Psychology of Singularity – Your Source for AI Information

The field of education has many books on how to apply and use AI in the classroom, but a missing piece is that there are fewer models. A model for AI in education must keep humans in the loop and be used alongside learning and not replace it. The following is the model I propose because the end goal is for the human to oversee AI. In order to do that, the learner must possess a certain level of mastery knowledge.

Pedagogy in the Age of Artificial Intelligence: An AI-Educational Framework for Higher Education

By: Dr. Jill Maschio

Copyright 2026

Artificial intelligence (AI) is changing how educators will teach and facilitate learning. The paramount challenge is determining how to use AI effectively as a pedagogical tool. The heart of this discussion is learning itself and whether AI can accelerate deep learning rather than diminish it. Educational psychologists have long proposed theories of how humans construct knowledge, including Jean Piaget’s (1896-1980) constructivist theory. Piaget argued that learners actively construct meaning by organizing and interpreting experiences into cognitive structures. As AI enters classrooms, foundational learning theories should remain a focal point of educational practice.

Learning occurs in the brain through the formation and strengthening of neural pathways. With each new experience, the brain reorganizes itself, creating connections that either fade quickly or stabilize into long-term memory. These networks integrate words, imagery, emotion, and meaning into unified cognitive structures. Because human thought deeply intertwines with language, the tools that shape language also shape cognition itself.

AI may influence pedagogy in ways no previous educational technology has due to its advanced linguistic capabilities. Since humans rely on language to organize thought, increasing dependence on AI-generated language may affect how individuals generate original ideas. Research by Bartoli et al. (2024) provides insight into the neural foundations of creativity and internally generated thought. Using intracranial EEG recordings and direct brain stimulation in thirteen epilepsy patients, the researchers examined brain activity during tasks involving mind wandering, creative idea generation, and sustained visual attention. Their findings showed that the Default Mode Network (DMN)—a brain network associated with internally directed cognition, imagination, and spontaneous thought—was activated early during creative idea generation. When activity within this network was dampened through stimulation, participants retained the ability to identify conventional uses for objects but demonstrated reduced ability to generate novel or creative ideas. The study further revealed that creative thinking involved rapid synchronization across multiple brain regions associated with cognitive control and decision-making (Friesen, 2024). These findings suggest that originality of thought emerges from complex internal neural coordination, raising important questions about how increasing reliance on AI-generated content may dampen human creativity and independent thought.

Educators who are aware of the potential risks of using AI can teach to foster deeper knowledge and prevent shallow learning. Emerging research suggests potential risks to human cognition and original thought, which include:

Cognitive offloading.

Cognitive Load Theory (CLT) posits that human working memory has a limited capacity, so overloading it can overwhelm learning (Sweller, 1988). According to an MIT study (2025), as demonstrated by an EGG, student participants who used a Large Language Model (LLM)  before writing an assignment showed less neural connectivity than those who first thought about the content and then used an LLM. Judges of the participants’ work also scored the LLM first group’s content higher than the other groups. The authors noted, “The convenience of instant answers that LLMs provide can encourage passive consumption of information, which may lead to superficial engagement, weakened critical-thinking skills, less deep understanding of the materials, and less long-term memory formation” (Kosmyna et al., p. 12). ). LLMs may reduce mental effort but compromise depth of learning (Stadler et al., 2024).

Hivemind.

Hivemind is a concept in which LLMs produce similar outputs or content for users. The similarity is due to computer scientists training chatbots on similar data sets. According to a study by Jiang et al. (2025), LLMs fail to produce diverse, open-ended information, as evidenced by intra-model repetition and inter-model homogeneity across 70 LLMs. Intra-model repetition is where LLMs fail to produce diverse outputs. Inter-model homogeneity is where different LLMs produce similar outputs. Using a pairwise statistical analysis of the outputs and qualitative analysis, the researchers reported that the LLMs produced highly repetitive outputs, sometimes with overlapping phrases, and that their responses showed similarities. These concerns have led researchers to question the homogenization of human thought and conformity and, since the introduction of LLMs, have shown that the diversity of human writing styles has declined (Sourati et al., 2025).

Critical thinking. 

