How Generative AI is Shaping Critical Thinking in Knowledge Work

Generative AI (GenAI) tools are not only transforming how knowledge workers approach everyday tasks but are also subtly reshaping the very nature of critical thinking. A recent study presented at CHI ’25 by Hank Lee, Advait Sarkar, Lev Tankelevitch, Ian Drosos, Sean Rintel, Richard Banks, and Nicholas Wilson offers a comprehensive exploration of this transformation. By surveying 319 knowledge workers and gathering 936 real-world examples of GenAI use, the study provides rich insights into both the promises and pitfalls of integrating AI into professional workflows.

The Promise: Efficiency Gains and Enhanced Quality

1. Cognitive Offloading and Reduced Effort

One of the most compelling findings is that a majority of participants reported a reduction in cognitive effort for a range of critical thinking activities. For tasks such as knowledge recall, comprehension, and even synthesis, the use of GenAI often meant that workers experienced “less effort” or “much less effort” compared to traditional methods. For example, when creating artifacts or generating content, workers could rely on GenAI to draft the initial version, allowing them to focus on refining and quality-checking the output rather than starting from scratch. This shift—from material production to critical integration—has been widely reported in both controlled lab studies and real-world settings, reinforcing earlier insights on AI-assisted workflows.

2. Enhanced Focus on Quality Assurance

The study also highlighted how GenAI tools serve as a catalyst for more focused quality control. Participants reported engaging in critical evaluation by verifying AI-generated outputs—be it code that needs to compile correctly or written content that must align with client specifications. In several cases, as seen with a survey participant (P278), the AI’s output was used as a first draft, which was then meticulously refined to meet strict quality criteria. This proactive approach ensures that even though the initial workload is reduced, the responsibility of final quality remains firmly with the human expert.

The Challenge: Over-reliance and Diminished Critical Engagement

1. Confidence in AI vs. Confidence in Self

A striking observation from the study was the dual role of confidence. Knowledge workers with high confidence in their own abilities were more likely to engage in critical thinking, even when using GenAI. In contrast, those who placed high confidence in the AI’s capabilities tended to show a decrease in critical evaluation. This inverse relationship suggests that while GenAI can support productivity, it might also foster a form of complacency if users start overtrusting the technology. One participant (P185) even shared how self-doubt in verifying certain outputs led to a default acceptance of AI responses—a phenomenon that could have significant repercussions in high-stakes scenarios.

2. The “Mechanised Convergence” Risk

Another critical insight revolves around the concept of “mechanised convergence.” This term refers to the tendency of GenAI tools to produce homogenised outputs, which may limit diverse, context-specific problem-solving approaches. When workers repeatedly use the same AI-generated responses, they might inadvertently forgo the opportunity to refine their independent problem-solving skills. The study underlines that while AI can reduce immediate cognitive load, it might simultaneously undermine the long-term cultivation of critical thinking skills—a warning echoed by early thinkers from Socrates to modern-day automation researchers.

3. Barriers to Critical Thinking: Awareness, Motivation, and Ability

The study further categorised inhibitors to effective critical thinking into three key areas:

? Awareness: Some workers did not perceive a need for critical thinking when tasks seemed trivial (e.g., drafting simple emails or social media posts). For instance, P147 noted that when using DALL-E for indirect visual references, there was little incentive to over-correct.

? Motivation: In high-pressure environments like sales, where speed is paramount, there is often little time left to scrutinise AI output. One sales development representative (P295) candidly mentioned that meeting daily quotas left no room for deep critical engagement.

? Ability: Even when motivated, some participants expressed challenges in verifying and improving GenAI outputs—especially in domains where they lacked specialized knowledge. As P290 explained, insufficient domain expertise can make it difficult to detect inaccuracies or biases in AI responses.

Key Takeaways and Implications for the Future

1. Reimagining Critical Thinking in the Age of AI

This study forces us to reframe our understanding of critical thinking—not as a static, innate ability, but as a dynamic skill that is influenced by technology. GenAI doesn’t simply automate tasks; it reshapes the way we think by shifting our cognitive load from creation to verification and integration. This realignment offers both opportunities for efficiency and risks of skill atrophy.

2. The Need for Adaptive AI Design

For developers and organisations, these insights underscore the importance of designing AI tools that not only assist with routine tasks but also encourage ongoing critical engagement. Features like interactive prompts, built-in verification checkpoints, and adaptive feedback mechanisms could help mitigate overreliance while promoting a more reflective, judicious use of AI.

3. Balancing Automation with Skill Development

As the study shows, there is a delicate balance between harnessing the power of AI to reduce cognitive load and ensuring that human critical thinking skills remain sharp. Organisations should consider training programs that foster independent problem-solving and continuous learning, even as they integrate AI into their workflows. Encouraging a culture where verification and quality control are valued as much as efficiency can help preserve essential cognitive skills.

Concluding Thoughts

The integration of generative AI into knowledge work is undeniably transformative. As we embrace these powerful tools, it is crucial to remain mindful of the subtle shifts they induce in our cognitive processes. The work of Lee et al. provides both a warning and a roadmap—a call to design AI that not only augments our capabilities but also safeguards the critical thinking skills that are vital for innovation and long-term success.

For professionals navigating this new landscape, the message is clear: leverage the efficiency of GenAI, but never at the expense of your own analytical rigor. By striking the right balance, we can harness the full potential of AI while continuing to build and refine the human mind.