Pedagogical Discernment in the Age of Generative AI: The BAIM Framework for Instructional Design

Darin M. Mather

doi:10.20944/preprints202607.1406.v1

Submitted:

17 July 2026

Posted:

20 July 2026

You are already at the latest version

Abstract

The rapid adoption of generative artificial intelligence (AI) in higher education has created new opportunities for learning while raising concerns about cognitive offloading, reduced mental effort, and the erosion of foundational academic skills. Although existing research highlights both the benefits and risks of AI, it provides limited guidance for instructors making practical decisions about when AI should be encouraged or constrained in instructional design.This conceptual article introduces the Bloom-Aligned AI Integration Model (BAIM), a pedagogical framework that serves as a decision-making model for AI use by aligning AI integration with the cognitive categories of Bloom’s Revised Taxonomy. The model distinguishes between independent (AI-constrained) and AI-collaborative learning activities and assessments across all levels of cognition. Bridging learning science and AI integration, BAIM draws on research related to cognitive effort, metacognition, and skill development to position AI as a variable that must be calibrated according to instructional goals. In doing so, it offers actionable instructional design guidance by linking cognitive objectives to decisions about AI use, enabling instructors to support both independent student learning and responsible engagement with AI in higher education.

Keywords:

generative artificial intelligence

;

instructional design

;

bloom’s taxonomy

;

assessment

;

higher education

;

AI integration

Subject:

Arts and Humanities - Other

Introduction

The rapid emergence of generative artificial intelligence (AI) has introduced both significant opportunities and complex challenges for higher education. Tools such as large language models are now capable of performing a wide range of cognitive tasks traditionally associated with student learning, including summarizing information, generating explanations, solving problems, and producing written work. As a result, long-standing assumptions about how students learn, demonstrate knowledge, and develop academic skills are being fundamentally reexamined. While early responses to AI in education often focused on concerns related to academic integrity and misuse, a growing body of research suggests that AI can also enhance learning when it is thoughtfully integrated into instructional design.

At the same time, this emerging literature highlights an important tension. On one hand, AI has been shown to support personalized learning, provide immediate feedback, and facilitate higher-order thinking when students actively engage with its outputs. On the other hand, unstructured or excessive reliance on AI may lead to cognitive offloading, reduced mental effort, weakened metacognitive awareness, and the erosion of foundational academic skills. These competing findings suggest that the educational value of AI does not reside in the technology itself, but in how its use is aligned with learning goals and cognitive processes.

Bloom’s Taxonomy of Educational Objectives has long served as a foundational framework for organizing learning outcomes, designing instruction, and aligning assessment with cognitive development. Over time, the taxonomy has been revised and extended to reflect advances in cognitive science and the integration of digital technologies. More recently, scholars have begun to explore how Bloom’s framework might be adapted for AI-mediated learning environments. These efforts have included mapping AI tools to cognitive tasks, reconceptualizing cognition as a human–AI partnership, and identifying risks associated with AI-supported learning. While these contributions provide valuable insights, they often emphasize either the potential of AI integration or its associated risks, with limited guidance for instructors making practical, day-to-day decisions about when AI should be encouraged, constrained, or integrated within specific assignments.

This gap points to the need for a pedagogically grounded framework that aligns decisions about AI use with clearly defined learning objectives. In particular, instructors require guidance that not only recognizes the dual role of AI as both a support and a potential substitute for cognitive work, but also provides concrete direction for structuring learning activities and assessments accordingly. Addressing this need, this article introduces the Bloom-Aligned AI Integration Model (BAIM), a framework that aligns AI use with the cognitive categories of Bloom’s Revised Taxonomy while distinguishing between independent (AI-constrained) and AI-collaborative modes of learning.

The BAIM model reframes AI not as a tool to be uniformly adopted or restricted, but as a pedagogical variable that must be calibrated to the intended learning outcome. By explicitly linking cognitive goals to decisions about when AI should be integrated and when it should be limited, the model provides instructors with a structured approach to designing assignments that preserve essential cognitive processes while leveraging the benefits of AI. In doing so, BAIM seeks to bridge the gap between theoretical discussions of AI and cognition and the practical realities of instructional design in contemporary higher education.

Literature Review

Bloom’s Taxonomy of Educational Objectives

Bloom’s Taxonomy of Educational Objectives is a framework for classifying levels of thinking and learning. Originally developed 70 years ago by educational psychologist, Benjamin Bloom (1956), it remains one of the most influential and frequently referenced curriculum design and assessment resources in the field of education (Houghton et al., 2004).

Organized hierarchically, Bloom's Taxonomy places cognitive skills onto a continuum that moves from lower-order to higher-order thinking. The taxonomy is divided into six ascending levels: knowledge, comprehension, application, analysis, synthesis, and evaluation. At the base, knowledge involves simple recall and recognition of facts. Comprehension requires understanding and translating these facts. Application involves using the learned material to solve new problems. Analysis entails dissecting information into simpler parts and understanding their interrelationships. Synthesis requires the integration of different elements to form a coherent whole, often resulting in innovative solutions. Lastly, evaluation is about critiquing and making judgments based on established criteria. (Bloom, 1956; Seaman, 2011).

Bloom’s model is recognized for its adaptability across a wide range of education contexts and for its role in promoting higher-order thinking skills in classroom instruction. It has been used by educators to design curriculum, write learning objectives, create assessments, and structure classroom activities that intentionally develop deeper levels of thinking. The taxonomy also provides a common language for educators to describe learning goals, helping faculty align instruction with assessment, and promoting more intentional course design (Hmoud and Shaqour, 2024).

Although Bloom’s Taxonomy, in its original form, has been a staple in education for decades, over the years there have been a number of efforts to improve the model by revising, reinterpreting or adapting it to new educational contexts. Fink (2003) reconceptualized learning as a holistic, integrated process that is not purely cognitive. Drawing from this broader perspective on learning, Fink incorporated affective and interpersonal dimensions such as caring, human interaction, and learning how to learn into the taxonomy. Later, Marzano and Kendall (2007) also extended and reconceptualized Bloom’s Taxonomy. They introduced a systems-based model focused on the internal mechanisms that govern learning including motivation, metacognitive regulation, and cognitive processing. While these and other scholars have contributed to the ongoing development of Bloom’s Taxonomy as a tool for learning, perhaps two of the most impactful revisions, and the most germane to this article, are the Revised Bloom’s Taxonomy and Bloom’s Digital Taxonomy.

Revised Bloom’s Taxonomy

In the early 2000’s, Lorin W. Anderson and David R. Krathwohl (2001) introduced an updated version of Bloom’s model, called the Revised Bloom’s Taxonomy (RBT) in response to ongoing developments in cognitive psychology and educational theory. Anderson, a former student of Bloom, and Krathwohl, one of the original contributors to the 1956 taxonomy, sought to revise the model to better reflect contemporary understandings of learning and cognition. The RBT maintained the general hierarchical structure of the original taxonomy, but introduced several important conceptual and structural changes intended to increase its clarity, flexibility, and applicability in modern educational contexts.

