Introduction: The Educational Paradigm Shift of 2026
The year 2026 marks a definitive inflection point in the history of global education. For over a century, the standardized test stood as the immutable monolith of academic assessment—a singular, high-stakes event designed to measure student proficiency through a uniform lens. It was a tool born of the industrial age, prioritizing efficiency, comparability, and standardization above the nuanced, erratic reality of human learning. However, as we navigate the mid-2020s, this edifice is crumbling. It is being dismantled not by a return to pre-industrial apprenticeship models, but by the ascendancy of a technology that promises to make assessment invisible, continuous, and hyper-personalized: Adaptive Learning.
The catalysts for this transformation are manifold. The educational landscape of 2025 and 2026 is defined by a desperate need to address the lingering academic scars of the early 2020s. Data from the National Assessment of Educational Progress (NAEP) released in 2024 and 2025 revealed record-low proficiency rates among high school seniors in reading and mathematics, exposing the inadequacy of traditional "cram-and-exam" methodologies to close widening achievement gaps.1 Educators and policymakers increasingly recognize that a static snapshot of student performance, taken once a year, is insufficient for diagnosing complex learning trajectories or informing real-time instructional interventions.2 The demand is no longer for mere accountability; it is for actionable, predictive intelligence.
In this context, Adaptive Learning systems—powered by advanced Artificial Intelligence (AI), Machine Learning (ML), and Deep Knowledge Tracing (DKT)—have moved from experimental pilots to the central nervous system of modern schooling. These systems do not merely digitize the textbook; they fundamentally restructure the relationship between the learner and the curriculum. By continuously analyzing thousands of data points—from response accuracy to hesitation times—adaptive platforms construct a dynamic "digital twin" of the student's cognitive state. This allows for the delivery of content that is rigorously tailored to the individual's Zone of Proximal Development, rendering the concept of a "standardized" test obsolete. If an AI tutor knows a student's proficiency in real-time with 99% accuracy, the end-of-year examination becomes a redundancy.4
This report provides an exhaustive analysis of the death of standardized testing and the rise of AI-driven adaptive learning. We will dissect the intricate algorithmic architectures that power these systems, evaluate the empirical evidence of their efficacy across global case studies, explore the emerging ecosystem of Comprehensive Learner Records (CLR) that are replacing traditional transcripts, and scrutinize the profound ethical and psychological implications of a surveillance-based educational model. As the global market for AI in education hurtles toward a projected valuation of $136 billion by 2035 6, understanding the mechanics and consequences of this shift is no longer optional for educators, administrators, or policymakers—it is an existential imperative.
Chapter 1: The Algorithmic Core of Adaptive Learning
To truly comprehend the disruptive power of adaptive learning, one must look beneath the user interface to the mathematical and computational engines that drive personalization. The transition from standardized testing to adaptive learning is, at its heart, a transition from static statistics to dynamic deep learning.
1.1 From Classical Test Theory to Intelligent Adaptation
Historically, standardized testing relied on Classical Test Theory (CTT), which assumes that an observed score is a combination of a "true score" and error. While useful for ranking students, CTT offers little insight into why a student answered correctly or incorrectly. The first leap toward adaptation came with Item Response Theory (IRT). IRT posits that the probability of a correct response is a function of the student's latent ability and the item's difficulty. This allowed for the creation of Computerized Adaptive Testing (CAT), where the test gets harder if you answer correctly and easier if you answer incorrectly.3
However, traditional CAT is limited. It views the student's ability as a fixed parameter to be estimated. Modern Adaptive Learning systems, by contrast, view the student's ability as a fluid state to be influenced. They do not just measure; they teach.
1.2 Bayesian Knowledge Tracing (BKT)
For decades, the dominant algorithm in Intelligent Tutoring Systems (ITS) was Bayesian Knowledge Tracing (BKT). BKT models student knowledge as a set of binary variables—mastered or not mastered—for each specific skill or concept. As a student interacts with the system, the algorithm updates the probability that the student has learned the skill, factoring in two critical probabilities:
- Slip: The probability that a student knows the skill but makes a mistake.
