Countering Digital Displacement: Evaluating A I-Enhanced Curriculum Studies Education through Philip’s Theory of Academic Performance Decline

1. Introduction

1.1 Background and Problem Statement: The Rise of Generative AI and the Threat of Digital Displacement in Educational Spaces

Since the public release of large language models able to produce coherent, context-sensitive text, code, and multimodal outputs, generative artificial intelligence has entered formal education at remarkable speed. Institutions have responded in markedly different ways some banning the tools outright, others redesigning entire programmes around GenAI as a cognitive partner. Beneath the language of efficiency and personalization, however, lies a deeper shift: the progressive displacement of the intellectual labour that once formed the heart of academic development. Digital displacement, as used here, describes the structural transfer of epistemic agency from the human learner to algorithmic systems. The result is a theoretically predictable decline in independent reasoning, evaluative judgement, and the capacity for sustained intellectual effort.

Empirical work already reveals the dual character of GenAI. Meta-analyses report moderate-to-large positive effects on selected learning outcomes when tools are embedded inside carefully structured pedagogical frameworks.¹ At the same time, experimental and correlational studies document cognitive offloading, reduced analytical autonomy, and lower performance on unaided assessments after unrestricted AI access.² The tension is conceptual as much as empirical: when the cognitive work required for knowledge construction is externalized, learners may experience an illusion of mastery that collapses once independent performance is required. The educational consequence is a quiet erosion of the very capacities curricula claim to cultivate.

Curriculum studies, concerned with the design, enactment, and evaluation of educational programmes, offers a particularly useful vantage point for diagnosing and countering this displacement. Traditional curriculum theory has long insisted that learning cannot be reduced to the production of artefacts; it is constituted by processes of inquiry, deliberation, and meaning-making situated in specific socio-cultural and material contexts. Uncritical adoption of GenAI risks flattening those processes into generic, decontextualized outputs that can be generated at scale without corresponding growth in human intellectual agency. The present inquiry therefore examines the vulnerabilities of contemporary curricula and proposes structural countermeasures grounded in a newly articulated diagnostic theory.

Urgency is heightened by the pace of adoption. Surveys conducted in 2025 and 2026 show that large majorities of secondary and tertiary students already use GenAI regularly for academic tasks, with a substantial proportion reporting daily reliance. Educators, meanwhile, express uncertainty about appropriate boundaries and note a relative scarcity of theoretically informed design principles capable of preserving the formative dimensions of learning. Without such principles, institutional responses risk oscillating between prohibition which is neither enforceable nor educationally optimal and laissez-faire integration that accelerates the very displacement the field should seek to mitigate. A diagnostic theory that clarifies mechanisms of decline and supplies criteria for redesign is therefore both timely and necessary.

1.2 Theoretical Foundation: PTAPD as a Diagnostic Instrument for Cognitive Regression

Philip’s Theory of Academic Performance Decline (PTAPD)³ is used here as a conceptual instrument for identifying, analyzing, and ultimately reversing patterns of cognitive regression associated with algorithmic mediation of learning. The theory holds that academic performance understood not merely as the production of assessable outputs but as the durable capacity for independent intellectual action declines when three interrelated conditions are present:

i. algorithmic over-reliance, in which learners systematically delegate core cognitive operations to AI systems;

ii. cognitive friction deficits, characterized by the removal of productive struggle essential for schema construction and transfer; and

iii. epistemic flattening, the homogenization of inquiry pathways that privileges generic, globally trained model outputs over localized, context-sensitive human reasoning.

These three parameters rest on both theoretical and empirical foundations. Cognitive-offloading research shows that habitual externalization of memory, calculation, and reasoning weakens the internal resources required for later independent performance.⁴ Desirable-difficulties research establishes that effortful engagement retrieval practice, spaced practice, interleaving, and generation produces superior long-term retention and transfer relative to fluent, low-effort processing.⁵ Scholarship on epistemic dependence and distributed cognition further reveals that when the locus of justification and verification migrates from the individual to an opaque algorithmic assemblage, the learner’s capacity for contestation, reconstruction, and intellectual-virtue formation is attenuated.⁶ PTAPD synthesizes these strands into a coherent diagnostic lens calibrated specifically to the educational consequences of GenAI.

Unlike purely descriptive models of AI impact, PTAPD is intentionally normative and interventional. It does not merely catalogue risks; it supplies criteria against which curriculum designs, assessment regimes, and pedagogical practices may be evaluated and redesigned. The theory thereby bridges critical analysis and constructive prescription, offering curriculum scholars a language for articulating why certain configurations of AI integration undermine educational purposes and how those configurations may be re-engineered. The term “performance decline” is deliberately chosen to emphasize durable capacity rather than momentary output quality: a student may produce high-scoring artefacts under AI mediation while simultaneously experiencing a decline in the independent performance that constitutes the deeper educational goal.

The architecture of PTAPD further treats the three parameters as mutually reinforcing rather than merely additive. Algorithmic over-reliance reduces the incidence of productive struggle; the resulting cognitive-friction deficits weaken the learner’s capacity to evaluate and contest AI outputs; and the consequent epistemic flattening further incentivises reliance on generic algorithmic solutions. This recursive dynamic helps explain why superficial interventions such as post-hoc detection of AI-generated text are insufficient. Structural redesign of the learning sequence itself is required if the cycle is to be interrupted.

1.3 Research Objectives: Assessing Vulnerabilities and Proposing Structural Countermeasures

The overarching aim of this study is to evaluate the vulnerabilities of contemporary curriculum-studies education under conditions of generative AI integration and to propose structural countermeasures that preserve human intellectual agency. Three specific objectives guide the inquiry. First, the study maps the mechanics of digital displacement by synthesizing empirical indicators of performance decline under uncritical AI use with the theoretical architecture of PTAPD. Second, it examines product-evaluation vulnerabilities, focusing on the failure of static, artefact-centred assessment metrics to discriminate between human intellectual labour and synthetic generation. Third, it documents processes of epistemic flattening and advances a re-engineering agenda centred on productive friction, aspectual diversification, and curriculum–practice convergence.

These objectives are pursued through a conceptual and critical analytical framework rather than primary empirical data collection. The justification for this design lies in the relative novelty of the diagnostic theory and the need for careful theoretical elaboration before large-scale empirical testing. By clarifying constructs, specifying evaluative criteria, and deriving actionable design principles, the paper establishes a foundation upon which subsequent empirical validation and design-based research may productively build. The conceptual approach also permits integration of heterogeneous evidence streams meta-analyses, experimental findings, qualitative interaction studies, and curriculum-theoretic arguments that would be difficult to synthesize within a single empirical design.

