Metacognition, a term first introduced by Flavell (Flavell 1976), refers to the self-awareness of one’s thinking and learning patterns. Think of it as learning how to learn. It is not just about acquiring information but understanding how we process and retain that information. For instance, when preparing for an exam, some students might find that they recall material better when they teach it to someone else, while others might benefit from visual aids or mnemonic devices. Recognizing and acting on these personal insights is the essence of metacognition. It encompasses activities like planning how to approach a particular task, monitoring one’s comprehension, and evaluating the effectiveness of strategies used. For example, if you are reading a complex article and realize you have not understood the last few paragraphs, metacognition would involve recognizing this lapse and re-reading the section or seeking external resources to clarify the concept.
The true strength of metacognition is its role in creating self-sufficient learners. These individuals not only absorb information but understand how they learn best, adjusting their methods accordingly. This self-awareness translates into effective learning strategies that can be applied in various scenarios throughout life (Rivas et al. 2022). Hence, metacognition is an important learning goal in education, but we still know little about how Artificial Intelligence (AI) could facilitate students’ metacognitive thinking.
The strength of metacognitive control
Metacognition revolves around two core components: metacognitive knowledge and metacognitive control (Fleur et al., 2021). Metacognitive Knowledge dives into one’s awareness and understanding of their cognitive processes. It is akin to knowing how our minds work and our ability to reflect upon these processes. It covers personal insights into one’s strengths, weaknesses, and biases (Velzen et al., 2015). For a practical example, consider a teacher aiming to learn a novel teaching approach. He might find that rather than just reading about a concept, he grasps it better when engaged in hands-on activities. With this insight, the teacher might opt for workshops or interactive sessions to deeply understand the new technique.
Metacognitive control emphasizes the self-regulation of learning. It is about guiding our cognitive processes, adapting, and fine-tuning based on the feedback we get. If one recognizes they are struggling with a particular task, they will modify their strategy for better outcomes (Bran & Balas 2011). Drawing from our previous example, let us say our teacher is now venturing into a new educational technology tool. He might start with a structured learning plan, allocating specific times to understand its various features. However, upon encountering a particularly challenging feature, he reassesses his approach. Instead of pushing ahead, he adjusts his schedule to allocate more time to this feature.
Significance of metacognitive skills in education
Metacognitive skills hold undeniable significance for both teachers and students. These skills foster self-reliance, and empower individuals to become lifelong learners. Rather than solely relying on instructors, learners equipped with metacognitive abilities can independently navigate information acquisition, problem-solving, skill mastery, and knowledge dissemination.
In our swiftly evolving world, it is nearly impossible for institutions to provide students with every requisite skill for the future. Consequently, the emphasis should shift from mere knowledge transfer to teaching students “how to learn.” Encouragingly, metacognitive skills can be cultivated and taught (Çakiroğlu & Betül 2023). Acknowledging their pivotal role, the revised Bloom’s taxonomy has integrated metacognitive skills across its levels (Anderson et al. 2001).
Using AI for boosting metacognitive skills
A landmark achievement in the AI arena is the emergence of Large Language Models (LLMs) such as GPT. Initially designed for text-based tasks, these models can generate coherent, human-like narratives, occasionally even outperforming human outputs. Their performance on various cognitive assessments is equally commendable. Moreover, LLMs have evolved beyond just text, as they can now interpret and produce content across multiple formats. Their influence is already making waves in the realms of education and research (Khan 2023).
The discussion surrounding AI’s cognitive abilities continues to evolve. While certain assessments indicate that LLMs, such as GPT, demonstrate qualities such as context awareness, problem-solving, and reasoning (Joublin; Orru; Anton 2023), AI has not yet reached the complexity of human cognition. However, the primary goal of AI is not to outperform humans, but to augment our capabilities. By incorporating generative AI tools based on LLMs in educational settings, instructors can craft tailored learning experiences that foster metacognitive reflection and consistently promote the growth of critical thinking skills (Hutson & Plate 2023).