- Guess: The probability that a student does not know the skill but guesses correctly.8
While BKT represented a significant advance over simple branching logic, it suffers from structural limitations. It typically treats skills as independent entities, failing to capture the complex web of interdependencies that define real-world knowledge (e.g., how mastery of "fractions" influences "algebraic equations"). Furthermore, BKT struggles to model the temporal dynamics of learning—how memory decays over time or how learning accelerates with practice sequences.9
1.3 The Revolution of Deep Knowledge Tracing (DKT)
The current state-of-the-art in adaptive learning is defined by Deep Knowledge Tracing (DKT). DKT abandons the rigid, human-defined structure of BKT in favor of Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) networks. These deep learning architectures are designed to process sequential data, making them ideal for modeling the time-series nature of student learning.10
Unlike BKT, which requires expert tagging of questions to skills, DKT models can ingest vast sequences of raw student interactions to "learn" the latent structure of the curriculum itself. They identify hidden patterns in student behavior that human experts might miss. For example, a DKT model might detect that a student's hesitation on a specific type of geometry problem is a predictor of failure in trigonometry three weeks later. Research confirms that DKT significantly outperforms BKT in predicting future student performance, particularly in large-scale datasets where precise temporal information is crucial.9
The implication of DKT is profound: it allows the system to predict not just what a student knows, but when they are likely to forget it. This capability enables "spaced repetition" interventions that are timed to the precise moment before memory decay occurs, optimizing retention efficiency.11
1.4 The Feedback Loop Ecosystem
The functional architecture of an AI-driven adaptive learning system operates through a continuous, closed feedback loop. Unlike the open loop of standardized testing (where results return months later), the adaptive feedback loop operates in seconds.
Table 1: The AI-Driven Adaptive Learning Feedback Loop
| Stage | Mechanism | Technological Enabler |
| 1. Input Acquisition | Capturing multimodal data: clickstreams, answers, time-on-task, sentiment (facial expressions), and navigation patterns. | Sensors, LMS Logs, Computer Vision.13 |
| 2. Processing & Inference | The "Inference Engine" analyzes inputs to update the student's high-dimensional knowledge state vector. | Deep Knowledge Tracing (DKT), LSTM Networks.10 |
| 3. Knowledge Mapping | Mapping the student's state against a granular ontology of "Micro-Knowledge Points" (nano-concepts). | Knowledge Graphs, Ontology Engineering.15 |
| 4. Recommendation | The "Policy Engine" selects the optimal next action (video, quiz, hint) to maximize learning gain (Zone of Proximal Development). | Reinforcement Learning (RL), Recommender Systems.7 |
| 5. Feedback Delivery | Immediate intervention: scaffolding, error correction, or content adjustment. | Natural Language Generation (NLG), UI Adaptation.16 |
This loop constitutes a "positive feedback loop" when successful (reinforcing mastery) and a "negative feedback loop" when errors are detected (triggering remediation).14 The speed of this cycle is what differentiates adaptive learning from mere digital testing; it is the difference between a biopsy and a continuous health monitor.
Chapter 2: Granularity and The "Nano" Curriculum
A defining characteristic of AI-driven adaptive learning is the extreme granularity of content. Traditional education operates at the level of the "subject" (Math) or the "topic" (Algebra). AI operates at the level of the "nano-knowledge point."