Critical thinking is a deliberate process of applying scientific thought to what we read or see before forming a conclusion (Ennis, 1964). A critical thinker will apply more scientific reasoning to information in the pursuit of truth rather than relying on unquestioned claims. A person can apply critical thinking to claims made by others as well as apply it to one’s own thoughts. With limited research, it is assumed that when information and answers are readily available from AI, learners may not take the hard road and apply critical thinking to its generated content, so knowledge may be reduced as critical thought is omitted from the cognitive process. 

Synthesis of Information.

LLMs have an interface that provides a synthesis of information. AI-generated synthesis of information is completed by the time the end user reads it, eliminating the need for the user to perform the cognitive task of synthesis themselves. This may ease the cognitive burden or complexity for the user, but there may be a cost in not synthesizing information for long-term memory. Students may offload other cognitive processes, such as metacognition, understanding, critical thinking, and deep engagement with content, when they do not use LLMs effectively for learning.

Echo Chamber Effect.

According to Kosmyna et al. (2025), an echo chamber is “where a person becomes trapped within information environments that reinforce existing beliefs while filtering out contradictory evidence” (p. 21). Voices may be excluded from LLM output because AI developers created algorithms designed to predict the most probable “token” in a sequence. Much like social media, information provided by LLMs may not reflect the public’s expressions or opinions, but instead reinforce similar narratives (Avin et al., 2024). This can limit a student’s ability to learn about alternative and diverse perspectives.

Shallow vs Deep Level of Processing Information.

According to the Level of Processing model developed by Fergus I. M. Craik and Robert S. Lockhart (1972), the durability of memory depends on the depth of mental processing involved. If the brain processes information at a minimal level (e.g., rote memorization or scrolling on a digital device for information that is not processed), then that kind of shallow learning may not lead to memory trace and consolidation compared to the deepest level of processing. The stability of the memory trace may influence the availability of information to retrieve from long-term memory (Craik & Tulving, 1975).

Aside from these factors, using AI as a tool without a purpose and a strategy for acquiring knowledge prevents real learning. Furthermore, with an effective strategy, educators can measure student learning in this new educational setting. A 2026 NACE report lists the AI-ready skills employers want employees to possess as:

• Identify and use an AI tool to complete tasks.
• Develop effective AI prompts that result in quality outputs.
• Analyze and revise AI’s output as needed.
• Develop and use AI tools to increase work productivity.

To meet these new demands in the workplace, AI learning theories and models must include opportunities for students to develop original thought, deeper long-term memory, metacognitive, critical thinking, and problem-solving skills. NACE suggests that higher education institutions incorporate AI into the curriculum and that educators stay abreast of their respective industries. AI in the workplace will entail humans overseeing AI and a symbiotic relationship to complete tasks. Therefore, educators need ideas for using it and for overseeing it so they can transfer their knowledge to students. To do that, educators must be savvy in their respective industries, as they are responsible for helping students build equivalent knowledge so that they have the necessary skills to oversee AI in the workplace. With an effective AI learning framework, students can reach intellectual milestones while learning to incorporate AI to meet the needs and demands of their current jobs and future roles.

AI can produce erroneous information, so human oversight is essential to ensure accurate decision-making. Without proper knowledge, people will automatically believe AI’s output as being correct and effective, with no need to think critically about it. For a person to oversee the output, they must engage in critical thought, and the level of knowledge must align with the expected ability to oversee AI’s outputs. Dreyfus and Dreyfus’ (1980) model of acquisition postulates five progressive levels of learning: novice, advanced beginner, competent, proficient, and mastery. If a worker is required to oversee AI’s output at a proficient level, that person must have acquired proficient-level knowledge and experience; otherwise, there is a mismatch that increases the potential for a breakdown in oversight.

Two-year higher education institutions can expect students’ knowledge to range from advanced beginner to proficient. Students can still master material at the associate or bachelor level, but in some industries, mastery is at an expert level, and becoming an expert can take years of training and education, as in medicine and psychology. Educators must evaluate their students’ knowledge levels to determine learning expectations that align with the ability to oversee AI.

It is equally important that, in the learning process, students manage metacognition and develop good critical-thinking skills and decision-making skills. This is critical for questioning AI’s output, identifying potential errors, addressing and/or fixing errors, or overseeing AI to make the necessary changes. The factors mentioned here are incorporated into the AI-learning framework below (see Image 1).