One of Anderson and Krathwohl’s key modifications was to the terminology that Bloom’s original model used to describe the different levels of learning. The original taxonomy used noun-based categories—Knowledge, Comprehension, Application, Analysis, Synthesis, and Evaluation. The RBT converted these nominal categories into verbs (Remember, Understand, Apply, Analyze, Evaluate, and Create) to emphasize that learning is an active, not a static process. In addition, the highest levels of the taxonomy were reordered, placing Create above Evaluate. This change recognized that the generation of new ideas or products often represents a more complex cognitive process than the evaluation of existing ones (Jain and Samuel, 2025).

Anderson and Krathwohl also introduced a new dimension to the taxonomy that distinguished between different types of knowledge. While the original taxonomy focused primarily on cognitive processes, the revised model incorporated a knowledge dimension consisting of four categories: factual knowledge, conceptual knowledge, procedural knowledge, and metacognitive knowledge. This two-dimensional structure allowed educators to classify educational objectives according to both the type of knowledge being addressed and the cognitive process required. This revised taxonomy provided a more nuanced framework that helped instructors better align learning goals, instructional strategies, and assessment practices. (Urgo et al., 2019). As such, it provided a common language for discussing levels of cognitive engagement that is widely applied in educational research, curriculum planning and instructional design (Jain and Samuel 2025).

Bloom’s Digital Taxonomy

Soon after the RBT was developed, Andrew Churches (2008) introduced Bloom’s Digital Taxonomy (BDT) to describe the ways that digital technologies could enhance higher-order thinking skills. In BDT, Churches mapped common digital activities and tools onto the RBT’s cognitive processes in an attempt to help instructors connect traditional learning objectives to various digital activities. For example, remembering is associated with bookmarking, highlighting, or simple searching with higher-order processes, such as creating are associated with podcasting, coding, and digital content production. In this sense, BDT functions more as an extended pedagogical guide for integrating technology into instruction than as a new taxonomy of learning itself.

BDT has been widely applied to technology-enhanced learning, instructional design, and digital literacy. It has helped educators move beyond using technology merely for content delivery by encouraging the intentional alignment of digital activities with learning objectives. BDT has also helped highlight the importance of digital skills—such as information filtering, online collaboration, and digital content creation—as integral components of modern learning environments.

While BDT has made many contributions, research on technology in education has raised concerns about the technologically integrated learning that it espouses. Some scholars argue that introducing digital tools risks conflating technological activities with cognitive processes, implying that the use of a particular tool necessarily reflects a specific level of thinking (Kirschner and De Bruyckere 2017; Selwyn 2016). The rapid pace of technological change also raises questions about the durability of frameworks like BDT, since specific tools and digital practices can quickly become outdated Selwyn 2016). As a result, Bloom’s Digital Taxonomy is likely best seen as a useful heuristic for thinking about technology integration rather than as a formal revision of Bloom’s cognitive framework.

Bloom’s Taxonomy - AI Integrations

The widespread availability of AI large language models and their rapid and wide-ranging impact on education has prompted a number of scholars to extend, revise, or reinterpret Bloom’s Taxonomy for AI-mediated learning environments. While these efforts share a recognition that generative AI has altered traditional assumptions about cognition, knowledge production, and instructional design, their approaches vary widely in focus, ranging from tool alignment and cognitive reconceptualization to diagnostic critiques and new process-oriented models. The paragraphs below summarize the scope of this new scholarship.

One stream of inquiry focuses on aligning AI capabilities with instructional tasks across Bloom’s levels. Hmoud and Shaqour (2024) propose an AIEd taxonomy that replaces Bloom’s cognitive categories with AI-functional stages such as collecting in place of remembering, adapting in place of understanding and innovating in place of creating. Similarly, Faraon, Granlund, and Rönkkö (2023) advocate for “AI-driven practices” by mapping specific AI tools—such as conversational agents, tutoring systems, and generative platforms—to each level of Bloom’s Digital Taxonomy. Their work aims to help educators intentionally use AI to support different types of learning outcomes by aligning different AI tools with Bloom’s cognitive levels. While these technology-integration models provide valuable insights into how AI can support learning, they primarily emphasize the academic potential of AI tools and offer limited guidance on when AI use should be restricted to promote cognitive development.

A second body of literature seeks to reconceptualize cognition itself in light of AI. Panthalookaran (2025) proposes a taxonomy for “AI-natives” that moves beyond Bloom’s original assumption that humans perform all cognitive tasks themselves to an acknowledgement that AI can perform many of these (especially lower and mid level) tasks and that much future cognition will be a human-AI collaboration. Based on these assumptions about new approaches to learning, Panthalookaran reimagines Bloom’s for AI-collaborative and AI-complementary modes of thinking. Jain and Samuel (2025) similarly reconceptualize Bloom’s for AI-mediated learning. Using the Revised Bloom’s Taxonomy as a foundation, they suggest new categories for a dual approach to “co-piloted” learning that balances AI’s computational capabilities with human judgement. These approaches offer important theoretical insights into the emerging nature of AI-mediated cognition. However, they often remain abstract, providing limited guidance for instructors designing specific assignments within existing curricular frameworks. They also fail to provide clear instructional guidance for when AI should be constrained versus integrated.

Complementing these conceptual models are diagnostic and risk-oriented perspectives that highlight potential drawbacks of AI integration through reinterpretations or extensions of Bloom’s Taxonomy. Gonsalves (2024) examines the impact of generative AI on critical thinking by revisiting Bloom’s higher-order categories, arguing that overreliance on AI may weaken higher-order cognitive skills by shifting effort away from independent reasoning. Similarly, Asbari (2025) extends Bloom’s framework by introducing the concept of a “Zero Order Thinking State” (C0), a level positioned below the traditional taxonomy, to describe passive engagement in AI-mediated environments, emphasizing the risk of cognitive disengagement. These works contribute important cautionary insights but remain largely problem-focused, offering limited prescriptive guidance for instructional design.

A third line of inquiry examines the distinction between individual and AI-supported cognition. Ayodele et al. (2026) propose the Augmented Cognition Framework (ACF), which conceptualizes each level of Bloom’s Taxonomy as operating in two modes: individual (unaided) and distributed (AI-supported). The model also introduces a seventh level, orchestration, which captures the ability to manage and evaluate human–AI interaction. ACF represents a significant conceptual advance by recognizing human cognition is both internal and AI assisted and by generating learning outcomes for each mode. However, its primary orientation is toward modeling cognition and assessment rather than specifically guiding instructional design. It clarifies what should be assessed but leaves underdeveloped how instruction should be systematically organized in response. It also does not specify when educators should privilege individual cognition over distributed cognition, nor how this distinction should shape assignment development.

