Protecting Human Cognition in the Age of AI

1 Introduction

• GenAI may reshape cognition as it makes synthesized, structured information widely accessible, while evidence also points to benefits and concerns about critical thinking and reasoning.Reported benefits include productivity, creative work, and learning experiences; concerns include possible long-term effects on cognition.Source: §1 · p. 1
• The analysis focuses on novices, such as students, and considers how stress and information overload interact with GenAI’s cognitive effects.The paper primarily examines potential hindrances to cognitive development, while recognizing that impacts depend on how GenAI is used.Source: §1 · p. 1
• The paper synthesizes literature on GenAI’s cognitive effects and uses Bloom’s Revised Taxonomy and Dewey’s reflective thinking to examine mechanisms and inform educational interventions.It proposes reflective thinking as a way to evaluate impacts on learners’ critical thinking and guide intervention design.Source: §1 · p. 1

2 Emerging Cognitive Challenges in the AI Era

• Human cognition includes memory, attention, reasoning, critical thinking, and creativity, and GenAI raises challenges across several of these abilities.The section situates GenAI within a longer history of technology affecting cognition.Source: §2 · p. 1

Knowledge Acquisition

• GenAI search may shift information seeking toward passive consumption of potentially one-sided responses, while alignment with prior beliefs can reinforce echo chambers and sideline less visible perspectives.A cited study found users selectively attended to or retained responses and spent more time on consonant than dissonant opinions.Source: Knowledge Acquisition · pp. 1–2

Reasoning

• Studies report over-reliance on AI advice and effects of AI-generated explanations on beliefs, while AI-supported performance may not translate into knowledge gain, transfer, or independent problem-solving.Students in some settings struggled when AI support was removed, and learners may mistake ease of task completion for understanding.Source: Reasoning · p. 2, Learning · p. 2

Learning

• GenAI’s effects on creativity depend on its definition and task: assistance can improve some stories while reducing collective diversity, and LLMs may help convergent but hinder divergent thinking.The section also cites an argument that widespread use may reduce the exercise of human creative abilities.Source: Creativity · p. 2

Creativity

• Cognitive offloading is associated with reduced cognitive engagement and critical thinking, and novices may face additional metacognitive difficulties when learning with GenAI.The cited work argues for explicitly teaching metacognitive skills to novice learners.Source: Metacognition and Critical Thinking · p. 2

3 Mechanisms of AI-Induced Cognitive Change

• Bloom’s framework distinguishes knowledge types from cognitive processes, while Dewey’s framework treats reflective thought as disciplined evaluation of evidence.Dewey’s prerequisites include inquiry, prior knowledge, persistent consideration, suspended judgment, and connecting and evaluating ideas.Source: Bloom’s Revised Taxonomy · p. 2, Dewey’s Theory of Reflective Thinking · p. 3

Bloom’s Revised Taxonomy

• GenAI can accelerate access to factual and procedural knowledge, but this may bypass the slower cognitive work of remembering, applying, analyzing, and evaluating that supports learning.The programming example contrasts sourcing and synthesizing web information with outsourcing some of those processes to an LLM, potentially impeding metacognitive skill development.Source: Bloom’s Revised Taxonomy · p. 3
• Reduced cognitive difficulty from over-reliance may limit deeper metacognitive processes, and novices may be especially susceptible because they have less experience with learning strategies.The paper calls this pattern “metacognitive laziness.”Source: Bloom’s Revised Taxonomy · p. 3
• Experts may be better positioned to use GenAI for high-level decisions because they have domain knowledge for effective prompting, though over-reliance may hinder deliberate practice.The paper describes potential expert uses including offloading lower-level tasks, iterative creative work, and text comprehension.Source: Bloom’s Revised Taxonomy · p. 3

Dewey’s Theory of Reflective Thinking

• GenAI may weaken reflective thinking when immediate answers reduce further doubt, reinforce existing beliefs, present confidence without uncertainty, or discourage deeper processing.The paper links these patterns to possible failures to detect inaccuracies and to superficial understanding.Source: Dewey’s Theory of Reflective Thinking · p. 3
• Stress and information overload can strain attention and regulatory capacities, making it harder to critically engage with GenAI-generated content.The paper also describes a possible interaction between AI’s human-like behavior, emotional dependence, and reflective reasoning.Source: Societal Factors Interacting with AI’s… · p. 4

4 Supporting Thinking & Learning in the AI Era

• Teaching and assessment should emphasize critical and evaluative skills, including activities in which students actively critique GenAI outputs.The paper notes that curricula and high-stakes tests currently prioritize formulaic decision-making, a skill at which AI excels.Source: Implications for Educators and Test… · p. 4

Implications for Educators and Test Designers

• Early learning should preserve cognitive effort and persistence, with minimal AI use primarily for formative feedback and guardrails that gradually increase learner-AI interaction.The proposed progression aims to let learners exercise independent judgment when seeking AI assistance.Source: Implications for Educators and Test… · p. 4

Promoting Active Engagement with GenAI Tools

• Dewey’s prerequisites for reflective thinking can help identify passive AI use and guide interventions, while evaluation should assess engagement, critical thinking, and depth of learning.The paper argues that traditional evaluation metrics may not capture these dimensions.Source: Promoting Active Engagement with GenAI… · p. 4
• To support inquiry and persistent exploration, tools could add friction before output use and connect new information to prior knowledge through navigable schemas.The paper suggests highlighting parts of responses and using knowledge graphs alongside conversational AI.Source: Promoting Active Engagement with GenAI… · p. 4
• Metacognitive prompts can encourage learners to pause, consider alternatives, and assess comprehension before accepting persuasive or belief-aligned AI responses.The paper also notes AI-generated provocations that highlight risks, biases, limitations, and alternatives.Source: Promoting Active Engagement with GenAI… · p. 4

• Current short-term studies make definitive conclusions about GenAI’s long-term effects difficult, underscoring the need for long-term studies.Source: §5 · p. 4
• The paper identifies implications for education, interaction design, and cognitive development in the AI era.Source: §5 · p. 4
• Younger users are particularly vulnerable to the negative effects of GenAI overuse.Source: §5 · p. 5