Metacognition encompasses several skills, with critical thinking and reflection standing out as pivotal. Envisioning a scenario where AI tools can facilitate the learning of these crucial skills, we can tailor it specifically for a Business ICT program. This model not only offers a technologically advanced approach but ensures that students in the Business ICT domain can effectively analyze complex business scenarios, make data-informed decisions, and reflect on their strategic choices, thereby enhancing their problem-solving capabilities in real-world contexts.
Scenario: Tech Startup Challenge
A crucial consideration for educators centers on the timing and methods through which students engage in metacognitive control. Further, it is pivotal to understand how AI can seamlessly fit into these advanced thinking processes. To illustrate, let us consider a specific scenario.
Business students are tasked with launching a mock tech startup with a €100,000 budget. Given resources costing a total of €150,000, they must strategically select which to prioritize, factoring in their startup’s nature, challenges, and potential ROI. They identify a target market, researching its demographics and preferences. Students will present their startup progress to virtual stakeholders, aiming to get feedback and understand stakeholder expectations. They navigate changes like consumer shifts and regulatory updates within a dynamic simulated market. Budget considerations also guide their tech integration decisions, such as adopting a chatbot or e-commerce platform.
Initiating with a literature review that highlights the significance of strategic decision-making in startup launches and the rising role of AI in business decision support, the project propels students into a simulated startup scenario. Several AI-driven tools that assist in summarizing, explaining, comparing, and finding studies based on semantic similarity address the “analyzing” level of Bloom’s taxonomy. AI-powered search engines such as Perplexity AI and Microsoft Bing Chat could be harnessed to fetch precise, up-to-date data on market trends and competition. Similarly, the data analytics tools like ChatGPT’s Code Interpreter can provide rapid insights into the underlying market patterns. For a comprehensive discussion on the relevant AI tools, please see our previous article (Khan 2023).
To touch upon Creating, the top level of Bloom’s taxonomy, predictive analytics tools allow students to simulate various scenarios. They can manipulate their resource allocations, drawing a parallel between divergent choices and their consequent outcomes. Several predictive analysis tools can be utilized such as IBM SPSS Modeler, Tableau, Microsoft Azure Machine Learning, and Alteryx Analytics, to name a few.
Post-completion of the mock startup launch scenario, students initiate their reflection journey by engaging in a structured session with a Conversational Agent (CA). With insights into the project’s parameters and individual student performance, the CA, might ask questions like, “Can you explain the reasoning behind your resource allocation?” If the CA detects a significant allocation to a particular aspect, like marketing, it dives into the metacognitive knowledge dimensions of Bloom’s taxonomy by querying, “Why did you prioritize marketing over product development?” As students articulate their rationale, the CA further probes, drawing out deeper layers of analysis. For instance, when a student reflects, “I believe we allocated too much to marketing,” the AI, anchoring to the Analyzing phase, counters with, “What specific outcomes or trends led you to this belief?” A domain-specific CA (Yager et al. 2023) could offer better feedback by aligning its queries with business ICT and best practices, enabling students to align their thought processes with real-world standards and expectations.
Following individual reflections, students converge in groups to discuss their insights, moving to the metacognitive level in their social thinking. Here, they compare and contrast their strategies, weighing the effectiveness of different approaches. This peer discussion not only fosters a broader perspective but also underscores the diverse strategies that their colleagues adopted.
The outcome of the reflective activity is the class-wide discussion steered by the instructor. By stitching together individual and group insights, the session ventures into the conceptual, procedural, and metacognitive knowledge dimensions of Bloom’s taxonomy. The instructor emphasizes emerging innovative strategies, solutions to common challenges, and novel insights gained through reflection. With this collective knowledge, students are encouraged to ideate on alternative strategies or solutions they could employ in similar future scenarios. These rich reflections further serve as valuable feedback for the educator. By discerning areas of strength and potential growth, they can tailor feedback to individual students. Moreover, these insights prove instrumental in refining future course structures or designing even more engaging and educational project scenarios.