2. Methodology

2.1 Research Design: A Conceptual and Critical Analytical Framework Utilizing PTAPD

The study adopts a conceptual and critical analytical research design. Conceptual analysis is warranted because PTAPD is recent; its constructs still require precise definition, internal-coherence testing, and differentiation from related theoretical resources before they can be operationalized in empirical studies. Critical analysis is required because the educational integration of GenAI is not a neutral technical development but a value-laden restructuring of epistemic relations that demands normative evaluation against the purposes of curriculum studies namely: the cultivation of informed, critical, and agentic participants in democratic and professional life.

The design proceeds in three analytical movements. The first constructs the theoretical architecture of PTAPD by synthesizing insights from cognitive psychology (desirable difficulties, cognitive offloading), epistemic cognition, and curriculum theory. The second applies the resulting diagnostic parameters to three domains of educational practice: traditional assessment regimes, curriculum structures, and pedagogical mediation models. The third derives structural countermeasures and articulates their implications for policy and pedagogy. Throughout, the analysis remains attentive to the dual potential of GenAI: under conditions of deliberate instructional design it may amplify learning; under conditions of uncritical adoption it functions as a cognitive substitute.⁷

This methodological stance is justified by the current state of the evidence base. Meta-analyses reveal substantial heterogeneity in GenAI effects, with prediction intervals spanning harm to benefit.⁸ Such heterogeneity indicates that outcomes are highly sensitive to design and contextual factors. A purely empirical aggregation therefore risks obscuring the mechanisms that determine whether AI augments or erodes cognitive development. Conceptual clarification of those mechanisms is a necessary prior step. Moreover, the rapid evolution of GenAI capabilities means that empirical findings from any single cohort or tool generation risk rapid obsolescence; a theoretically robust diagnostic framework retains relevance across technological iterations.

The critical dimension of the analysis is further justified by the recognition that educational technology is never value-neutral. Decisions about where to place AI within a learning sequence, what forms of cognitive work to protect from offloading, and which forms of knowledge to privilege as authentic are simultaneously technical, pedagogical, and ethical. PTAPD makes these value commitments explicit by foregrounding cognitive sovereignty the durable capacity for independent intellectual action as the normative criterion against which integration strategies are judged.

2.2 Analytical Parameters: Criteria for Evaluating Algorithmic Over-Reliance, Cognitive Friction Deficits, and Epistemic Flattening

Three analytical parameters, derived from PTAPD, structure the evaluation. Algorithmic over-reliance is operationalized as the systematic delegation of core cognitive operations problem framing, evidence evaluation, argument construction, and verification to GenAI systems, accompanied by reduced learner capacity to perform those operations independently. Indicators include performance decrements on unaided post-tests following AI-supported practice, self-reported cognitive offloading, and behavioural patterns of prompt-and-copy rather than iterative dialogue.⁹ The parameter is further refined by distinguishing productive reliance, in which AI serves as a scaffold that leaves the learner more capable, from harmful dependence, in which the relation is difficult to contest and weakens transfer.¹⁰

Cognitive friction deficits refer to the engineered removal of productive struggle from learning sequences. Productive struggle is defined, following Hiebert and Grouws and subsequent elaborations, as effortful engagement with tasks that are challenging yet attainable, involving strategy generation, error detection, and revision.¹¹ Deficits arise when AI systems provide immediate, complete solutions, thereby short-circuiting the encoding, retrieval, and consolidation processes that build durable knowledge. The parameter is evaluated by examining whether curriculum designs preserve first-attempt effort and final unaided performance checks while permitting scaffolded AI use in intermediate phases.¹² The justification for prioritizing friction preservation draws on extensive evidence that desirable difficulties enhance long-term learning even when they temporarily reduce performance fluency.

Epistemic flattening denotes the homogenization of inquiry pathways and the displacement of localized, human-centric knowledge construction by generic model outputs trained on global corpora. Indicators include reduced diversity of reasoning strategies across learners, diminished attention to context-specific evidence, and the treatment of fluent AI text as authoritative without contestation.¹³ The parameter is assessed by analyzing the extent to which curricula require learners to anchor claims in hyper-local data, tactile experience, or culturally specific interpretive traditions that current GenAI systems cannot fully replicate. Selection of this parameter is justified by the distinctive contribution of curriculum studies to the valuation of context, practice, and democratic deliberation dimensions systematically under-represented in the training distributions of large language models.

2.3 Evaluative Scope: Reviewing Traditional Assessment Vulnerabilities, Curriculum Structures, and Pedagogical Mediation Models

The evaluative scope encompasses three interrelated domains. Traditional assessment vulnerabilities are examined through the lens of product-oriented metrics essays, reports, problem sets, and portfolios that historically served as proxies for intellectual labour but are now readily synthesized by GenAI. Curriculum structures are analyzed with respect to sequencing, task design, and the distribution of cognitive demand across learning trajectories. Pedagogical mediation models are scrutinized for the degree to which they position AI as a cognitive substitute versus a carefully constrained thinking partner. Selection of these domains is justified by their centrality to curriculum studies and by their documented susceptibility to the three PTAPD parameters.

Sources informing the analysis include recent systematic reviews and meta-analyses of GenAI effects on learning outcomes (2024–2026), experimental studies documenting performance trade-offs, conceptual contributions on cognitive friction and epistemic dependence, and curriculum-theoretic literature on productive struggle and local knowledge. The synthesis is selective rather than exhaustive; priority is given to studies that illuminate mechanisms of decline or that supply design principles capable of countering displacement. Where findings diverge, the analysis privileges those that specify boundary conditions structured versus unstructured use, short-term versus durable outcomes, and product versus process measures because such specificity is essential for the diagnostic and interventional purposes of PTAPD.