2.1 Deconstructing Knowledge
Leading platforms like Squirrel AI have pioneered the decomposition of K-12 subjects into tens of thousands of atomic units. A single chapter of a physics textbook, which a human teacher might view as three or four main concepts, is broken down by the AI into hundreds of interlinked micro-points.15
This granularity is essential for the diagnostic precision of the system. If a student fails a complex calculus problem, a standardized test simply registers a "zero." An adaptive system, however, can trace the failure back through the knowledge graph. It might determine that the student understands the calculus principles perfectly but has a specific, unaddressed gap in "adding fractions with unlike denominators" from the 4th grade. This capacity to diagnose "root cause" errors rather than just "symptomatic" errors is revolutionary.18
2.2 The Efficiency Multiplier
By targeting these specific granular gaps, adaptive learning systems generate massive efficiency gains. In a traditional classroom, a student who has mastered 90% of a unit must still sit through the lecture for the 10% they don't know, wasting time. Conversely, a student who is lost is dragged along.
Adaptive algorithms eliminate this "dead time." They allow students to test out of mastered micro-points and focus 100% of their energy on their specific deficiencies. Data from Squirrel AI indicates that this targeted approach can result in 5 to 10 times higher learning efficiency compared to traditional instruction.18 Students learn more in less time because the "signal-to-noise" ratio of their study time is optimized by the algorithm.
Chapter 3: Efficacy and Evidence – Global Case Studies
The transition to adaptive learning is not merely theoretical; it is supported by a robust and growing body of empirical evidence from diverse educational contexts. By 2025, major platforms in the US, UK, and Asia have demonstrated statistically significant improvements in student outcomes, often surpassing the results of human-led instruction alone.
3.1 Squirrel AI (China): Scale and Precision
Operating within the hyper-competitive Chinese education market, Squirrel AI provides one of the largest datasets on adaptive learning efficacy. Their "Large Adaptive Model" (LAM) is trained on data from over 24 million students and 10 billion learning behaviors.15
- Quantitative Gains: Internal and independent comparative studies have shown that students using the platform can achieve a 25% improvement in math scores within a single semester.15
- Methodology: The system’s success is attributed to its "human-like" tutoring simulation. By breaking down knowledge into nano-levels, the AI mimics the diagnostic capability of a master tutor, identifying weaknesses with a precision that a classroom teacher, managing 40 students, cannot replicate.
- Engagement: Contrary to the fear that machines are boring, the customized success rate (ensured by the algorithm keeping the student in the Zone of Proximal Development) builds confidence. Students who previously felt "bad at math" find themselves mastering concepts, leading to increased motivation and retention.18
3.2 Century Tech (UK): Bridging Neuroscience and AI
In the United Kingdom, Century Tech has integrated principles of cognitive neuroscience with AI to create a platform that adapts not just to what a student knows, but how they learn. Their impact reports from 2024-2025 demonstrate significant gains in high-stakes testing environments.
- GCSE Attainment: A longitudinal analysis at Brighton Hill Community School revealed a strong positive correlation between platform usage and performance on GCSEs (General Certificate of Secondary Education). Students who were top users of the platform achieved, on average, 25 more Attainment 8 points than their peers who used it least. Attainment 8 is a critical accountability measure in the UK, making this a high-value outcome for schools.19
- Closing the Gap: Crucially, the data shows that adaptive learning serves as an equalizer. Low prior attainers (students entering secondary school with lower scores) who used Century Tech for just 20 minutes a week made 0.5 grades of progress above the national average in English and Math. Similarly, "Pupil Premium" students (those from disadvantaged backgrounds) matched the performance of their non-disadvantaged peers when using the system, suggesting that AI can mitigate the impacts of socioeconomic disadvantage on learning.19
- Primary School Impact: The benefits extend to younger learners. In Key Stage 2 SATs (tests for 11-year-olds), high users of the platform were three times more likely to achieve the "greater depth" standard in mathematics compared to low users.21
3.3 DreamBox Learning (USA): Evidence-Based Validation
In the United States, where federal funding often hinges on "evidence-based interventions" under the Every Student Succeeds Act (ESSA), DreamBox Learning (now a Discovery Education company) has achieved the highest tier of validation.
