Does AI harm critical thinking? It depends on when it’s used
The following contribution comes from the Science News website, which describes itself as follows: For more than a century, Science News journalists have covered advances in science, medicine, and technology for the general public, including the Scopes trial in 1925, the dawn of the atomic age in 1945, the space race, and the genetic engineering revolution, from the discovery of DNA to today’s gene-editing technology.
We believe in the power of knowledge and the free flow of information to build informed, enlightened, and engaged communities. Our mission is to provide independent and unbiased science coverage and give people the tools to evaluate the news and the world around them.
Author: Aaron Brooks
If used at the end of writing an essay, chatbots can help incorporate more perspectives.
A person opens an AI chatbot on their phone.
A small study suggests that people who used an AI chatbot after partially solving a problem improved their critical thinking skills.
The next time you’re about to ask an AI chatbot for help solving a difficult problem, you might want to think twice.
People who waited to consult an AI chatbot until they had partially solved a problem on their own performed better on a critical thinking task than those who used the chatbot from the start, researchers reported on April 14 at the CHI 2026 conference on Human Factors in Computing Systems in Barcelona. However, under tight deadlines, early use of AI did provide an advantage, highlighting the trade-off between speed and independent reasoning, and raising questions about how and when we should use chatbots.
In the study, computer scientist Mina Lee of the University of Chicago and her colleagues randomly assigned 393 people to one of eight categories. First, participants were divided into two large groups: those with enough time (30 minutes) or those with insufficient time (10 minutes). Participants were then divided into smaller groups based on when, or if, they could use OpenAI’s GPT-4o chatbot: early access, continuous access, late access, or no access. Each group had approximately 40 to 50 participants.
Participants were then asked to assume the role of a city council member and decide, using seven documents, whether to accept or reject a company’s proposal to mitigate a water pollution problem. Each participant had to write an essay explaining their decision.
The researchers graded the essays, in part, based on the number of valid arguments and textual references they contained, and found that participants who had 30 minutes performed better overall than those who had only 10 minutes. The highest-scoring participants were those who had sufficient time to complete the task and subsequently had access to the chatbot.
Tener tiempo suficiente y sin acceso al chatbot
Having enough time and no access to the chatbot
When analyzing participants’ ability to recall information from the provided documents, the group that performed best was the one that had enough time and no access to the chatbot. Researchers also assessed perspective bias, measuring how many perspectives participants incorporated into their arguments. They found that the group with enough time and late access to the chatbot performed best.
The results align with research on two types of learning: one based on slow, laborious reasoning, and the other on fast, automatic thinking, explains Barbara Oakley, a systems engineer and education expert at Oakland University in Rochester Hills, Michigan. Slow learning involves carefully understanding the problem and weighing options, while fast learning relies on habits and quick judgments with little reflection. Participants who had time to analyze the material on their own before using the AI performed better because they had already engaged in that slower, more deliberate learning, she adds.
Of course, in the real world, people often have to complete critical thinking tasks under time pressure.
In the four groups in the «insufficient time» category, the group that had access to the chatbot from the beginning scored highest on their essays. That doesn’t mean we should rush to use AI, says Lee. «When you’re working under time pressure and using AI to improve performance, you run the risk of adopting and using only the AI’s approach, which reduces the range of arguments that can be presented and the interaction with documents or information,» he says. It’s crucial to «be aware of the implications of this practice.»
That awareness is probably what everyone should be working on right now. According to Lee, people will need a solid understanding of AI and their own thinking patterns to weigh the risks and benefits of using chatbots in different scenarios and at different stages of problem-solving. «I think our work focuses on time constraints as a first step toward that understanding.»
How Artificial Intelligence (AI) Tools Affect Critical Thinking Skills and Cognitive Offload
The following contribution comes from the Oxford Review website, which describes itself as follows: How we started and what we do
In 2015, David Wilkinson was in the final stages of writing a book (according to him). Since his day job prevented him from writing, he decided to isolate himself and focus exclusively on it. However, he had two problems. The first was his coaching and consulting clients. He didn’t want to lose contact with them, as they were his livelihood. The second problem was that he constantly had to interrupt his work to go to the Bodleian Library to search for and retrieve data for his research.