3. Results / Findings

3.1 The Mechanics of Digital Displacement: Empirical and Theoretical Indicators of Performance Decline under Uncritical AI Integration

Synthesis of recent evidence reveals a consistent pattern: when GenAI is introduced without pedagogical guardrails, short-term performance on AI-assisted tasks improves while subsequent unaided performance declines. In one randomized experiment, high-school students granted unrestricted chatbot access during mathematics practice scored approximately 17 per cent lower on a later unaided examination relative to peers who practised without AI; a version of the same model redesigned to withhold complete answers eliminated the harm.¹⁴ Parallel findings emerge from studies of essay writing, scientific inquiry, and critical-thinking assessments, where AI-supported groups exhibit lower neural connectivity, weaker argumentation, and reduced ownership of produced text.¹⁵

These empirical patterns align with the theoretical predictions of PTAPD. Algorithmic over-reliance is evidenced by the substitution of generative processes for retrieval, generation, and verification. Cognitive friction deficits are observable in the reduction of time-on-task and the bypassing of errorful, iterative engagement. Epistemic flattening appears in the convergence of student outputs toward model-preferred phrasings and structures, accompanied by diminished capacity to reconstruct reasoning without the tool. The Dual-Mechanism Model articulated in a 2026 systematic review of higher-order cognitive skills captures the contingency: GenAI functions as an amplifier under structured conditions and as a substitute under unguided use.¹⁶

Further support for the mechanics of displacement is found in mediation analyses linking AI-use frequency to critical-thinking scores through cognitive offloading.¹⁷ The youngest cohorts, who have had the least opportunity to develop independent cognitive routines prior to AI exposure, exhibit the strongest offloading tendencies. Classroom observation studies likewise report that general-purpose chatbots, more than pedagogically constrained tutoring systems, are associated with surface-level exchange and passive agreement rather than sustained intellectual engagement (Instruction Partners observations reported in 2026 syntheses). Collectively, these findings indicate that displacement is not an inevitable consequence of AI presence but a contingent outcome of design choices that remove productive struggle and fail to require independent performance.

Table 1 summarizes key empirical indicators mapped onto the three PTAPD parameters. The table is constructed to highlight the diagnostic utility of the theory rather than to claim exhaustive coverage of all studies.

Table 1: Empirical Indicators of Digital Displacement Mapped to PTAPD Parameters

Note. Indicators are drawn from experimental, correlational, and systematic-review evidence published between 2024 and 2026. The mapping is interpretive and theory-driven.

The mechanics of displacement are therefore neither uniform nor inevitable. They emerge preferentially under conditions of low instructional intentionality, high tool accessibility, and assessment regimes that reward final products irrespective of process. Where curricula and pedagogies actively regulate the placement of AI within the learning sequence securing first-attempt effort and final unaided checks the risk of decline is substantially attenuated.²⁷ This contingency underscores the value of a diagnostic theory capable of distinguishing harmful from productive configurations of AI use.

3.2 Product Evaluation Vulnerabilities: How Static Assessment Metrics Fail against Synthetic Generation and Automated Shortcuts

Traditional assessment practices in curriculum studies and related fields have long relied upon the production of textual or multimodal artefacts essays, research reports, lesson plans, reflective portfolios as evidence of learning. These product-oriented metrics presuppose a relatively stable relationship between the artefact and the underlying intellectual labour required to produce it. GenAI severs that relationship. Fluent, coherent, and formally correct outputs can now be generated with minimal human contribution, rendering the artefact an unreliable proxy for cognitive development.

The vulnerability is structural rather than merely technical. Detection tools remain imperfect and ethically fraught; more importantly, the educational problem is not solely academic integrity but the displacement of the formative processes that assessment was intended to capture and promote. When students can obtain high-quality products without engaging in problem framing, evidence synthesis, or iterative revision, the assessment ceases to function as a learning experience and becomes an exercise in prompt engineering or simple delegation. Empirical observations of metacognitive laziness and reduced ownership of AI-assisted writing corroborate this analysis.²⁸

PTAPD directs attention to three diagnostic questions that expose these vulnerabilities. First, does the assessment require evidence of independent cognitive operations that cannot be fully offloaded? Second, does it preserve productive friction by demanding first-attempt effort and subsequent revision under conditions of constrained AI support? Third, does it require anchoring in local, experiential, or context-specific knowledge that generic models systematically under-represent? Assessments that answer these questions negatively are classified as high-risk under the theory.

A further dimension of vulnerability concerns the temporal structure of assessment. When high-stakes evaluation occurs solely at the end of a learning sequence and focuses exclusively on final products, students have strong incentives to maximise AI assistance throughout the preceding work. Process-oriented alternatives portfolios that include first-attempt artefacts, revision traces, oral defences, and unaided performance samples alter incentive structures by making intermediate cognitive labour visible and evaluable. Such alternatives are not merely technical fixes; they reassert the formative purpose of assessment within a PTAPD framework.

Table 2 contrasts conventional product metrics with PTAPD-informed alternatives. The selection of alternatives prioritizes process visibility, friction preservation, and local anchoring—features that increase the cost of pure algorithmic substitution while remaining pedagogically legitimate.

Table 2: Conventional Product Metrics versus PTAPD-Informed Assessment Alternatives

Note. Alternatives are derived from principles of productive struggle, process-oriented assessment, and epistemic localization consistent with PTAPD.

3.3 Epistemic Flattening: Documenting the Loss of Localized, Human-Centric Inquiry and Critical Agency in Modern Learning Ecosystems

Epistemic flattening constitutes the third and most culturally consequential dimension of digital displacement. GenAI systems are trained on vast corpora that privilege statistically dominant patterns of language, argument, and knowledge organization. When learners systematically adopt these patterns as authoritative, the diversity of interpretive traditions, local knowledge systems, and idiosyncratic lines of inquiry is compressed. The result is a gradual convergence toward a narrow band of acceptable reasoning styles and content emphases, accompanied by a weakening of the learner’s disposition to contest, reconstruct, or situate knowledge within particular communities of practice.

Evidence of flattening appears in multiple forms. Qualitative analyses of student–AI interaction logs reveal modal relationships of validation rather than genuine dialogue: students seek confirmation of pre-formed answers more frequently than they engage in sustained interrogation of model outputs.²⁹ Correlational studies link heavier AI use to lower critical-thinking scores mediated by cognitive offloading.³⁰ Conceptual analyses of distributed epistemic displacement describe a structural condition in which the locus of justification migrates into a human–AI assemblage, leaving individual learners less capable of independent argument construction and intellectual-virtue formation.³¹

Within curriculum studies, the stakes are particularly high. The field has historically valued the reconstruction of educational experience in light of local contexts, democratic deliberation, and the interplay of theory and practice. When AI-generated generic analyses of curriculum problems displace site-specific inquiry classroom ethnography, community consultation, practitioner reflection—the distinctive contribution of curriculum scholarship is diluted. PTAPD therefore treats epistemic flattening not as an incidental side-effect but as a core diagnostic indicator requiring deliberate curricular countermeasures.