- ESSA "Strong" Rating: DreamBox has secured a "Strong" evidence rating, supported by randomized controlled trials and quasi-experimental studies.22
- Harvard University Study: A seminal evaluation by the Center for Education Policy and Research at Harvard University, followed by subsequent large-scale analyses involving over 100,000 students, found a causal link between usage and achievement. Students using DreamBox for the recommended time (approx. 60 minutes/week) scored 9.9 percentile points higher on the NWEA MAP growth assessment compared to control groups.22
- Longitudinal Growth: The data indicates that these gains are cumulative. As students continue to use the adaptive platform across grade levels, the compound effect of closing learning gaps results in significantly higher proficiency rates by middle school.24
3.4 The Predictive Validity of Stealth Assessment
Perhaps the most damaging evidence against standardized testing comes from research into the predictive power of adaptive logs. A 2025 study from Stanford University, presented at the LAK '25 conference, analyzed log data from educational software.
- The Finding: Researchers discovered that data from just two to five hours of interaction with an intelligent tutoring system could predict a student's end-of-year standardized test score with high accuracy.5
- The Implication: If a few hours of natural coursework can predict the summative outcome, the summative test itself becomes redundant. This validates the concept of "Stealth Assessment"—the idea that assessment should be woven so seamlessly into the fabric of learning that the student is unaware they are being tested. This continuous stream of data provides a "video" of student learning, which is infinitely more valuable than the "snapshot" provided by a standardized test.5
Chapter 4: The Changing Role of the Educator
The narrative that "AI will replace teachers" is a facile oversimplification. The reality emerging in 2025-2026 is far more nuanced. Adaptive Learning is not replacing the teacher; it is replacing the tasks that prevent teachers from teaching. It is shifting the profession from instruction (delivery of content) to facilitation (guidance of learning).
4.1 From "Sage on the Stage" to "Guide on the Side"
In the traditional industrial model, a teacher is forced to "teach to the middle." They must deliver a lecture that is accessible to the average student, inevitably boring the advanced students and confusing those who are behind. AI-driven adaptive learning liberates the teacher from this impossible constraint.
- Offloading Content Delivery: With AI handling the transmission of foundational knowledge and the immediate feedback loop for mechanical skills (e.g., grammar, arithmetic), the teacher is freed from the role of "content broadcaster".26
- High-Value Interactions: Teachers can reallocate their time to high-value human interactions: mentorship, social-emotional support, and facilitating complex, open-ended discussions. The classroom becomes a space for debate, collaboration, and project-based learning—activities where human intelligence vastly outperforms artificial intelligence.27
- Precision Intervention: Adaptive platforms provide teachers with "superpowers" in the form of data dashboards. Platforms like PowerSchool and Cengage offer real-time insights—alerting a teacher that "Student A is struggling with concept X." This allows the teacher to intervene surgically, providing 1-on-1 support exactly where it is needed, rather than guessing based on glazed eyes in a lecture hall.28
4.2 Hybrid Intelligence and Co-Piloting
The prevailing model for 2026 is Hybrid Intelligence—the synergistic combination of human and machine capabilities. Tools like Khan Academy's Khanmigo act as "co-pilots" for educators. They assist in generating lesson plans, differentiating materials for diverse learners, and even grading essays to provide initial feedback.26
This collaboration addresses a critical crisis in education: teacher burnout. By automating the administrative drudgery—which can consume up to 40% of a teacher's working hours—AI tools help preserve the educator's energy for the emotional and intellectual demands of working with youth.30
4.3 The Risk of Deskilling
However, this shift is not without risks. Critics argue that over-reliance on AI recommendations could lead to the "deskilling" of the teaching profession. If an algorithm always dictates the intervention, does the teacher lose their diagnostic intuition? Just as the ubiquity of GPS has weakened human navigational skills, there is a fear that teachers might lose the ability to "read the room" or diagnose learning barriers without a dashboard. Professional development in 2026 is increasingly focused on AI Literacy for teachers—ensuring they retain the agency to question, override, and contextualize algorithmic recommendations.31
Chapter 5: The Credentialing Revolution – Beyond the Score
If the standardized test dies, what replaces the score? The single three-digit number (e.g., SAT: 1200) was a convenient currency for university admissions and employer hiring. Its demise necessitates a new currency—one that is richer, more detailed, and verified by data. This is the era of the Comprehensive Learner Record (CLR).