Image 1: AI-Learning Hierarchy

Level One: Human Thought, Ideas, Decisions, and Solutions

All learning objectives must start with humans having original thoughts and ideas about the subject or topic matter. This lays the foundation for students to engage in the learning process at the onset through thought, words, and language. People need to think and engage in the struggle to produce words to remember them (Art of the Problem, 2025). Original thought is a hallmark of humanity that sets us apart from AI.

Instructors should guide students by first asking low-level questions. For example, when teaching psychology, I ask students the following prior to a lecture on the material. What comes to mind when I say the word adapt? Do you recall a time when you experienced stress – what was that like? Do you think the unconscious mind exists? Why? Although a student may have limited or no concrete knowledge about the subject/topic, it is still important for the educator to guide low-level questioning to help the brain begin schema-building and foster curiosity.

Level Two: Human-AI Expansion of Knowledge and/or Skill

Students use AI to expand their knowledge of the subject/topic. The student is moving away from Level One to Level Two, where schemas expand. Two things are important at this level: AI literacy and metacognition. AI-literacy centers on effective prompting. Prompting should not influence a hivemind or cognitive offloading, but facilitates the acquisition of knowledge.

The end-user is trained to prompt AI systems, enabling a symbiotic AI-human learning process. The student prompts AI for initial information. The student then reads and comprehends the information, assimilating it into an existing schema. To avoid the risks mentioned above, the student can learn to prompt AI multiple times. The following are prompts that lead to higher learning outcomes, and the examples are based on social science. Listing a word count requirement in the prompt will help ensure the AI’s output reflects deeper learning.

Prompt 1:

Gather Initial Information.

The user prompts AI for initial information about a topic. The prompt should include who you are and what information you want.

 Example:

I am a student learning about the history of psychology, what behavior is, and the seven major perspectives on human behavior. Tell me about these topics based on Dr. Maschio’s textbook: General Psychology: Foundations and Perspectives of the Human Mind and Behavior. Minimum 600 words.

Prompt 2: Dive Deeper.

Take AI’s output and read it with the goal of prompting another question. A question to help “dive” deeper into the material.

 Example:

Tell me more about (ABC concept) and provide research to help me learn more about the topic. Minimum 600 words.

Prompt 3: Explanations.

The user prompts AI to explain, elaborate, and support how concepts work.

Example:

Explain the following concepts and how they apply to the field of psychology (each minimum of 250 words) with examples. Provide sources in APA format:

• What is psychology?
• What is behavior?
• Wilhelm Wundt.
• William James.
• Charles Darwin.
• Sigmund Freud.
• Watson and Skinner from the Behaviorist perspective and their work and contributions (classical and operant conditioning)
• Information processing model.

 Prompt 4. Critical Thinking.

The user prompts AI to ask them questions about what they should know about the concepts or subject.

Example:

Ask me five questions about these topics, and I will reply to them. Then, the user should prompt the AI for clarification of the provided answers and confirm whether the student answered correctly. Next, the user prompts the AI to provide a complex problem for them to reason about based on their level of knowledge. The “conversation” between the user and the AI may continue, encouraging more critical thinking across different scenarios.

 Prompt 5: Metacognition.

The user prompts the AI to help them identify what they still need to learn about the topic.

Example:

From the last prompt, help me identify what I know and what I do not know about these topics.

Prompt 6: Opposing and Diverse Views.

The user prompts AI to help them become familiar with opposing and diverse views.

Example:

List and explain opposing views to [     ]. Then, the user can take this prompt further to increase their knowledge.

Effective prompting does not stop once AI generates content. It also requires information to be assimilated into cognitive schema so that neural networks of knowledge can expand and become more integrated. Prompt #5 addresses this factor. Effective prompting and content assimilation depend heavily on metacognitive skills. In 1979, John H. Flavell proposed that humans possess a limited ability to reflect upon and monitor their own cognition. Flavell argued that metacognition involves both awareness of one’s knowledge and awareness of one’s cognitive experiences. This includes the ability to evaluate whether one utterly understands information, recognize confusion or gaps in comprehension, and revise existing knowledge structures to integrate new information into more organized and unified cognitive frameworks.

Metacognition is an important metaskill for both learning and problem-solving because it enables learners to recognize which cognitive strategies are needed to solve problems effectively (Mayer, 1998). Metacognition strategies aim to strengthen learning by giving individuals greater agency over their thinking processes and self-regulation. In AI-mediated environments, these skills may become increasingly important, as independent thinking and self-monitoring can help workers and students critically evaluate AI systems, assess AI-generated outputs, and use AI effectively without becoming cognitively dependent on it.

