI is having a widespread impact on the entire human thought process, a major challenge has become clear. The AI thinking design training offered by the GTF Thinking Academy is an evidence-based program that counters this problem with a 3-step process: Self-thinking -> AI augmentation -> Gap verification.

Sources

Lee et al. (2025) — ACM Digital Library Lee et al. (2025) — Microsoft Research PDF Gerlich (2025) — MDPI Societies CHI Conference Impact Factor — Resurchify Harvard Gazette: Is AI dulling our minds? EDUCAUSE: The Paradox of AI Assistance

About GTF Thinking Academy

Since its establishment in 2001, GTF Corporation has accumulated a track record of training over 20,000 executives and interns, with a core business of corporate restructuring and revitalization through a team-dispatch hands-on approach, primarily for listed companies.

GTF Thinking Academy is Japan's first* critical thinking platform that systematizes this practical knowledge to seamlessly provide 'Training (CT 4 modules + AI thinking design)', 'Measurement (proprietary GTF-CTPA test)', and 'Augmentation (prescriptive mini-modules)'. It is designed based on global standard academic evidence (e.g., APA Delphi Report, Abrami meta-analysis).

* As a comprehensive platform integrating training, measurement, and augmentation that covers all 6 core skills defined by Facione (1990) (GTF research, as of May 2026)

FACT BOX

  • Source: PR TIMES
  • Category: Survey
  • Organizations: Microsoft Research / Harvard Gazette / EDUCAUSE
  • Dates in source: 2025
  • Products / services: GTF Thinking Academy / GTF-CTPA