Finally, process-oriented models explore how learners interact with AI systems during learning. Elim (2026) proposes a reflective model in which students engage in cycles of questioning, AI interaction, reflection, and refinement, using Bloom’s Taxonomy as a scaffold for cognitive engagement. Ng and his co-authors (2021) similarly emphasize the importance of AI literacy, proposing a framework of competencies—including understanding, application, evaluation, and ethical awareness—grounded in broader literacy traditions and informed by cognitive frameworks such as Bloom’s Taxonomy. Hui (2024) provides empirical evidence of how students’ cognitive processes manifest in AI-supported environments, demonstrating that generative AI can stimulate a range of cognitive activities while also revealing patterns of reliance on lower-order thinking in complex tasks. While these models offer valuable insights into learning processes and AI-supported cognition, they generally assume AI integration and do not address when AI use should be intentionally constrained.

Taken together, these studies demonstrate a growing recognition that AI reshapes both the nature of cognition and the structure of learning activities. However, they share a critical limitation. Existing approaches either (1) focus on how AI may be integrated into learning, (2) reconceptualize cognition in AI-augmented environments, or (3) identify risks associated with AI use. Few provide clear, actionable guidance for instructors making day-to-day instructional decisions about when AI use should be encouraged or constrained within specific assignments, and if utilized, how it should be structured. This gap points to a missing layer in the literature: a pedagogically grounded decision-making model that aligns AI use or constraint with cognitive objectives and provides instructors with clear guidance for structuring assignments.

While the preceding literature highlights the need for more structured guidance on AI use in instructional design, a complementary body of research examines the underlying pedagogical principles that should inform such decisions. In particular, scholarship in learning science provides insight into when AI augmentation may enhance learning and when they may interfere with cognitive processes. The following section reviews this literature, focusing on the conditions under which AI use should be constrained and the contexts in which it can be productively integrated to enhance learning.

Potential Benefits of AI-Assisted Learning

Although early academic discussion of generative artificial intelligence often centered on academic dishonesty, dependency, and epistemic reliability, a growing body of research suggests that AI also has the potential to enhance learning when it is thoughtfully integrated. Several potential benefits have emerged in recent studies, including personalized learning, immediate feedback, support for higher-order thinking, academic writing assistance, development of professional competencies, and expanded access to learning resources.

One of the most consistently identified benefits of AI in education is its capacity to support personalized and adaptive learning (Maphalala and Ajani 2025). AI-driven systems can analyze student performance and adjust instructional activities in response to individual needs, pacing, and knowledge gaps. Zhou (2023) found that students using an AI-based personalized learning platform demonstrated significant improvements in academic performance across a range of disciplines, including mathematics, computer science, English, sociology, and management. Similarly, Zouhaier (2023) argues that AI positively affects the learning experience by enabling more tailored instruction and offering timely responses to students’ needs. Kenchakkanavar (2023) summarizes the personalizing potential of AI by noting that it can analyze student performance to identify learning gaps and adapt instruction in real time. It also enhances engagement by creating more interactive and tailored learning experiences through tools like virtual instructors, gamification, and simulations.

A related benefit is the provision of immediate formative feedback. Feedback has long been recognized as a central component of effective learning (Pereira et al., 2016; Sancho-Vinuesa et al., 2013), and AI tools can make such feedback more rapid and accessible than is often possible in conventional classroom settings. Hooda and colleagues (2022) present an exploratory and comparative study of how AI can enhance student learning through assessment and feedback. They highlight the importance of timely and effective feedback, showing how AI feedback and learning analytics can support improved learning outcomes and student engagement. Two other meta-analyses (Steenbergen-Hu and Cooper, 2014; Ma et al., 2014) found that AI intelligent tutoring systems (ITS), which operate by supplying students with immediate, or near-immediate feedback, outperform other instructional methods across a wide range of disciplines. ITS had a significant positive effect on learning when compared to traditional classroom instruction, reading print or digital material, homework assignments and other instructional methods. Of all the methods examined, only individual human tutoring compared to the outcomes derived from ITS, with one (Ma et al., 2014) finding no significant difference between the two and the other (Steenbergen-Hu and Cooper, 2014) finding ITS less effective than individualized human tutoring.

Research also suggests that AI may support higher-order thinking when students actively engage with it rather than passively accept its outputs. Daniel, Msambwa, and Wen’s (2025) review of 158 studies reports consistent improvements in analytical reasoning, critical thinking, and metacognition in higher education, with stronger outcomes when students interact with AI critically. Supporting this, a meta-analysis of experimental studies (Zhao et al., 2025) finds that generative AI has a moderate positive effect on higher-order thinking—particularly in problem solving and critical analysis—especially in structured learning contexts over sustained periods. Additional reviews (Patrick et al., 2025) indicate that AI can function as a collaborative learning partner, promoting exploration, feedback, and application of knowledge, though its effectiveness depends on thoughtful and critical student engagement. Together, these findings suggest that, with appropriate structure and guardrails, AI has the potential to serve as a productive cognitive partner when students are required to question, interpret, and evaluate its outputs.

Another area of promise is academic writing support. Recent studies suggest that AI writing tools hold significant promise for improving student writing when used as a support for feedback and revision. Evidence from reviews and empirical studies indicates that AI-assisted writing environments can enhance grammatical accuracy, coherence, organization, and argumentation, while also providing timely, individualized feedback that supports iterative drafting and revision ( Deep and Chen, 2025; Malik et al., 2023; Nazari et al., 2021). In addition, these tools appear to strengthen the writing process itself by promoting student engagement, self-efficacy, and autonomy, particularly through support for brainstorming, outlining, and self-editing (Sanz-Tejeda et al., 2025; Yang et al., 2025). Importantly, the literature consistently emphasizes that these benefits are conditional. AI is most effective when embedded within structured instructional contexts that require students to interpret and apply feedback rather than simply accept generated text (Sanz-Tejeda et al., 2025; Yang et al., 2025). At the same time, scholars caution that improper use of AI writing assistants may hinder learning by reducing cognitive effort, weakening independent writing skills, and encouraging overreliance on generated content, thereby limiting the development of critical thinking and metacognitive awareness (Sanz-Tejeda et al., 2025; Yang et al., 2025).

AI use may also support the development of digital, professional, and workplace-relevant competencies. Hasan, Nasreen, and Rasul (2025) found that higher levels of AI exposure were associated with stronger collaboration, communication, and critical thinking skills among students. Reviews of the scientific literature (Daniel et al., 2025; Weng et al., 2024) have supported these findings, reporting that generative AI tools can foster not only technical skills, such as writing, programming and data analysis, but also life and career enhancing interpersonal skills such as teamwork, organization, and communication. Together, these findings suggest that AI-assisted learning may help students develop the kinds of interdisciplinary competencies increasingly required in AI-augmented professional environments.

Potential Harms of AI Use for Learning

While an ever-expanding body of literature suggests that AI tools can enhance student learning in many ways, another growing branch of literature identifies substantial risks associated with unstructured or excessive AI use in educational settings. These risks appear across multiple dimensions of the learning process, including reduced cognitive effort, weakened metacognitive awareness, erosion of higher-order thinking, and disruption of foundational academic skills.