Metacognition is important for the success of stages 2, 3, and 4 of the AI-Learning framework presented here. The process of acquiring new knowledge involves struggling with new thoughts and ideas. Producing new thoughts and ideas, and expanding on them, may feel uncomfortable because it requires deeper thinking, focus, and attention. And yet, it is an essential process that makes associated schemas more sophisticated. For the educator, it means directing students to reflect on what they know and do not know so that they can begin to identify information that leads to a deeper understanding of the subject matter.

Level Three: Human-AI  Critical-Thinking & Decision-Making for Problem Solving

Knowledge alone is not sufficient to oversee AI’s output, but it must align with the skill level of the job. Humans will also need to possess strong critical thinking and problem-solving skills. Once learners have reached a certain level of knowledge, they can begin to work symbiotically with AI to solve problems and make sound decisions. But to do so, humans must Socratically think by asking essential questions, reasoning well, thinking logically, and determining the best solution. In 1960, J. C. R. Licklider laid the foundation for the term “Man-Computer Symbiosis”. Licklider, a psychologist and computer scientist, envisioned a future in which humans and machines work together. Like a partnership, humans and computing machines work closely to be as effective as possible.

Level Four: Human Oversight & Productivity

According to NACE (2026), future workers will need the ability to use AI effectively in the workplace. Businesses across industries are increasingly integrating AI systems, such as AI agents. AI agents can execute multiple tasks, make decisions, and automate processes across departments, such as customer support and supply chain management. Organizations are adopting these systems to improve efficiency, workflow, and operational costs. The global AI agent market, valued at around $5-7 billion in 2024-2025, is projected to grow exponentially by 2023 to over $50 billion (Grandviewresearch, n.d.). As AI systems become more embedded within organizations, they will require workers with sufficient expertise to oversee, evaluate, and collaborate effectively with AI technologies.

In this emerging environment, individuals with proficient or expert-level knowledge in their field may become increasingly valuable. AI itself may assist individuals in acquiring essential knowledge for problem solving and innovation (WEF, 2025). Educational institutions, therefore, have an opportunity not only to teach students AI skills but also to use AI to develop advanced cognitive and metacognitive skills necessary for working alongside AI.

Research on expertise demonstrates that experts process information differently from novices. Experts tend to recognize patterns more efficiently, solve problems more rapidly, and employ forward-thinking strategies to anticipate outcomes. Patel and Groen (1986), for example, found that physicians used forward reasoning by moving from patient symptoms toward a diagnosis, whereas medical students were more likely to begin with assumptions and work backward to determine whether symptoms fit those assumptions. Similarly, Adriaan de Groot (1966) found that master chess players searched multiple moves ahead and retained meaningful board patterns far longer than novices. Research by Schempp and Woorons (2018) further showed that expert tennis coaches recognized critical movements and strategic patterns more effectively than novice coaches. Together, these findings suggest that expertise involves highly organized cognitive structures, pattern recognition, strategic anticipation, and the ability to evaluate information efficiently—skills that may become increasingly important in AI-mediated workplaces.

Educational institutions can play a central role in helping students acquire AI-related competencies by integrating AI into learning as a personalized, adaptive educational tool. However, overseeing AI systems requires more than technical familiarity; it requires the cognitive ability to critically evaluate AI-generated outputs. AI developers have acknowledged that AI systems can generate biased or inaccurate information. Biases may emerge when training data fail to adequately represent broader populations, with real-world consequences if left unaddressed (Chapman University, 2026). AI systems have also demonstrated the capacity to hallucinate by generating fabricated news, non-existent court cases, and fake academic references. These hallucinations can pose significant threats in critical sectors like healthcare if left unchecked or unmonitored by a human.

As a result, future workers must develop the ability to identify flawed outputs before acting upon AI-generated information. Pedagogy that integrates AI into the learning process while simultaneously strengthening human cognition, metacognition, critical thinking, and expertise development may help education adapt to these emerging workforce demands. An AI learning framework,  such as the one proposed here, may help educational institutions prepare students for a future in which humans and AI systems operate in a collaborative, interdependent relationship grounded in human oversight, judgment, and expertise.