The AI-driven reflection approach offers advantages over human instructors. AI agents can simultaneously engage multiple students, offering personalized feedback based on each student’s project performance. They ensure consistent feedback unaffected by fatigue or mood and can instantly analyze vast amounts of data, making the reflection process immediate and relevant. Accessible anytime and adaptable, AI tailors its approach based on a student’s grasp of concepts. It not only benefits students but also provides educators with structured, data-driven insights on student performance, enabling better course design. Operating on objective standards, AI ensures equitable evaluation of all students. Furthermore, by handling the primary reflection phase, AI allows instructors to concentrate on higher-value tasks, optimizing resource allocation.
In summary, metacognitive thinking equips students to introspectively assess their thought processes, problem-solving tactics, and their actual methodologies in tackling issues. While students can also address the given business planning scenario using established routines and pre-learned methods without introspection, metacognitive thinking will help in critically evaluating concerns such as their comprehension of the context, the suitability of their current methods, the need for innovative solutions, or whether they have genuinely weighed all viable alternatives instead of making impulsive decisions.
AI’s strength lies in its capacity to prompt students to re-evaluate their initial concepts, selected business strategies, and methodologies for gathering and utilizing new information. However, several algorithms, particularly on social media and e-commerce platforms, inadvertently reinforce groupthink by forming “social bubbles.” Thus, there’s an increasing need for AI-augmented teaching and learning methods to support the growth of students’ metacognitive abilities.
Balancing the promise and perils of AI in metacognitive education
By tailoring AI tools for specific domains, we can pave the way for students to analyze complex scenarios more effectively, make informed decisions, and reflect on their strategic choices, thereby refining their problem-solving capabilities in real-world contexts. However, as with all technological advancements, the integration of AI in education is not without its challenges. Ethical issues, especially concerning data privacy and biases in AI algorithms, are bound to emerge. Furthermore, the inherent nature of generative AI models to produce plausibly looking false outputs poses a significant challenge. While AI tools can provide tailored learning experiences, relying solely on them without human oversight could lead to misinformation or skewed learning experiences.
This also requires policies and strategies to transform the existing education system. Ensuring digital accessibility, fostering technological awareness, promoting inclusiveness, and capacitating teachers to leverage AI tools effectively are equally important. Without these foundational elements in place, the true potential of AI in enhancing metacognitive learning will remain unrealized. The future of education, underpinned by AI, is bright, but only if we navigate its complexities with foresight, responsibility, and inclusivity.
Anderson, Lorin, et al. 2001. A Taxonomy for Teaching, Learning, and Assessment: a revision of Bloom’s taxonomy of educational objectives, Pearson.
Bran, C.-N. and Balas, E.-C. 2011. Metacognitive regulation and in-depth learning. A study on the students preparing to become teachers. Procedia-Social Behav. Sci., vol. 11, pp. 107–111.
Çakiroğlu, Ü. and Betül, E. R. 2023. A model to develop activities for teaching programming through metacognitive strategies. Thinking Skills and Creativity vol. 48.
Flavell, J. H. 1976. Metacognitive aspects of problem solving. Nat. Intell.
Fleur, D. S., Bredeweg, B. and van den Bos, W. 2021. Metacognition: ideas and insights from neuro-and educational sciences. NPJ Sci. Learn., vol. 6, no. 1, p. 13.
Hutson, J. and Plate, D. 2023. Human-AI Collaboration for Smart Education: Reframing Applied Learning to Support Metacognition. IntechOpen.
Joublin, F. et al. 2023. A Glimpse in ChatGPT Capabilities and its impact for AI research. arXiv preprint arXiv:2305.06087.
Khan, U.A. 2023. The unstoppable march of artificial intelligence: The dawn of large language models. eSignals Pro. Haaga-Helia.
Korinek, A. 2023. Exploring the impact of language models on cognitive automation with David Autor, ChatGPT, and Claude. Brookings.
Orru, G. et al. 2023. Human-like problem-solving abilities in large language models using ChatGPT.” Frontiers in Artificial Intelligence, vol. 6.
Rivas, S. F., Saiz, C. and Ossa, C. 2022. Metacognitive strategies and development of critical thinking in higher education. Front. Psychol., vol. 13, p. 913219.
Van Velzen, J. 2015. Metacognitive learning. New York, NY: Springer International Publishing.
Yager, K. G. 2023. Domain-specific ChatBots for Science using Embeddings. arXiv preprint arXiv:2306.10067.