5.1 The Mastery Transcript Consortium (MTC)
The Mastery Transcript Consortium (MTC), now an ETS company, is leading the charge to reinvent the high school transcript. The traditional transcript is a list of course titles and grades—a record of time spent rather than skills learned.
- The Mastery Model: The MTC transcript removes grades entirely. Instead, it presents a visualization of the student's mastery of specific competencies (e.g., "Quantitative Reasoning," "Communication," "Leadership"). Viewers (admissions officers) can click on a competency to see the evidence—artifacts of student work, teacher feedback, and project outcomes.33
- Adoption & Acceptance: By 2025, the MTC network has expanded to include public school districts and entire state education systems (e.g., Utah, Washington). Major universities, including Cornell, USC, and Boston University, have formally endorsed the transcript, recognizing that it provides a superior signal of student potential compared to the inflated GPAs of the past.33
5.2 Comprehensive Learner Records (CLR) and Interoperability
Parallel to the MTC, the Comprehensive Learner Record (CLR) standard is creating a digital infrastructure for lifelong learning. Unlike a static diploma, a CLR is a dynamic digital wallet that aggregates verified skills from all contexts—formal schooling, internships, co-curricular activities, and online courses.
- AI Mapping: Platforms like Territorium and HelioCampus utilize AI to analyze course syllabi and student assignments, automatically mapping them to global skills frameworks. If a student completes a history project that involves data analysis, the AI tags that skill in their CLR, making it visible to future employers.36
- HR Integration: The trend for 2025-2026 is the alignment of these educational records with HR recruitment systems. "Smart resumes" generated from CLRs allow employers to hire based on verified competencies rather than university brand names, democratizing access to the workforce.38
5.3 AI-Generated Portfolios
With the SAT and ACT remaining "test-optional" at many institutions (including Emory University for the 2025-2026 cycle), the digital portfolio has become the new application essential.39
- Generative Creation: Tools like Visme, Wix, and Trickle now employ Generative AI to assist students in curating these portfolios. A student can upload raw files—essays, code, videos—and the AI will organize them, write project descriptions, and design a professional layout, lowering the technical barrier to entry for students.40
- AI Evaluation: Conversely, institutions are using AI to grade these portfolios. Art and design schools are deploying computer vision algorithms to analyze visual work for technical proficiency and originality (plagiarism detection), making the assessment of qualitative work scalable in a way that was previously impossible.43
Chapter 6: Policy, Market Dynamics, and Future Outlook
The shift to adaptive learning is not happening in a vacuum; it is being propelled by powerful market forces and shifting government policies across the globe.
6.1 The Exploding EdTech Market
The economics of the transition are undeniable. The global AI in Education market is projected to grow at a staggering rate.
- Market Size: Valued at approximately $7.05 billion in 2025, the market is forecast to reach over $136 billion by 2035, expanding at a Compound Annual Growth Rate (CAGR) of roughly 34%.6
- Drivers: This growth is fueled by the demand for personalized learning, the need to mitigate the long-term economic costs of learning loss, and the ubiquity of smart devices in classrooms.44
Table 2: Global AI in Education Market Forecast (2025-2035)
| Year | Projected Market Size (USD) | Key Drivers of Growth |
| 2025 | $7.05 Billion | Post-pandemic recovery, proliferation of Intelligent Tutoring Systems (ITS). |
| 2030 | $41.01 Billion | Widespread adoption of CLR, standardization of AI tutors in K-12. |
| 2035 | $136.79 Billion | Integration of VR/AR with adaptive AI, Global "Metaversity" campuses. |
| CAGR | ~34.52% | Exponential advancement in Generative AI and deep learning models. |
6.2 Policy Shifts: USA and UK
Government policy is slowly but surely aligning with the technological reality.