One of the most frequently cited concerns is cognitive offloading and reduced mental effort. Zhai, Wibowo, and Li (2024) found, in a systematic review of the literature, that when AI systems provide immediate answers, explanations, or solutions, students may bypass important mental processes required for deep learning. Miranda and colleague (2025) similarly describe this pattern of cognitive offloading, suggesting that increased dependence on AI may reduce cognitive engagement and contribute to declines in students’ critical academic skills, as learners rely more heavily on AI to perform tasks. These findings are consistent with longstanding cognitive research suggesting that durable learning depends on effortful processing and that bypassing such effort may reduce retention and transfer (Bjork and Bjork, 2020; Nelson and Eliasz, 2023).

A second concern relates to students’ metacognition, their ability to accurately judge what they understand. Research in the learning sciences has long documented the phenomenon of “illusions of competence,” in which learners overestimate their mastery when material is presented fluently or feels familiar (Bjork et al., 2013). AI-generated explanations and polished prose may intensify this effect by providing clear, well-structured responses that can be easily recognized but not necessarily internalized Empirical studies across a wide range of disciplines consistently show a pattern in which students’ engagement with AI-generated responses increase their confidence without improving their understanding (Kumar et al., 2026).

A related concern is the potential decline of students’ thinking skills. When students depend on AI to summarize, explain, argue, or solve problems, it may reduce their willingness and ability to analyze information critically, draw conclusions independently, and form logical arguments (George et al., 2024; Zhai, Wibowo, and Li, 2024). Szmyd and Mitera (2024) found that, when surveyed, 83% of students themselves believed excessive reliance on AI could weaken their ability to think independently and make responsible decisions. Although some research shows that AI can support critical thinking under properly structured conditions, its misuse or overuse may undermine the very cognitive skills it is intended to develop (Girma, 2025). Without intentional pedagogical design, AI may shift learning from active knowledge construction to passive consumption of generated responses.

Finally, the literature also raises significant concerns about the weakening of foundational academic skills, especially in writing, research, and problem solving. Hassan and Funsho (2025) report that AI-supported writing environments can reduce students’ capacity to produce high-quality academic discourse. They link this decline to students’ increased reliance on AI tools, bypassing iterative processes such as drafting, revising, and argument development, which are the underlying skills required for fully developing effective academic writing. Basha (2024) similarly argues that excessive reliance on AI tools may hinder students’ decision-making, problem-solving abilities, and practical skill development. In mathematics education, Opesemowo and Ndlovu (2024) warn that heavy reliance on AI-assisted problem solving may weaken students’ creativity and independent problem-solving skills. These findings are important because they point not merely to performance issues but to interruptions in the developmental processes through which academic skills are formed.

Introducing the BAIM Model

Taken together, the benefits and harms literature suggests that AI is neither inherently beneficial nor inherently detrimental to learning. Rather, its educational value depends on how its use is aligned with instructional purpose and cognitive demand. Research supporting AI integration highlights its value as a scaffold, feedback provider, dialogical partner, and generator of practice opportunities, while cautionary literature emphasizes the importance of internalization, retrieval, productive struggle, and the development of foundational academic skills. This tension underscores the need for a framework to help determine when AI should extend learning and when it should be constrained to preserve essential cognitive ownership.

To address this gap, the Bloom-Aligned AI Integration Model (BAIM) addresses this need by aligning decisions about AI use with Bloom’s cognitive categories. Instead of treating AI as something to be broadly adopted or restricted, BAIM frames AI as a pedagogical variable that must be calibrated to the intended learning outcome. It prompts instructors to determine when AI participation supports learning and when it risks displacing essential cognitive processes.

At each level of Bloom’s Taxonomy, BAIM differentiates between two possible pathways: Independent (AI-constrained) learning, which requires unaided student performance, and AI-Collaborative learning, which incorporates AI as a support for cognitive development. Importantly, the model does not assume a fixed relationship between cognitive level and AI use. Both constrained and integrated approaches may be appropriate at any Bloom’s level, depending on instructional goals.

The model, presented in Table 1, operationalizes this framework by linking each cognitive category to specific instructional decisions. For each level of Bloom’s Revised Taxonomy, it outlines the primary learning goal, the rationale for constraining or integrating AI, and examples of both Independent and AI-Collaborative activities and assessments.

Interpreting the BAIM Model

The BAIM model provides a structured framework for aligning AI use with both learning processes and assessment across Bloom’s Taxonomy. Central to the model is the distinction between independent and AI-collaborative work, applied not only to how students learn, but also to how their learning is evaluated. By differentiating between learning activities and assessed outcomes at each cognitive level, BAIM translates insights from learning science into practical decisions about when AI should support learning and when it should be constrained to preserve essential cognitive processes.

At the level of remembering, the model focuses on the internalization of foundational knowledge through effortful retrieval. Research shows that durable learning depends on actively recalling information rather than simply re-exposing oneself to it. As a result, BAIM prioritizes activities such as self-testing and recall exercises, with assessment requiring unaided demonstration of knowledge. AI can support rehearsal through practice questions and review activities and can also administer assessments by delivering prompts, scoring responses, and providing feedback. In all cases, however, students must generate answers independently so that retrieval is strengthened rather than replaced.

Understanding, by contrast, centers on the construction of meaning through explanation, interpretation, and connection-making. Findings from generative learning research highlight the importance of processes such as self-explanation and paraphrasing in building accurate mental models. Students are therefore expected to articulate ideas in their own words without assistance. AI may contribute by offering alternative explanations, examples, and feedback, and it may also evaluate student-generated explanations during assessment. Still, the responsibility for explaining and interpreting concepts must remain with the student, ensuring that understanding is constructed rather than borrowed.

When students move into application, the focus shifts to transfer—using knowledge and skills to solve problems in new contexts. Learning science indicates that this kind of transfer depends on students independently selecting and executing appropriate strategies, rather than relying on previously demonstrated solutions. Accordingly, in the BAIM model independent work emphasizes problem-solving, case-based tasks, and procedural execution without assistance, with assessment requiring accurate performance in unfamiliar situations. AI can broaden practice by generating varied scenarios, modeling approaches, and offering immediate feedback. In collaborative contexts, however, evaluation centers on how well students assess, adapt, and improve AI-generated responses while still carrying out the work themselves.

At the level of analysis, the emphasis turns to reasoning—identifying patterns, examining relationships, and interpreting structure. Research on cognitive offloading suggests that when AI performs these functions, students may bypass the effort needed to develop analytical skill. For this reason, BAIM foregrounds tasks such as data interpretation, argument deconstruction, and case analysis, with assessment focused on students’ ability to construct and justify their interpretations. AI may be used to supply material for analysis, including data and alternative perspectives, and may also assist in assessment by presenting materials and evaluating reasoning. Even so, students must independently build and defend their analyses so that AI serves as a stimulus rather than a substitute for thinking.