A further manifestation of flattening is the erosion of tolerance for uncertainty and ambiguity. GenAI systems characteristically produce fluent, confident, and complete-seeming responses even when underlying knowledge is incomplete or contested. Learners who internalize this mode of presentation may develop reduced disposition to dwell with unresolved questions, to hold multiple interpretive possibilities in tension, or to acknowledge the provisional character of educational knowledge. Such dispositions are central to the reflective practice that curriculum theory has long sought to cultivate. Their attenuation constitutes a form of intellectual loss that is difficult to measure with conventional achievement metrics yet consequential for the long-term quality of educational thought and practice.

4. Discussion

4.1 Re-engineering Curriculum Studies through PTAPD: Operationalizing Productive Friction and Aspectual Diversification

The diagnostic findings of the preceding section generate a constructive agenda. Re-engineering curriculum-studies education through PTAPD begins with the deliberate reintroduction of productive friction at those points in the learning sequence where cognitive development is most vulnerable to offloading. Design principles derived from the Effortless Trap framework³² and Friction-by-Design models³³ prescribe a consistent placement rule: secure the first hard attempt and the final unaided performance check; permit guarded AI support hints, examples, feedback only in intermediate phases. This placement preserves the encoding and retrieval processes that build durable knowledge while still allowing efficiency gains where they do not undermine learning.³⁴

Empirical support for this sequencing logic continues to accumulate. Bastani et al.³⁵ demonstrated that simply withholding complete answers from a generative model during practice eliminated the subsequent unaided-performance decrement observed under unrestricted access. Parallel evidence from Stadler et al.³⁶ and Fan et al.³⁷ shows that learners who retain responsibility for initial problem framing and final verification display stronger argumentation, greater ownership of produced text, and higher neural engagement during production. These findings align with the broader desirable-difficulties literature, which has long established that temporary reductions in performance fluency are often the price of superior long-term retention and transfer.³⁸ PTAPD simply translates that principle into the specific design language required by GenAI environments.

Aspectual diversification constitutes a second operational strategy. Rather than permitting a single, AI-generatable product to stand as evidence of mastery, curricula should require learners to demonstrate competence across multiple complementary aspects of a domain: theoretical analysis, empirical observation, practical enactment, critical evaluation of AI outputs, and reflective articulation of personal intellectual growth.³⁹ Diversification raises the cost of pure algorithmic substitution because no single model prompt can simultaneously satisfy the full range of demands, particularly when some aspects require embodied or localized activity. Recent classroom studies confirm that multi-aspect assessment regimes substantially reduce the incidence of prompt-and-copy behaviour and increase the frequency of iterative, critical dialogue with AI systems.⁴⁰

Justification for these selections rests on the dual-mechanism evidence base articulated in the Alubthane⁴¹ systematic review: structured pedagogical conditions convert GenAI from substitute to amplifier; unstructured conditions reverse the effect. By embedding friction and diversification as non-negotiable design features, curriculum designers shift the default from displacement toward cognitive sovereignty. The approach does not reject AI; it subordinates AI to the educational purposes that remain distinctively human. Moreover, the strategies are scalable: they can be implemented at the level of individual assignments, course sequences, or programme-level curriculum mapping, and they remain compatible with a range of disciplinary traditions within curriculum studies.⁴²

Operationalization further requires attention to the fading of scaffolding. As learners demonstrate increasing capacity for independent performance, the intensity and availability of AI support should be progressively reduced. This principle, long established in scaffolding research, acquires renewed importance under conditions of GenAI, where the temptation to maintain continuous high-level assistance is strong.⁴³ PTAPD supplies a diagnostic criterion for determining when fading is warranted: the point at which continued high support would begin to produce measurable cognitive-friction deficits or indicators of over-reliance. Longitudinal tracking of unaided performance samples provides a practical means of calibrating that threshold in real instructional settings.⁴⁴

4.2 Curriculum–Practice Convergence: Anchoring Learning in Hyper-Localized and Tactile Realities to Neutralize Generic AI Output

A further structural countermeasure is the systematic convergence of curriculum and practice through hyper-localized and tactile anchoring. GenAI systems excel at producing fluent discourse about generic educational phenomena; they remain limited in their capacity to incorporate real-time, site-specific observational data, embodied practitioner experience, and community knowledge that has not been digitized at scale.⁴⁵ Curricula that mandate such anchoring classroom micro-ethnographies, material analysis of local artefacts, collaborative inquiry with practising teachers, or community-based curriculum audits create zones of inquiry that resist complete algorithmic substitution.⁴⁶

This strategy is justified on both epistemic and pedagogical grounds. Epistemically, it restores the connection between knowledge claims and the particular contexts that give them meaning, countering the flattening tendency of global model corpora.⁴⁷ Pedagogically, it reintroduces productive struggle through the demands of observation, negotiation, and situated judgement activities that cannot be fully offloaded.⁴⁸ The resulting learning experiences cultivate the very capacities that PTAPD identifies as vulnerable: independent problem framing, evidence evaluation under uncertainty, and the intellectual virtues of careful attention and responsibility.⁴⁹

Recent design-based research reinforces the practical viability of localization strategies. Studies that required students to interrogate AI-generated curriculum analyses against their own observational data report higher levels of critical agency, greater diversity of interpretive moves, and stronger retention of conceptual frameworks.⁵⁰ Parallel findings emerge from programmes that embed tactile and material tasks manipulation of classroom artefacts, collaborative construction of physical curriculum prototypes, or site-based community consultations within otherwise discursive courses.⁵¹ These activities function as natural barriers to complete displacement while simultaneously enriching the quality of educational understanding.

Implementation requires careful calibration. Hyper-localization should not become parochialism; learners must still engage broader theoretical resources and comparative perspectives.⁵² The point is not to exclude AI but to position it as a tool for expanding, rather than replacing, the learner’s engagement with concrete educational realities. When students are required to interrogate AI-generated analyses against their own observational data, the interaction becomes a site of critical agency rather than passive reception. In this configuration, AI may accelerate certain preparatory tasks literature scanning, for example while the core intellectual work of interpretation, critique, and synthesis remains with the human learner.⁵³

Tactile and embodied dimensions of learning warrant particular emphasis. Curriculum studies has long recognized that educational knowledge is not solely discursive; it is also enacted, felt, and materially situated. Activities that require physical presence in educational settings, manipulation of material artefacts, or collaborative construction of tangible products introduce forms of cognitive and perceptual demand that current GenAI systems cannot replicate.⁵⁴ Such activities therefore function as natural barriers to complete displacement while simultaneously enriching the quality of educational understanding and reinforcing the distinctive contribution of curriculum scholarship.