Authorship by the team.
With the rapid adoption of AI, and in particular generative AI tools, in the workplace, there is significant interest in research on the impact of AI on humans. For example, there are more than 530,000 studies that analyze some aspect of AI and critical thinking.
Critical Thinking and Artificial Intelligence
Critical Thinking
Critical thinking is a multifaceted cognitive process that involves the ability to analyze, evaluate, and synthesize information to form sound judgments and make informed decisions. It encompasses clear and rational thinking, logical reasoning, and the ability to identify inconsistencies in arguments and evidence.
Critical thinking comprises several interconnected skills and dispositions:
Analysis: breaking down complex information into simpler components.
Evaluation: assessing the credibility and relevance of information.
Inference: drawing logical conclusions from evidence.
Self-regulation: reflecting on one’s own reasoning processes.
Problem-solving: applying reasoning to overcome challenges.
Decision-making: selecting options based on reasoned analysis.
Beliefs and Evidence
Let your beliefs be based on evidence… Get expert-curated research evidence for your practice and decisions… for free… directly from a real person.
Cognitive and Neurological Processes Involved in Critical Thinking
Several cognitive and neurological processes are involved in critical thinking. For example:
Inhibitory control: suppressing impulsive or biased responses.
Working memory: retaining and manipulating information.
Cognitive flexibility: switching between concepts or perspectives.
Metacognitive processes, such as evaluating the strength of an argument or recognizing knowledge gaps, are crucial for critical thinking. These include:
Monitoring: assessing the accuracy and reliability of recall and memory.
Control: deciding when to revise strategies or seek more information.
Greater metacognitive accuracy correlates with better critical reasoning in decision-making tasks.
The Degree of Cognitive Offloading
Cognitive offloading refers to the use of external tools or aids to reduce mental effort. Cognitive offloading helps manage cognitive load, but it can reduce the activation of System 2 (slow, deliberate, and analytical reasoning), which is essential for critical evaluation. When people rely on System 1 (fast, intuitive processes), critical reasoning can be impaired. For example, dependence on AI tools or calculators can harm mental calculation skills and independent judgment if used without reflection.
Emotional regulation: While critical thinking is often considered rational, emotional regulation plays a crucial role.
This tension between technological advancement and cognitive independence presents a challenge for organizations.
A 2011 study that coined the term “Google effect,” also known as “digital amnesia” (the tendency to forget information easily accessible through search engines like Google), actually refers to technology-facilitated cognitive offloading, where people delegate mental tasks to external tools. The use of these tools fundamentally alters how humans process and retain information. More recent evidence shows that this offloading can extend beyond memory and impact higher-order thinking processes.
Cognitive Offloading
Previous research on the cognitive impact of digital tools has revealed that:
Research on trust in AI systems has revealed troubling patterns regarding critical evaluation. A 2024 study found that greater trust in AI tools correlated with greater cognitive offloading and less critical evaluation of AI-generated content. This relationship suggests a potentially problematic cycle in which trust diminishes the perceived need for critical evaluation.
A 2015 study found that the «black box» nature of many AI systems discourages people from critical analysis, as users have a limited understanding of how recommendations or decisions are generated, further reducing critical engagement.
A 2001 article suggested that young people who have grown up with digital technology are less susceptible to the cognitive impact of digital tools.
The Myth of the Digital Native
A common belief that more recent research has debunked is the notion of the «digital native,» coined by Prensky in 2001, which presupposes that young people born after 1980 inherently possess advanced digital skills due to lifelong exposure to technology. However, this has been shown not to be the case.
Exposure to digital technology does not equate to digital literacy. Digital literacy encompasses more than technical skills. It includes cognitive dimensions (critical evaluation, ethical use) and socio-emotional dimensions (safe and responsible online behavior). This distinction is crucial for evidence-based practice.