The evaluation level introduces a further shift, emphasizing judgment through the application of criteria and the defense of reasoned conclusions. The literature on critical thinking underscores both the importance of active evaluative engagement and the risks of deferring judgment to external sources. Students are therefore expected to assess arguments, evaluate sources, and defend positions on their own. AI may enrich this process by presenting competing viewpoints and modeling evaluative frameworks, and it may also provide critique during assessment. Nevertheless, the act of applying criteria and justifying conclusions must remain the student’s responsibility, preserving the development of independent discernment.

Finally, creation places the emphasis on authorship, ownership, and the sequencing of the creative process. Research on writing and iterative feedback shows that learning is strengthened through cycles of drafting, revision, and reflection. BAIM reflects this by constraining AI use during initial idea generation to preserve originality, while incorporating it later to support refinement and improvement. Students are responsible for originating and developing their ideas, while AI contributes to revision, expansion, and feedback. Assessment, in turn, focuses on authentic student work and clear evidence of independent authorship, often supported by documentation of the creative process and reflective justification.

Across all levels, BAIM brings into practice a central insight from the literature: AI has the potential to enhance learning when it provides feedback, practice, and cognitive stimulation, but it can undermine learning when it replaces effortful thinking, encourages cognitive offloading, or creates illusions of understanding. Beyond supporting learning processes, AI can also assist in assessment by administering tasks, evaluating student-generated responses, and providing timely feedback. The model addresses this tension by using the distinction between independent and AI-collaborative work as a guiding heuristic for both instruction and assessment design. Rather than treating AI as uniformly beneficial or harmful, BAIM aligns its use with cognitive purpose, ensuring that students develop independent intellectual capacity while also learning to engage AI critically and responsibly.

Implications for Practice

The BAIM model offers practical guidance for instructors, program designers, and institutions seeking to integrate artificial intelligence thoughtfully in higher education. It enables instructors to move beyond blanket protocols of AI prohibition or unrestricted use and instead provides a structured approach to assignment design that aligns AI use with specific cognitive objectives. The model shifts instructional decision-making from a tool-centered to a learning-centered approach, prompting educators to focus on the type of cognitive work required for a task and to determine whether AI participation would support or undermine that work. This approach can be operationalized through relatively simple design questions: What must students be able to do independently? Where might AI provide beneficial scaffolding or feedback? At what stage of the learning process should AI be introduced? Such questions can guide the creation of both independent and AI-collaborative assignments within a single course, allowing for more nuanced and intentional instructional design.

The model also has implications for assessment practices. As AI-generated outputs become increasingly difficult to distinguish from student work, traditional assessments that focus primarily on final products may no longer provide valid evidence of student learning. BAIM highlights the importance of designing assessments that capture process as well as product. This may include in-class performance tasks, oral explanations, iterative drafts, reflective components, or assignments that require students to critique or improve AI-generated content. By intentionally structuring assignments at various cognitive levels with AI in mind, instructors can better assess students’ learning while incorporating AI use as appropriate for the learning objectives.

The model further supports the development of AI literacy in a more meaningful sense than tool proficiency alone. By engaging students in both constrained and integrated contexts, BAIM helps students learn not only how to use AI, but when and why to use it. This includes developing the ability to critically evaluate AI outputs, recognize limitations such as bias or inaccuracy, and make informed decisions about appropriate use. In this way, AI literacy becomes a matter of discernment and judgment rather than mere technical competence.

At the institutional level, the model supports more flexible and coherent AI policies. Rather than applying uniform rules, programs can tailor AI use to disciplinary and course-level needs—for example, emphasizing skill development in foundational courses and more AI integration in advanced coursework. Such an approach allows institutions to move beyond reactive or compliance-based policies toward more pedagogically grounded strategies.

Finally, BAIM has implications for faculty development and professional learning. Many instructors report uncertainty about how to respond to AI in their teaching, often oscillating between permissive and restrictive approaches. The model provides a common language and conceptual structure that can support faculty conversations, workshops, and curriculum redesign efforts. By grounding these discussions in established principles of learning and cognition, BAIM can help educators move from reactive responses to intentional pedagogical practice.

Conclusion

The emergence of generative artificial intelligence marks a significant turning point for higher education, challenging traditional assumptions about learning, assessment, and student work. As the literature suggests, AI is neither inherently beneficial nor harmful. Rather, its educational impact depends on how it is integrated into instructional contexts and how it shapes the cognitive processes involved in learning. While existing research has clarified AI’s relationship to cognition and identified its opportunities and risks, a key gap remains in providing practical guidance for instructors. The BAIM model addresses this need by offering a framework that aligns AI use with Bloom’s Taxonomy and distinguishes between contexts requiring independent thinking and those suited for AI-supported augmentation. Although this model is developed with higher education in mind, its underlying principles may extend to other educational settings where similar tensions between independent cognition and AI-supported learning exist.

By introducing the distinction between independent and AI-collaborative learning activities across all levels of the taxonomy, BAIM translates theory into actionable instructional design. It shifts the focus from managing AI use to aligning its use with learning goals, helping preserve essential processes such as effortful learning and skill development while also fostering students’ ability to engage AI in productive and responsible ways.

As a conceptual and pedagogical framework, BAIM has several limitations. It has not yet been empirically tested across disciplines or institutional contexts, and its effectiveness will depend on how it is implemented in practice, including faculty training, course design, and student engagement with AI tools. Additionally, the model assumes access to AI technologies and a level of institutional support that may not be uniformly available. Variability in disciplinary norms, technological infrastructure, and student readiness may also influence how the model functions in different settings.

Future research is needed to examine the implementation and impact of BAIM in diverse educational contexts. Empirical studies could investigate how different configurations of AI-constrained and AI-integrated assignments affect cognitive development, skill acquisition, and academic integrity. Further work should also explore how students develop discernment in AI-mediated environments and how faculty adapt instructional practices when using the model. In addition, research examining the scalability of BAIM, its applicability across disciplines, and its adaptability to evolving AI technologies would help refine and strengthen the framework.

In conclusion, the integration of artificial intelligence into education requires new forms of pedagogical practice. The BAIM model offers one approach to this challenge by helping educators design learning environments in which students develop both independent intellectual capacity and the ability to engage AI responsibly and effectively. As higher education continues to adapt to the presence of AI, frameworks that support intentional, learning-centered decision-making will be essential for ensuring that technological innovation serves, rather than supplants, the goals of education.

Author’s Statement: Given the focus of this study on AI-integrated learning, the author used an AI-based language model (ChatGPT, OpenAI) as a support tool during manuscript development. The central concept, theoretical framing, and structural design of the BAIM framework, as well as the overall organization and argument of the article, were conceived and directed by the author. The AI tool was used to assist with idea elaboration and language refinement. Outputs generated by the tool were treated as provisional and were critically evaluated, revised, or discarded as appropriate. The author retained full intellectual control over the development of the model and the manuscript and assumes full responsibility for the accuracy, integrity, and originality of the work.