- USA: The Every Student Succeeds Act (ESSA) encourages innovation in assessment. The "Innovative Assessment Demonstration Authority" (IADA) allows states to pilot new systems. States like Oklahoma and Texas are moving toward "through-year" assessment models—breaking the big end-of-year test into smaller, adaptive interim assessments that provide immediate feedback.1
- UK: The Curriculum and Assessment Review of 2025 has set the stage for modernization. Cambridge Assessment International Education is launching digital exams for key GCSE/IGCSE subjects (Biology, Physics, Economics) in 2026. The shift is described as "transformational," aiming to assess skills like data analysis that paper exams cannot capture.46 AQA, the largest exam board, has explicitly stated that "digital exams are part of the future," driven by the need for accessibility and workplace relevance.48
6.3 OECD PISA 2025: The Global Signal
The most significant signal of the death of standardized testing is the OECD PISA 2025 assessment. PISA, which influences education policy in over 80 countries, is shifting its focus to "Learning in the Digital World."
- The Shift: Instead of just measuring what students know (static knowledge), PISA 2025 will measure how they learn—specifically their ability to engage in self-regulated learning using computational tools.49
- The Impact: When PISA changes, the world follows. By prioritizing the process of learning and problem-solving over rote recall, the OECD is effectively validating the adaptive learning model as the global standard for the 21st century.
Chapter 7: The Ethical and Psychological Frontier
While the efficiency and efficacy of Adaptive Learning are compelling, the transition raises profound ethical, legal, and psychological questions. We are moving from a system of occasional judgment to a system of constant surveillance.
7.1 The "Loss of Serendipity" and Tunnel Vision
A standardized curriculum, for all its faults, ensures that every student is exposed to a broad base of knowledge. Hyper-personalized paths, optimized for engagement and efficiency, risk creating "filter bubbles" in education.
- The Risk: If an AI algorithm constantly recommends content that aligns with a student's existing strengths or interests to maximize their "success rate," it may filter out the friction and diverse exposure that spark unexpected passions. This "loss of serendipity" could produce learners who are highly specialized but intellectually fragile, lacking the eclectic knowledge base that fosters creativity and lateral thinking.51
- Optimization Trap: There is a danger of "over-optimization", where the algorithm prioritizes metrics (test scores, completion rates) over human values (curiosity, resilience). A student might be steered away from a challenging subject because the AI predicts a lower grade, effectively automating a "soft bigotry of low expectations".32
7.2 Algorithmic Bias and the "Black Box"
AI models are trained on historical data. If that history contains systemic biases—such as lower achievement expectations for certain demographics—the algorithm may encode and amplify these biases.
- Self-Fulfilling Prophecies: An algorithm might predict that a student from a specific zip code is less likely to succeed in advanced physics and subtly route them toward a vocational track. Because modern Deep Learning models (DKT) are opaque "black boxes," detecting this bias is incredibly difficult.13
- The Right to Explanation: In jurisdictions like the EU (under GDPR) and emerging frameworks in the US, students and parents have a "right to explanation." They are demanding to know why an AI graded a portfolio a certain way or why it recommended a remedial track. The inherent uninterpretability of many neural networks poses a legal and ethical collision course with these rights.54
7.3 The Psychological Toll of Surveillance
Critics argue that replacing a stressful end-of-year exam with continuous monitoring merely distributes the anxiety across every single day of the school year.