4.3 Policy and Pedagogical Implications: Transitioning Educational Design from Output Efficiency to Cognitive Sovereignty

The implications of the analysis extend beyond individual course design to institutional policy and professional formation. Educational systems that continue to prioritize output efficiency speed of production, volume of artefacts, surface correctness will systematically incentivise the very forms of AI use that PTAPD diagnoses as deleterious.⁵⁵ A transition toward cognitive sovereignty requires the reorientation of success metrics around durable independent performance, process quality, and the capacity to exercise judgement under conditions of uncertainty and incomplete information.⁵⁶

Policy levers include the revision of assessment regulations to require process evidence and unaided performance samples; the development of AI-use guidelines that specify permitted and prohibited phases of cognitive work; investment in professional learning that equips educators to design friction-preserving tasks; and the incorporation of epistemic literacy understanding the limits, biases, and knowledge conditions of AI systems into core curriculum.⁵⁷ Pedagogically, the shift demands that educators reclaim the design of productive struggle as a central professional responsibility rather than outsourcing it to tool vendors whose commercial incentives favour seamless, low-friction interaction.⁵⁸

These implications are advanced with recognition of genuine constraints. Not all institutions possess equivalent resources for redesign; not all learners enter with equal prior preparation; and the rapid evolution of AI capabilities continually alters the boundary between what can and cannot be offloaded.⁵⁹ Nonetheless, the normative commitment of curriculum studies to the formation of agentic, critically capable persons supplies a stable reference point against which technological developments must be evaluated. PTAPD offers one theoretically articulated means of conducting that evaluation and of guiding the re-engineering that follows.⁶⁰

A final implication concerns the professional identity of curriculum scholars and educators. In an environment saturated with AI-generated educational content and analysis, the distinctive contribution of human curriculum work lies increasingly in the capacity for situated judgement, ethical deliberation, and the cultivation of intellectual agency in others.⁶¹ Preserving and strengthening these capacities is not a nostalgic resistance to technological change but a necessary condition for education to remain a human enterprise oriented toward human flourishing. Longitudinal research that tracks both cognitive outcomes and professional-identity formation under redesigned curricula will be essential for refining and validating the policy recommendations advanced here.⁶²

5. Conclusion

This paper has advanced Philip’s Theory of Academic Performance Decline as a diagnostic and interventional framework for understanding and countering digital displacement in AI-enhanced curriculum-studies education. Through conceptual and critical analysis organized around three parameters algorithmic over-reliance, cognitive friction deficits, and epistemic flattening the inquiry has mapped the mechanisms by which uncritical GenAI integration erodes independent intellectual agency, exposed the vulnerabilities of product-oriented assessment regimes, and documented the homogenizing effects of generic model outputs on localized inquiry. In response, structural countermeasures have been proposed: the operationalization of productive friction through deliberate placement of AI support, aspectual diversification of demonstration requirements, and curriculum–practice convergence anchored in hyper-localized and tactile realities. Collectively, these measures aim to reorient educational design from the pursuit of output efficiency toward the preservation of cognitive sovereignty the durable capacity of learners to frame problems, evaluate evidence, construct arguments, and exercise judgement without continuous algorithmic scaffolding.

The contribution is primarily theoretical and diagnostic. Empirical validation of PTAPD constructs, longitudinal tracking of cognitive outcomes under redesigned curricula, and comparative studies of friction-preserving versus friction-removing AI integrations constitute necessary next steps. What the present analysis establishes is a coherent conceptual architecture and a set of justified design principles capable of guiding those subsequent inquiries. In an era of rapid technological change, the responsibility of curriculum scholarship remains the defence and cultivation of the human intellectual agency that no algorithm can fully replace. Future research should prioritize the development of validated instruments for measuring the three PTAPD parameters, design-based investigations of friction-preserving pedagogical sequences, and comparative analyses of curriculum programmes that have and have not implemented localization and diversification strategies. Cross-cultural and cross-level studies will be particularly valuable for determining the boundary conditions under which the theory holds and the adaptations required for different educational contexts. Through such cumulative work, PTAPD may evolve from a conceptual proposal into a robust, empirically grounded framework for safeguarding cognitive sovereignty in an age of generative artificial intelligence.

Endnotes

1. Bayaro, M., Olorunfemi, P.A., Michael, C.I., & Adeleke, H. A. (2026). Beyond Native Norms: Decolonizing AI Speech Recognition for Accent Equity in Nigerian ESL Classrooms. Center of Artificial Intelligence, 1(2), 111-121. https://doi.org/10.65591/CAI-206-2026; Olorunfemi, P.A, Nnaji, C.M & Michael, C.I. (2026). An Investigation into AI-Assisted Aspect-Based and Integrated English Teaching Approaches Using ADLL and CPCT Theories. Center of Artificial Intelligence. 1(2). 84-95. https://doi.org/10.65591/CAI-147-2026; Jing, Y., Wang, J., Wang, P., & Xu, J. (2026). The Impact of Generative Artificial Intelligence on Learning Outcomes in Higher Education: A Meta-Analysis. Journal of Computer Assisted Learning 42, no. 5: e70329. https://doi.org/10.1002/jcal.70329.; Ma, N., & Zhong, Z. (2025). A Meta-Analysis of the Impact of Generative Artificial Intelligence on Learning Outcomes. Journal of Computer Assisted Learning 41, no. 5: e70117. https://doi.org/10.1111/jcal.70117.; Nie, X., Tian, Y., Liu, M., Wu, D., & Guo, Y. (2025). The impact of generative artificial intelligence on students’ higher order thinking: Evidence from a three-level meta-analysis. Education and Information Technologies, 30(17), 25359–25390. https://doi.org/10.1007/s10639-025-13735-x

2. Bastani, H., Bastani, O., & Sungu, A. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), e2422633122. https://doi.org/ 10.1073/pnas.2422633122; Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15, Article 6.https://doi.org/10.3390/soc15010006; Stadler, M., Bannert, M., & Sailer, M. (2024). Cognitive ease at a cost: LLMs reduce mental effort but compromise depth in student scientific inquiry, Computers in Human Behavior, Volume 160, 108386, https://doi.org/10.1016/j.chb.2024.108386.