Appendix

Appendix: Sample Activities and Assessments Across the BAIM Model

Bloom’s Category	Independent – Learning Activities	Independent – Assessment	AI-Collaborative – Learning Activities	AI-Collaborative – Assessment
Remember	Low-stakes retrieval quizzes Flashcard recall (no AI) Brain dumps Diagram labeling Timeline reconstruction	Closed-book quizzes/exams Short-answer recall questions Matching or identification tasks Vocabulary/terminology tests Timed recall exercises	AI-generated practice quizzes AI-created flashcards or vocab lists AI mnemonic devices AI-generated diagrams for labeling AI as a quiz partner	AI-administered recall quizzes (no assistance) Adaptive AI retrieval quizzes Timed AI recall exercises AI cumulative review assessments
Understand	Writing summaries in own words Explaining concepts Concept maps (no support) Compare/contrast activities Interpreting graphs/data	Short-answer explanations Concept maps (closed-resource) Paraphrasing assessments Compare/contrast responses Oral explanations	Comparing AI explanations Dialogue with AI for clarification AI critique of explanations AI scaffolding for concept maps	AI-evaluated explanations AI-assisted scoring of short answers AI-analyzed concept maps Adaptive AI questioning
Apply	Solving problems independently Case-based exercises Applying formulas/procedures Structured problem sets Technical/lab tasks	Problem-solving exams Case-based questions Skill demonstrations Timed problem sets Scenario-based responses	AI-modeled problem-solving AI-generated practice scenarios Guided hints (not solutions) AI feedback on work Iterative problem-solving	AI-evaluated problem-solving AI-assisted grading of cases AI simulations Adaptive AI problem sets AI performance reports
Analyze	Breaking texts into parts Identifying patterns/trends Comparing perspectives Analyzing arguments Mapping relationships	Analytical essays Source analysis Data interpretation Case analysis Logical reasoning tasks	Comparing student vs AI analysis AI-generated interpretations Testing AI-identified patterns AI critique of student analysis	AI-assisted evaluation of analysis AI-supported grading AI feedback on reasoning Adaptive AI questioning AI diagnostic reports
Evaluate	Critiquing arguments Evaluating sources Defending positions Ranking using criteria Identifying fallacies	Argumentative essays Source credibility analysis Position papers Debates Ethical decision tasks	Comparing judgments with AI AI-generated viewpoints Evaluate AI arguments AI ethical simulations	AI-assisted evaluation of arguments AI feedback on reasoning Adaptive AI questioning AI diagnostic reports
Create	Writing original work Designing projects/solutions Developing arguments Creating presentations Producing creative/technical work	Research papers Capstone projects Project-based assessments Creative productions	AI-assisted brainstorming Iterative drafting with AI Exploring alternatives AI critique of drafts Simulated audience feedback	AI-supported draft evaluation AI analysis of structure/coherence Revision tracking AI-supported originality checks