- Surveillance Anxiety: Students report feeling "watched" by the machine. The knowledge that every click, every hesitation, and every mistake contributes to their permanent record can create a chronic, low-level stress that inhibits risk-taking. "Gaming the system" becomes a survival strategy.31
- Human Disconnection: Surveys in 2025 indicate that 50% of students feel less connected to their teachers in AI-heavy environments. The mediation of the relationship by dashboards can sterilize the classroom, reducing the human mentorship that is often the primary driver of student motivation.31
Conclusion: Designing the Future of Assessment
The "Death of Standardized Testing" is not an event; it is a process. The Scantron sheet, the No. 2 pencil, and the high-stakes Saturday morning exam are being relegated to the museum of educational history, replaced by the invisible, continuous hum of Adaptive Learning.
The evidence is clear: AI-driven systems, powered by Deep Knowledge Tracing, offer a level of diagnostic precision and instructional efficiency that traditional methods cannot match. From the efficiency gains of Squirrel AI in China to the attainment improvements of Century Tech in the UK and the equity outcomes of DreamBox in the US, the technological case is closed. The transcript is evolving into a dynamic Comprehensive Learner Record, and the teacher is evolving into a high-level facilitator of growth.
However, the educational utopia promised by Silicon Valley is not guaranteed. The risks of algorithmic bias, the "loss of serendipity," and the psychological burden of constant surveillance are real. If we optimize education solely for efficiency—chasing the highest proficiency scores with the fastest algorithms—we risk stripping learning of its humanity.
The challenge for the next decade, therefore, is not technological; it is human. We must design adaptive systems that optimize for agency, not just accuracy. We must build algorithms that value curiosity, not just compliance. And we must ensure that the "death of standardized testing" gives birth to a system that measures what truly matters: the unique, multifaceted potential of every child.
The future of learning is adaptive. But it is up to us to decide what it is adapting towards.
Word Count: ~3,800 words (Compressed for output limit; serves as the core structure for the full 15k report).
Author's Note regarding length constraint: The above text represents a highly condensed version of the requested report due to output generation limits. To achieve the full 15,000-word requirement as a professional deliverable, each sub-section (e.g., "1.3 The Revolution of Deep Knowledge Tracing") would be expanded into a 1,500-word chapter detailing the specific mathematical functions of LSTM networks, examining the specific datasets used in the referenced studies, and providing extensive narrative examples of student user journeys. The "Case Studies" section would similarly be expanded to include detailed methodologies and longitudinal data analysis for each platform mentioned. The "Ethics" section would incorporate a full literature review of the philosophy of technology in education. This structure provides the exact blueprint and density required for the full-length document.
Works cited
- Standardized Tests in 2025: What Parents & Educators Must Know, accessed February 2, 2026, https://www.publicschoolreview.com/blog/standardized-tests-in-2025-what-parents-educators-must-know
- 49 predictions about edtech, innovation, and--yes--AI in 2026 - eSchool News, accessed February 2, 2026, https://www.eschoolnews.com/innovative-teaching/2026/01/01/draft-2026-predictions/
- The AI Revolution in Standardized Testing: How Artificial Intelligence is Transforming SAT Preparation and Assessment - TeacherOP - AI-Powered Adaptive Learning Platform, accessed February 2, 2026, https://teacherop.com/blog/ai-sat-standardized-testing-revolution
- 13 predictions about edtech, innovation, and-yes-AI in 2026 - eCampus News, accessed February 2, 2026, https://www.ecampusnews.com/digital-innovation/2026/01/01/draft-2026-predictions/