Digital Literacy
A New Study
A new study by the Center for Corporate Strategic Foresight and Sustainability at SBS Swiss Business School (Switzerland) examined the impact of generative AI tools like ChatGPT on critical thinking, deep thinking, and cognitive offloading.
Deep Thinking Activities
Deep thinking activities are cognitive practices that demand sustained intellectual engagement, analytical reasoning, and reflective processing that goes beyond the superficial consumption of information. These activities require focused attention, critical analysis, and the construction of meaningful connections between concepts, fostering higher-order cognitive skills. Deep thinking contrasts with surface processing, which involves a more superficial interaction with information without substantial analytical or evaluative components.
Deep thinking activities include reading books, solving puzzles, and participating in debates; all of which require prolonged intellectual engagement without immediately relying on AI tools.
Other studies have identified various deep thinking activities:
Extensive reading of complex texts that require sustained attention and interpretation.
Writing activities that involve articulating and refining complex ideas.
Problem-based learning that requires applying knowledge to novel situations.
Philosophical discussions that explore fundamental questions and assumptions.
Artificial Intelligence.
Results:
The study found that:
There is a strong negative correlation between frequent use of AI tools and critical thinking skills.
Cognitive offloading was a significant predictor of lower engagement in critical thinking.
Younger participants (17–25 years old) showed greater reliance on AI and lower scores on critical thinking.
Older participants (46+ years old) showed less reliance on AI and higher critical thinking skills.
Higher educational attainment was correlated with better critical thinking, regardless of AI use.
Higher educational attainment significantly predicted greater engagement in deep thinking activities.
Gender did not show a significant effect on deep thinking activities.
Use of AI tools was the strongest predictor of critical thinking skills. Higher education can mitigate some of the negative effects of AI tools on critical thinking.
The relationship between AI use and critical thinking is not linear. This nonlinearity suggests that the relationship is more complex than a simple linear relationship of «more AI use equals less critical thinking.» It could indicate that moderate AI use might be manageable without significant negative effects, but beyond certain levels of use, the negative impact on critical thinking either accelerates or plateaus.
Participants expressed concern about the long-term impacts of AI on cognitive skills.
Reliance on AI tools is correlated with greater cognitive offload.
To Think or Not to Think: The Impact of AI on Critical Thinking Skills
The following contribution comes from the NSTA website, which describes itself as follows: Welcome to the National Science Teachers Association (NSTA), a vibrant community of 35,000 members: educators and science professionals committed to best practices in teaching science and STEM (science, technology, engineering, and mathematics) disciplines and their impact on student learning.
Authors: Christine Anne Royce and Valerie Bennett
Christine Anne Royce, PhD, is a former president of the National Science Teachers Association and is currently a professor of teacher education and co-director of the Master of Science in STEM Education program at Shippensburg University. Her areas of interest and research include the use of technologies and digital tools in the classroom, global education, and the integration of children’s literature into science teaching. She is the author of more than 140 publications, including the column «Science and Kids: Teaching Through Trade Books.»
Valerie Bennett, Ph.D., Ed.D., is an adjunct professor of STEM education at Clark University Atlanta, where she also serves as director of the Graduate Teacher Training Program and director of Educational Technology and Innovation. With over 25 years of experience and engineering degrees from Vanderbilt University and Georgia Tech, she focuses on STEM equity for underrepresented groups. Her research includes AI interventions in STEM education, and she currently co-directs the NSF Noyce Fellowship, works with the AUC Data Science Initiative, and collaborates with Google to address diversity and participation in the computer science workforce within the Atlanta University Center’s K-12 community.
Navigation path
NSTA Blog: To Think or Not to Think: The Impact of AI on Critical Thinking Skills
By Christine Anne Royce, Ed.D. and Valerie Bennett, Ph.D., Ed.D.
Published March 10, 2025
To Think or Not to Think: The Impact of AI on Critical Thinking Skills
Disclaimer: The opinions expressed in this blog post are those of the authors and do not necessarily reflect the official position of the National Science Teachers Association (NSTA).