References

Anderson, L.W.; Krathwohl, D. R. A taxonomy for learning, teaching, and assessing: A revision of Bloom’s taxonomy of educational objectives; Longman, 2001. [Google Scholar]
Asbari, M. From C0 to C6: Expanding Bloom’s taxonomy to diagnose passive learning in AI-mediated classrooms. Indones. J. Manag. Econ. Res. 2025, 2(02), 38–43. [Google Scholar] [CrossRef]
Ayodele, K. P.; Obayiuwana, E.; Lawal, A. R.; Bamimore, A.; Offiong, F. B.; Peter, E. A. Revising Bloom's taxonomy for dual-mode cognition in human-AI systems: The augmented cognition framework. arXiv 2026, arXiv:2602.00697. [Google Scholar]
Basha, J. Y. The negative impacts of AI tools on students in academic and real-life performance. Int. J. Soc. Sci. Commer. 2024, 1(3), 1–16. [Google Scholar]
Bennett, S.; Maton, K.; Kervin, L. The ‘digital natives’ debate: A critical review of the evidence. Br. J. Educ. Technol. 2008, 39(5), 775–786. [Google Scholar] [CrossRef]
Bjork, R. A.; Bjork, E. L. Desirable difficulties in theory and practice. J. Appl. Res. Mem. Cogn. 2020, 9(4), 475. [Google Scholar] [CrossRef]
Bjork, R. A.; Dunlosky, J.; Kornell, N. Self-regulated learning: Beliefs, techniques, and illusions. Annu. Rev. Psychol. 2013, 64(1), 417–444. [Google Scholar] [CrossRef] [PubMed]
Bloom, B. S. (Ed.) Taxonomy of educational objectives: The classification of educational goals. In Handbook I: Cognitive domain; David McKay Company, 1956. [Google Scholar]
Churches, A. Bloom’s digital taxonomy. 2008. Available online: http://burtonslifelearning.pbworks.com/w/file/fetch/26327358/BloomDigitalTaxonomy2001.pdf.
Daniel, K.; Msambwa, M. M.; Wen, Z. Can generative AI revolutionise academic skills development in higher education? A systematic literature review. Eur. J. Educ. 2025, 60(1), e70036. [Google Scholar] [CrossRef]
Deep, P. D.; Chen, Y. The role of AI in academic writing: Impacts on writing skills, critical thinking, and integrity in higher education. Societies 2025, 15(9), 247. [Google Scholar] [CrossRef]
Elim, E. H. S. Y. Promoting cognitive skills in AI-supported learning environments: the integration of Bloom’s taxonomy. Educ. 3-13 2026, 54(3), 612–622. [Google Scholar]
Faraon, M.; Granlund, V.; Rönkkö, K. Artificial intelligence practices in higher education using Bloom's digital taxonomy. In 2023 5th International Workshop on Artificial Intelligence and Education (WAIE); IEEE, September 2023; pp. 53–59. [Google Scholar]
Fink, L. D. Creating significant learning experiences: An integrated approach to designing college courses; John Wiley & Sons, 2013. [Google Scholar]
George, A. S.; Baskar, T.; Srikaanth, P. B. The erosion of cognitive skills in the technological age: How reliance on technology impacts critical thinking, problem-solving, and creativity. Partn. Univers. Innov. Res. Publ. 2024, 2(3), 147–163. [Google Scholar]
Girma, A. H. The Role of Artificial Intelligence in Shaping Human Interaction and Cognitive Function. Kotebe J. Educ. 2025, 3(1), 69–88. [Google Scholar] [CrossRef]
Gonsalves, C. Generative AI’s impact on critical thinking: Revisiting Bloom’s taxonomy. J. Mark. Educ. 2026, 48(1), 4–19. [Google Scholar]
Hasan, S.; Nasreen, S.; Rasul, S. S. U. Leveraging artificial intelligence (ai) in higher education: fostering soft skills communication, collaboration, creativity and critical thinking among university students. Insights-J. Life Soc. Sci. 2025, 3(4), 1–7. [Google Scholar] [CrossRef]
Hassan, M.; Funsho, A. A. Exploring the Impact of Integrative Artificial Intelligence Tools on the Writing Proficiency of University Students in Kano State. Glob. ONLINE J. Acad. Res. (GOJAR) 2025, 4(1). [Google Scholar]
Hmoud, M.; Shaqour, A. AIEd Bloom’s taxonomy: A proposed model for enhancing educational efficiency and effectiveness in the artificial intelligence era. Int. J. Technol. Learn. 2024, 31(2). [Google Scholar] [CrossRef]
Hooda, M.; Rana, C.; Dahiya, O.; Shet, J. P.; Singh, B. K. Integrating LA and EDM for improving students Success in higher Education using FCN algorithm. Math. Probl. Eng. 2022, 2022(1), 7690103. [Google Scholar] [CrossRef]
Houghton, J.; Steele, C.; Henty, M. Research practices and scholarly communication in the digital environment. Learn. Publ. 2004, 17(3), 231–249. [Google Scholar] [CrossRef]
Hui, E. S. Y. E. Incorporating Bloom’s taxonomy into promoting cognitive thinking mechanism in artificial intelligence-supported learning environments. Interact. Learn. Environ. 2025, 33(2), 1087–1100. [Google Scholar]
Jain, Jasmine; Samuel, Moses. "Bloom meets gen ai: Reconceptualising Bloom's taxonomy in the era of co-piloted learning. 2025. [Google Scholar] [CrossRef]
Kenchakkanavar, A. Y. Exploring the artificial intelligence tools: Realizing the advantages in education and research. J. Adv. Libr. Inf. Sci. 2023, 12(4), 218–224. [Google Scholar]
Kirschner, P. A.; De Bruyckere, P. The myths of the digital native and the multitasker. Teach. Teach. Educ. 2017, 67, 135–142. [Google Scholar] [CrossRef]
Kumar, S.; Mikayelyan, A.; Vorfolomeyeva, O. Fluency Illusion: A Review on Influence of ChatGPT in Classroom Settings. Information 2026, 17(3), 299. [Google Scholar] [CrossRef]
Ma, W.; Adesope, O. O.; Nesbit, J. C.; Liu, Q. Intelligent tutoring systems and learning outcomes: A meta-analysis. J. Educ. Psychol. 2014, 106(4), 901. [Google Scholar] [CrossRef]
Malik, A. R.; Pratiwi, Y.; Andajani, K.; Numertayasa, I. W.; Suharti, S.; Darwis, A. Exploring artificial intelligence in academic essay: higher education student's perspective. Int. J. Educ. Res. Open 2023, 5, 100296. [Google Scholar] [CrossRef]
Maphalala, M. C.; Ajani, O. A. Leveraging artificial intelligence as a learning tool in higher education. Interdiscip. J. Educ. Res. 2025, 7(1), a01–a01. [Google Scholar] [CrossRef]
Marzano, R. J.; Kendall, J. S. (Eds.) The new taxonomy of educational objectives; Corwin Press, 2006. [Google Scholar]
Miranda, J. P. P.; Cruz, M. A. D.; Fernandez, A. B.; Balahadia, F. F.; Aviles, J. S.; Caro, C. A.; Gaña, E. P. Erosion of critical academic skills due to AI dependency among tertiary students: A path analysis. Pitfalls AI Integr. Educ. Ski. Obsolescence Misuse Bias 2025. [Google Scholar] [CrossRef]
Nazari, N.; Shabbir, M. S.; Setiawan, R. Application of Artificial Intelligence powered digital writing assistant in higher education: randomized controlled trial. Heliyon 2021, 7(5). [Google Scholar] [CrossRef] [PubMed]
Nelson, A.; Eliasz, K. L. Desirable difficulty: theory and application of intentionally challenging learning. Med. Educ. 2023, 57(2), 123–130. [Google Scholar] [PubMed]
Ng, D. T. K.; Leung, J. K. L.; Chu, S. K. W.; Qiao, M. S. Conceptualizing AI literacy: An exploratory review. Comput. Educ. Artif. Intell. 2021, 2, 100041. [Google Scholar] [CrossRef]
Opesemowo, O. A.; Ndlovu, M. Artificial intelligence in mathematics education: The good, the bad, and the ugly. J. Pedagog. Res. 2024, 8(3), 333–346. [Google Scholar] [CrossRef]
Panthalookaran, V. A revised taxonomy of educational objectives for AI-natives: reimagining thinking skills in the age of AI. High. Educ. Future 2025, 12(1), 51–64. [Google Scholar]
Patrick, P. M.; Yip, S. Y.; Campbell, C. Artificial Intelligence and Higher-Order Thinking: A Systematic Review of Educator and Student Experiences and Perspectives in Higher Education. High. Educ. Q. 2025, 79(4), e70069. [Google Scholar] [CrossRef]
Pereira, D.; Flores, M. A.; Simão, A. M. V.; Barros, A. Effectiveness and relevance of feedback in Higher Education: A study of undergraduate students. Stud. Educ. Eval. 2016, 49, 7–14. [Google Scholar] [CrossRef]
Sancho-Vinuesa, T.; Escudero-Viladoms, N.; Masià, R. Continuous activity with immediate feedback: A good strategy to guarantee student engagement with the course. Open Learn. J. Open Distance E-Learn. 2013, 28(1), 51–66. [Google Scholar] [CrossRef]
Sanz-Tejeda, A.; Domínguez-Oller, J. C.; Baldaquí-Escandell, J. M.; Gómez-Díaz, R.; García-Rodríguez, A. The impact of generative AI on academic reading and writing: a synthesis of recent evidence (2023–2025). In Frontiers in Education; Frontiers Media SA, 2025; Vol. 10. [Google Scholar]
Seaman, M. Bloom’s taxonomy: Its evolution, revision, and use in the field of education. In Curriculum and teaching dialogue; Flinders, D. J., Uhrmacher, P. B., Eds.; Emerald Publishing Limited, 2011. [Google Scholar] [CrossRef]
Selwyn, N. Education and technology: Key issues and debates; Bloomsbury Publishing, 2021. [Google Scholar]
Steenbergen-Hu, S.; Cooper, H. A meta-analysis of the effectiveness of intelligent tutoring systems on college students’ academic learning. J. Educ. Psychol. 2014, 106(2), 331. [Google Scholar] [CrossRef]
Szmyd, K.; Mitera, E. The impact of artificial intelligence on the development of critical thinking skills in students. Eur. Res. Stud. J. 2024, 27(2), 1022–1039. [Google Scholar] [CrossRef]
Urgo, K.; Arguello, J.; Capra, R. Anderson and krathwohl's two-dimensional taxonomy applied to task creation and learning assessment. In Proceedings of the 2019 ACM SIGIR International Conference on Theory of Information Retrieval, 2019, September; pp. 117–124. [Google Scholar]
Weng, X.; Qi, X. I. A.; Gu, M.; Rajaram, K.; Chiu, T. K. Assessment and learning outcomes for generative AI in higher education: A scoping review on current research status and trends. Australas. J. Educ. Technol. 2024, 40(6), 37–55. [Google Scholar] [CrossRef]
Yang, K.; Raković, M.; Liang, Z.; Yan, L.; Zeng, Z.; Fan, Y.; Chen, G. Modifying AI, enhancing essays: How active engagement with generative AI boosts writing quality. In Proceedings of the 15th international learning analytics and knowledge conference, 2025, March; pp. 568–578. [Google Scholar]
Zhai, C.; Wibowo, S.; Li, L. D. The effects of over-reliance on AI dialogue systems on students' cognitive abilities: a systematic review. Smart Learn. Environ. 2024, 11(1), 28. [Google Scholar] [CrossRef]
Zhao, Y.; Yue, Y.; Sun, Z.; Jiang, Q.; Li, G. Does Generative Artificial Intelligence Improve Students’ Higher-Order Thinking? A Meta-Analysis Based on 29 Experiments and Quasi-Experiments. J. Intell. 2025, 13(12), 160. [Google Scholar] [CrossRef] [PubMed]
Zhou, C. Integration of modern technologies in higher education on the example of artificial intelligence use. Educ. Inf. Technol. 2023, 28(4), 3893–3910. [Google Scholar]
Zouhaier, S. The impact of artificial intelligence on higher education: An empirical study. Eur. J. Educ. Sci. 2023, 10(1), 17–33. [Google Scholar] [CrossRef]