3. Olorunfemi, P. A., & Bayaro, M. (2025). Mitigating Academic Decline in English Language Skills through Aspectual Diversification and Curriculum Practice Convergence: A Multi-Theoretical Perspective Based on PTAPD, ADLL, and CPCT. Iconic Research and Engineering Journals, 9(6) https://doi.org/10.64388/IREV9I6-1713032

4. Risko, E. F., & Gilbert, S. J. (2016). Cognitive offloading. Trends in Cognitive Sciences, 20(9), 676–688. Pages 676-688, https://doi.org/10.1016/j.tics.2016.07.002.; Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15, Article 6.https://doi.org/10.3390/soc15010006

5. Bjork, E., & Bjork, R. (2011). Making Things Hard on Yourself, but in a Good Way: Creating Desirable Difficulties to Enhance Learning. In M. A. Gernsbacher, R. W. Pew, L. M. Hough, & J. R. Pomerantz (Eds.), Psychology and the Real World: Essays Illustrating Fundamental Contributions to Society (pp. 56-64). Worth Publishers.; Bjork, R. A., Dunlosky, J., & Kornell, N. (2013). Self-Regulated Learning: Beliefs, Techniques, and Illusions. Annual Review of Psychology, 64, 417-444.https://doi.org/10.1146/annurev-psych-113011-143823

6. Hussain, S., & Khan, I. (2026). Thinking by proxy: Distributed epistemic displacement (DED) in AI-mediated education. Research Consortium Archive, 4(2), 1784–1797. https://rcresearcharchive.com/index.php/Journal/article/view/967; Fan, Y., Tang, L., Le, H., Shen, K., Tan, S., Zhao, Y., Shen, Y., Li, X.,& Gašević, D. (2025). Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology, 56, 489–530. https://doi.org/10.1111/bjet.13544

7. Alubthane, F.O. (2026). Amplifier or substitute? A systematic review of generative AI’s impact on higher-order cognitive skills among university students. Frontiers in Psychology. 17:1863931. https://doi.org/10.3389/fpsyg.2026.1863931

8. Nie, X., Tian, Y., Liu, M., Wu, D., & Guo, Y. (2025). The impact of generative artificial intelligence on students’ higher order thinking: Evidence from a three-level meta-analysis. Education and Information Technologies, 30(17), 25359–25390. https://doi.org/10.1007/s10639-025-13735-x; Fang L., Li Z., Xiong, B., & Lu, Z. (2026). Effects of generative artificial intelligence on students' learning motivation: evidence from a meta-analysis. Frontiers in Psychology. 17:1832451. https://doi.org/10.3389/fpsyg.2026.1832451

9. Bastani, H., Bastani, O., & Sungu, A. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), e2422633122. https://doi.org/ 10.1073/pnas.2422633122; Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15, Article 6.https://doi.org/10.3390/soc15010006

10. Hussain, S., & Khan, I. (2026). Thinking by proxy: Distributed epistemic displacement (DED) in AI-mediated education. Research Consortium Archive, 4(2), 1784–1797. https://rcresearcharchive.com/index.php/Journal/article/view/967

11. Bjork, R. A., Dunlosky, J., & Kornell, N. (2013). Self-Regulated Learning: Beliefs, Techniques, and Illusions. Annual Review of Psychology, 64, 417-444.https://doi.org/10.1146/annurev-psych-113011-143823; Bussell, B., & Moyle, P. (2025). Friction by design: Centering learning in the age of AI [Webinar]. WestEd. https://www.wested.org/resource/friction-by-design-centering-learning-in-the-age-of-ai/

12. Brcic, M., & Frljic, S. (2026). The effortless trap: Productive struggle, AI, and the illusion of learning. arXiv preprint arXiv:2606.26181.; arXiv constraint-first design pilot, 2026

13. Hussain, S., & Khan, I. (2026). Thinking by proxy: Distributed epistemic displacement (DED) in AI-mediated education. Research Consortium Archive, 4(2), 1784–1797. https://rcresearcharchive.com/index.php/Journal/article/view/967; Suazo Galdames, I. C., Santiesteban Velázquez, M., Saracostti, M., & Chaple Gil, A. M. (2026). Between cognitive offloading and critical autonomy: A systematic review of the epistemic implications of generative AI in higher education. Frontiers in Education, 11, Article 1878665. https://doi.org/10.3389/feduc.2026.1878665

14. Bastani, H., Bastani, O., & Sungu, A. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), e2422633122. https://doi.org/ 10.1073/pnas.2422633122

15. MIT Media Lab studies summarized in 2025–2026 reviews; Stadler, M., Bannert, M., & Sailer, M. (2024). Cognitive ease at a cost: LLMs reduce mental effort but compromise depth in student scientific inquiry, Computers in Human Behavior, Volume 160, 108386, https://doi.org/10.1016/j.chb.2024.108386.

16. Alubthane, F.O. (2026). Amplifier or substitute? A systematic review of generative AI’s impact on higher-order cognitive skills among university students. Frontiers in Psychology. 17:1863931. https://doi.org/10.3389/fpsyg.2026.1863931

17. Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15, Article 6.https://doi.org/10.3390/soc15010006

18. Bastani, H., Bastani, O., & Sungu, A. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), e2422633122. https://doi.org/ 10.1073/pnas.2422633122

19. Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15, Article 6.https://doi.org/10.3390/soc15010006

20. Fan, Y., Tang, L., Le, H., Shen, K., Tan, S., Zhao, Y., Shen, Y., Li, X.,& Gašević, D. (2025). Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology, 56, 489–530. https://doi.org/10.1111/bjet.13544

21. Stadler, M., Bannert, M., & Sailer, M. (2024). Cognitive ease at a cost: LLMs reduce mental effort but compromise depth in student scientific inquiry, Computers in Human Behavior, Volume 160, 108386, https://doi.org/10.1016/j.chb.2024.108386.

22. Brcic, M., & Frljic, S. (2026). The effortless trap: Productive struggle, AI, and the illusion of learning. arXiv preprint arXiv:2606.26181.

23. Bussell, B., & Moyle, P. (2025). Friction by design: Centering learning in the age of AI [Webinar]. WestEd. https://www.wested.org/resource/friction-by-design-centering-learning-in-the-age-of-ai/

24. Hussain, S., & Khan, I. (2026). Thinking by proxy: Distributed epistemic displacement (DED) in AI-mediated education. Research Consortium Archive, 4(2), 1784–1797. https://rcresearcharchive.com/index.php/Journal/article/view/967

25. Frontiers Educ. (2026)

26. Gill-Simmen, L. (2026). AI doesn't just help us think, it thinks instead of us: What this means for the process of learning. The Conversation. https://theconversation.com/ai-doesnt-just-help-us-think-it-thinks-instead-of-us-what-this-means-for-the-process-of-learning-284038

27. arXiv Effortless Trap framework, 2026

28. Fan, Y., Tang, L., Le, H., Shen, K., Tan, S., Zhao, Y., Shen, Y., Li, X.,& Gašević, D. (2025). Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology, 56, 489–530. https://doi.org/10.1111/bjet.13544; MIT Media Lab findings

29. arXiv critical-thinking pilot, 2026

30. Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15, Article 6.https://doi.org/10.3390/soc15010006

31. Hussain, S., & Khan, I. (2026). Thinking by proxy: Distributed epistemic displacement (DED) in AI-mediated education. Research Consortium Archive, 4(2), 1784–1797. https://rcresearcharchive.com/index.php/Journal/article/view/967

32. Brcic, M., & Frljic, S. (2026). The effortless trap: Productive struggle, AI, and the illusion of learning. arXiv preprint arXiv:2606.26181.

33. Bussell, B., & Moyle, P. (2025). Friction by design: Centering learning in the age of AI [Webinar]. WestEd. https://www.wested.org/resource/friction-by-design-centering-learning-in-the-age-of-ai/

34. Bjork, R. A., Dunlosky, J., & Kornell, N. (2013). Self-Regulated Learning: Beliefs, Techniques, and Illusions. Annual Review of Psychology, 64, 417-444.https://doi.org/10.1146/annurev-psych-113011-143823; Bjork, E., & Bjork, R. (2011). Making Things Hard on Yourself, but in a Good Way: Creating Desirable Difficulties to Enhance Learning. In M. A. Gernsbacher, R. W. Pew, L. M. Hough, & J. R. Pomerantz (Eds.), Psychology and the Real World: Essays Illustrating Fundamental Contributions to Society (pp. 56-64). Worth Publishers.

35. Bastani, H., Bastani, O., & Sungu, A. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), e2422633122. https://doi.org/ 10.1073/pnas.2422633122

36. Stadler, M., Bannert, M., & Sailer, M. (2024). Cognitive ease at a cost: LLMs reduce mental effort but compromise depth in student scientific inquiry, Computers in Human Behavior, Volume 160, 108386, https://doi.org/10.1016/j.chb.2024.108386.

37. Fan, Y., Tang, L., Le, H., Shen, K., Tan, S., Zhao, Y., Shen, Y., Li, X.,& Gašević, D. (2025). Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology, 56, 489–530. https://doi.org/10.1111/bjet.13544

38. Bjork, R. A., Dunlosky, J., & Kornell, N. (2013). Self-Regulated Learning: Beliefs, Techniques, and Illusions. Annual Review of Psychology, 64, 417-444.https://doi.org/10.1146/annurev-psych-113011-143823

39. Hussain, S., & Khan, I. (2026). Thinking by proxy: Distributed epistemic displacement (DED) in AI-mediated education. Research Consortium Archive, 4(2), 1784–1797. https://rcresearcharchive.com/index.php/Journal/article/view/967; Gill-Simmen, L. (2026). AI doesn't just help us think, it thinks instead of us: What this means for the process of learning. The Conversation. https://theconversation.com/ai-doesnt-just-help-us-think-it-thinks-instead-of-us-what-this-means-for-the-process-of-learning-284038

40. Suazo Galdames, I. C., Santiesteban Velázquez, M., Saracostti, M., & Chaple Gil, A. M. (2026). Between cognitive offloading and critical autonomy: A systematic review of the epistemic implications of generative AI in higher education. Frontiers in Education, 11, Article 1878665. https://doi.org/10.3389/feduc.2026.1878665; Instruction Partners syntheses, 2026

41. Alubthane, F.O. (2026). Amplifier or substitute? A systematic review of generative AI’s impact on higher-order cognitive skills among university students. Frontiers in Psychology. 17:1863931. https://doi.org/10.3389/fpsyg.2026.1863931

42. Nie, X., Tian, Y., Liu, M., Wu, D., & Guo, Y. (2025). The impact of generative artificial intelligence on students’ higher order thinking: Evidence from a three-level meta-analysis. Education and Information Technologies, 30(17), 25359–25390. https://doi.org/10.1007/s10639-025-13735-x; Fang L., Li Z., Xiong, B., & Lu, Z. (2026). Effects of generative artificial intelligence on students' learning motivation: evidence from a meta-analysis. Frontiers in Psychology. 17:1832451. https://doi.org/10.3389/fpsyg.2026.1832451

43. Risko, E. F., & Gilbert, S. J. (2016). Cognitive offloading. Trends in Cognitive Sciences, 20(9), 676–688. Pages 676-688, https://doi.org/10.1016/j.tics.2016.07.002.; Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15, Article 6.https://doi.org/10.3390/soc15010006

44. Brcic, M., & Frljic, S. (2026). The effortless trap: Productive struggle, AI, and the illusion of learning. arXiv preprint arXiv:2606.26181.

45. Hussain, S., & Khan, I. (2026). Thinking by proxy: Distributed epistemic displacement (DED) in AI-mediated education. Research Consortium Archive, 4(2), 1784–1797. https://rcresearcharchive.com/index.php/Journal/article/view/967; Suazo Galdames, I. C., Santiesteban Velázquez, M., Saracostti, M., & Chaple Gil, A. M. (2026). Between cognitive offloading and critical autonomy: A systematic review of the epistemic implications of generative AI in higher education. Frontiers in Education, 11, Article 1878665. https://doi.org/10.3389/feduc.2026.1878665

46. Olorunfemi, P. A., & Bayaro, M. (2025). Mitigating Academic Decline in English Language Skills through Aspectual Diversification and Curriculum Practice Convergence: A Multi-Theoretical Perspective Based on PTAPD, ADLL, and CPCT. Iconic Research and Engineering Journals, 9(6) https://doi.org/10.64388/IREV9I6-1713032