Table 1. The BAIM Model (Bloom-Aligned AI Integration Model).

Bloom’s Category	Learning Goal	Rationale for Inhibiting AI	Rationale for Integrating AI	Independent – Learning Activities	Independent – Assessment	AI-Collaborative – Learning Activities	AI-Collaborative – Assessment
Remember	Recall foundational knowledge	Foundational knowledge must be stored in long-term memory to enable fluency, and support higher-order thinking without reliance on external tools.	AI can accelerate exposure, provide varied rehearsal, and generate retrieval practice, helping students encode and reinforce foundational knowledge more efficiently.	Students complete retrieval tasks such as recall activities, flashcard drills and diagram labeling and in which they produce answers from memory without prompts or assistance.	Students complete unsupported recall tasks such as quizzes and identification exercises to demonstrate their ability to accurately reproduce essential knowledge from memory.	Students use AI to generate practice questions, organize key information, and simulate retrieval exercises to assist the development of independent recall.	Students complete AI-administered recall assessments in which AI delivers prompts, scores responses, and provides feedback, while students generate answers independently, ensuring that performance reflects internalized knowledge.
Understand	Explain and interpret concepts	Understanding requires students to construct meaning by connecting new ideas to prior knowledge; outsourcing this process inhibits the ability to form accurate mental models.	AI can provide multiple explanations, examples, and analogies, helping students compare perspectives, identify misconceptions, and refine their understanding.	Students practice writing summaries, creating concept maps, and generating examples without prompts, to develop their ability to make meaning from the material.	Students complete explanation-based tasks in which they interpret and explain concepts clearly and accurately without assistance.	Students use AI to explore alternative explanations, test their understanding, and identify gaps, then revise their own explanations to reflect deeper, self-constructed meaning.	Students engage with AI-generated content (e.g., critique AI explanations or compare interpretations) in order to assess their ability to independently articulate clear, accurate, and well-reasoned explanations.
Apply	Transfer knowledge to new contexts	Independent application ensures that students can select and execute appropriate methods on their own, demonstrating true transfer rather than reliance on external problem-solving.	AI can generate diverse scenarios, model problem-solving processes, and provide immediate feedback, expanding opportunities for guided practice across varied contexts.	Students solve problems, complete case exercises, and carry out procedures using learned methods in new situations, selecting and executing appropriate steps without assistance.	Students complete application tasks such as problem sets and case responses to demonstrate their ability to correctly apply knowledge and procedures in unfamiliar or varied contexts without support.	Students use AI to observe modeled solutions, practice with varied scenarios, and receive feedback, while still performing the problem-solving steps and decision-making themselves.	Students complete AI-administered tasks in which AI presents problems, evaluates responses, and provides feedback to assess independent application of knowledge.
Analyze	Examine patterns and relationships	Analysis requires students to independently deconstruct information, identify relationships, and form reasoned interpretations; relying on AI can bypass the cognitive processes that develop analytical judgment.	AI can generate complex datasets, alternative interpretations, and contrasting perspectives, providing rich material for students to examine, critique, and refine their analytical thinking.	Students study texts, datasets, or cases to practice identifying patterns, comparing components, mapping relationships, and breaking information into parts without guidance.	Students complete analytical projects such as writing essays, examining data or deconstructing arguments to demonstrate their ability to independently identify relationships, interpret structure, and draw reasoned conclusions.	Students use AI to generate data, surface alternative interpretations, and test hypotheses to enhance their ability to develop and evaluate their analytical skills.	Students complete AI-administered analytical tasks in which AI presents materials, evaluates reasoning, and provides feedback, while students independently construct and defend their analysis.
Evaluate	Make judgments based on criteria	Evaluation depends on students internalizing and applying appropriate standards; relying on AI can replace personal judgment and weaken the development of discernment.	AI can present competing perspectives, model evaluative frameworks, and surface strengths and weaknesses, helping students clarify and refine their own criteria for judgment.	Students critique arguments, assess sources, rank options using criteria or write reviews without assistance.	Students complete evaluative tasks in which they demonstrate their ability to apply criteria, make justified decisions, and defend their reasoning independently.	Students use AI to explore alternative viewpoints, compare evaluative frameworks, and receive critique, to strengthen their ability to determine, apply, and defend their own standards and judgments.	Students complete AI-administered evaluative tasks in which AI presents criteria, applies rubric-based evaluation, and provides feedback, while students independently make judgments, apply criteria, and defend their reasoning.
Create	Produce original work	Creation requires students to originate ideas, synthesize knowledge, and take ownership of their work; overreliance on AI can obscure authorship and limit the development of creative and generative thinking.	AI can support brainstorming, iteration, and refinement, helping students expand ideas, explore possibilities, and improve the quality of their work through feedback and revision.	Students develop original products such as essays, projects, designs, or presentations by generating ideas, organizing content, and refining their work without assistance.	Students complete creation tasks such as papers, projects and portfolios in which they produce original work that demonstrates synthesis, coherence, and clear ownership of ideas.	Students use AI to brainstorm, explore alternatives, and refine their work through feedback and iteration, while maintaining responsibility for idea generation, direction, and final decisions.	Students complete AI-assisted assessments in which AI evaluates the final product using defined criteria and provides feedback, while authorship and originality remain clearly the student’s.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

© 2026 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.