47. Gill-Simmen, L. (2026). AI doesn't just help us think, it thinks instead of us: What this means for the process of learning. The Conversation. https://theconversation.com/ai-doesnt-just-help-us-think-it-thinks-instead-of-us-what-this-means-for-the-process-of-learning-284038; Fan, Y., Tang, L., Le, H., Shen, K., Tan, S., Zhao, Y., Shen, Y., Li, X.,& Gašević, D. (2025). Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology, 56, 489–530. https://doi.org/10.1111/bjet.13544

48. Bjork, R. A., Dunlosky, J., & Kornell, N. (2013). Self-Regulated Learning: Beliefs, Techniques, and Illusions. Annual Review of Psychology, 64, 417-444.https://doi.org/10.1146/annurev-psych-113011-143823; Bussell, B., & Moyle, P. (2025). Friction by design: Centering learning in the age of AI [Webinar]. WestEd. https://www.wested.org/resource/friction-by-design-centering-learning-in-the-age-of-ai/

49. Hussain, S., & Khan, I. (2026). Thinking by proxy: Distributed epistemic displacement (DED) in AI-mediated education. Research Consortium Archive, 4(2), 1784–1797. https://rcresearcharchive.com/index.php/Journal/article/view/967

50. arXiv critical-thinking pilot, 2026; Instruction Partners, 2026

51. Nie, X., Tian, Y., Liu, M., Wu, D., & Guo, Y. (2025). The impact of generative artificial intelligence on students’ higher order thinking: Evidence from a three-level meta-analysis. Education and Information Technologies, 30(17), 25359–25390. https://doi.org/10.1007/s10639-025-13735-x

52. Ma, N., & Zhong, Z. (2025). A Meta-Analysis of the Impact of Generative Artificial Intelligence on Learning Outcomes. Journal of Computer Assisted Learning 41, no. 5: e70117. https://doi.org/10.1111/jcal.70117.; Jing, Y., Wang, J., Wang, P., & Xu, J. (2026). The Impact of Generative Artificial Intelligence on Learning Outcomes in Higher Education: A Meta-Analysis. Journal of Computer Assisted Learning 42, no. 5: e70329. https://doi.org/10.1002/jcal.70329.

53. Bastani, H., Bastani, O., & Sungu, A. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), e2422633122. https://doi.org/ 10.1073/pnas.2422633122; Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15, Article 6.https://doi.org/10.3390/soc15010006

54. Bussell, B., & Moyle, P. (2025). Friction by design: Centering learning in the age of AI [Webinar]. WestEd. https://www.wested.org/resource/friction-by-design-centering-learning-in-the-age-of-ai/; Brcic, M., & Frljic, S. (2026). The effortless trap: Productive struggle, AI, and the illusion of learning. arXiv preprint arXiv:2606.26181.

55. Fan, Y., Tang, L., Le, H., Shen, K., Tan, S., Zhao, Y., Shen, Y., Li, X.,& Gašević, D. (2025). Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology, 56, 489–530. https://doi.org/10.1111/bjet.13544; Stadler, M., Bannert, M., & Sailer, M. (2024). Cognitive ease at a cost: LLMs reduce mental effort but compromise depth in student scientific inquiry, Computers in Human Behavior, Volume 160, 108386, https://doi.org/10.1016/j.chb.2024.108386.

56. Hussain, S., & Khan, I. (2026). Thinking by proxy: Distributed epistemic displacement (DED) in AI-mediated education. Research Consortium Archive, 4(2), 1784–1797. https://rcresearcharchive.com/index.php/Journal/article/view/967; Gill-Simmen, L. (2026). AI doesn't just help us think, it thinks instead of us: What this means for the process of learning. The Conversation. https://theconversation.com/ai-doesnt-just-help-us-think-it-thinks-instead-of-us-what-this-means-for-the-process-of-learning-284038

57. Alubthane, F.O. (2026). Amplifier or substitute? A systematic review of generative AI’s impact on higher-order cognitive skills among university students. Frontiers in Psychology. 17:1863931. https://doi.org/10.3389/fpsyg.2026.1863931; Bussell, B., & Moyle, P. (2025). Friction by design: Centering learning in the age of AI [Webinar]. WestEd. https://www.wested.org/resource/friction-by-design-centering-learning-in-the-age-of-ai/

58. Brcic, M., & Frljic, S. (2026). The effortless trap: Productive struggle, AI, and the illusion of learning. arXiv preprint arXiv:2606.26181.; Bjork, R. A., Dunlosky, J., & Kornell, N. (2013). Self-Regulated Learning: Beliefs, Techniques, and Illusions. Annual Review of Psychology, 64, 417-444.https://doi.org/10.1146/annurev-psych-113011-143823

59. Nie, X., Tian, Y., Liu, M., Wu, D., & Guo, Y. (2025). The impact of generative artificial intelligence on students’ higher order thinking: Evidence from a three-level meta-analysis. Education and Information Technologies, 30(17), 25359–25390. https://doi.org/10.1007/s10639-025-13735-x; Fang L., Li Z., Xiong, B., & Lu, Z. (2026). Effects of generative artificial intelligence on students' learning motivation: evidence from a meta-analysis. Frontiers in Psychology. 17:1832451. https://doi.org/10.3389/fpsyg.2026.1832451

60. Hussain, S., & Khan, I. (2026). Thinking by proxy: Distributed epistemic displacement (DED) in AI-mediated education. Research Consortium Archive, 4(2), 1784–1797. https://rcresearcharchive.com/index.php/Journal/article/view/967; Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15, Article 6.https://doi.org/10.3390/soc15010006

61. Gill-Simmen, L. (2026). AI doesn't just help us think, it thinks instead of us: What this means for the process of learning. The Conversation. https://theconversation.com/ai-doesnt-just-help-us-think-it-thinks-instead-of-us-what-this-means-for-the-process-of-learning-284038; Suazo Galdames, I. C., Santiesteban Velázquez, M., Saracostti, M., & Chaple Gil, A. M. (2026). Between cognitive offloading and critical autonomy: A systematic review of the epistemic implications of generative AI in higher education. Frontiers in Education, 11, Article 1878665. https://doi.org/10.3389/feduc.2026.1878665

62. Bastani, H., Bastani, O., & Sungu, A. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), e2422633122. https://doi.org/ 10.1073/pnas.2422633122; Ma, N., & Zhong, Z. (2025). A Meta-Analysis of the Impact of Generative Artificial Intelligence on Learning Outcomes. Journal of Computer Assisted Learning 41, no. 5: e70117. https://doi.org/10.1111/jcal.70117.