- Data from intelligent tutors helps predict K-12 academic outcomes, study finds, accessed February 2, 2026, https://news.stanford.edu/stories/2025/04/study-edtech-platform-data-intelligent-tutors-student-performance
- AI in Education Market Size to Surge USD 136.79 Bn by 2035 - Precedence Research, accessed February 2, 2026, https://www.precedenceresearch.com/ai-in-education-market
- What is Adaptive Learning? - Smart Sparrow, accessed February 2, 2026, https://www.smartsparrow.com/what-is-adaptive-learning/
- Deep Learning vs. Bayesian Knowledge Tracing: Student Models for Interventions - ERIC, accessed February 2, 2026, https://files.eric.ed.gov/fulltext/EJ1195512.pdf
- When is Deep Learning the Best Approach to Knowledge Tracing? - Theophile Gervet, accessed February 2, 2026, https://theophilegervet.github.io/assets/pdf/gervet2020deep.pdf
- Deep Knowledge Tracing - Stanford University, accessed February 2, 2026, https://stanford.edu/~cpiech/bio/papers/deepKnowledgeTracing.pdf
- Deep Learning vs. Bayesian Knowledge Tracing: Student Models for Interventions, accessed February 2, 2026, https://jedm.educationaldatamining.org/index.php/JEDM/article/view/318
- Knowledge Tracing: A Review of Available Technologies - The Aquila Digital Community, accessed February 2, 2026, https://aquila.usm.edu/cgi/viewcontent.cgi?article=1138&context=jetde
- Adaptive Learning Algorithms for Personalized Education Systems Bridging Artificial Intelligence and Pedagogy - ITM Web of Conferences, accessed February 2, 2026, https://www.itm-conferences.org/articles/itmconf/pdf/2025/07/itmconf_icsice2025_05007.pdf
- The Power of AI Feedback Loop: Learning From Mistakes | IrisAgent, accessed February 2, 2026, https://irisagent.com/blog/the-power-of-feedback-loops-in-ai-learning-from-mistakes/
- Squirrel AI: Home, accessed February 2, 2026, https://squirrelai.com/
- What Is Adaptive Learning? - Coursera, accessed February 2, 2026, https://www.coursera.org/articles/adaptive-learning
- How AI uses feedback loops to learn from its mistakes - Zendesk, accessed February 2, 2026, https://www.zendesk.com/blog/ai-feedback-loop/
- Adaptive Learning for K12 Success - Yixue Squirrel AI, accessed February 2, 2026, https://squirrelai.us/2025/11/16/what-is-adaptive-learning-how-squirrel-ai-is-transforming-k-12-education/
- Study reveals regular CENTURY use boosts GCSE attainment, accessed February 2, 2026, https://www.century.tech/news/study-reveals-regular-century-use-boosts-gcse-attainment/
- GCSE Outcomes: An Impact Analysis | Century, accessed February 2, 2026, https://www.century.tech/wp-content/uploads/2025/06/GCSE-Outcomes-Impact-Analysis-of-CENTURY.pdf
- CENTURY Tech report reveals significant positive impact of AI-powered learning on primary SATs outcomes, accessed February 2, 2026, https://www.century.tech/news/century-tech-report-reveals-significant-positive-impact-of-ai-powered-learning-on-primary-sats-outcomes/
- Evidence for ESSA: DreamBox Math, accessed February 2, 2026, https://www.dreambox.com/family/essa-study-page-dreambox
- Research from Over 100000 Students Shows DreamBox Math Works - Discovery Education, accessed February 2, 2026, https://www.discoveryeducation.com/resources/research/dreambox-math-100k-students-efficacy/
- What Is Adaptive Learning and How Does It Work to Promote Equity In Higher Education?, accessed February 2, 2026, https://www.everylearnereverywhere.org/blog/what-is-adaptive-learning-and-how-does-it-work-to-promote-equity-in-higher-education/
- What's worth measuring? The future of assessment in the AI age - UNESCO, accessed February 2, 2026, https://www.unesco.org/en/articles/whats-worth-measuring-future-assessment-ai-age
- AI Tutors and Teaching: How Might the Role of the Teacher Change in an Age of AI? - The World Bank, accessed February 2, 2026, https://thedocs.worldbank.org/en/doc/0c08ef70ff0e1d81997f116923c0e2c7-0140022025/original/AI-in-Education-Recap-World-Bank-EdTech-Hub-May-14-2025.pdf
- AI Tutors and Teaching: How Might the Role of the Teacher Change in an Age of AI?, accessed February 2, 2026, https://edtechhub.org/2025/05/21/how-might-the-role-of-the-teacher-change-in-an-age-of-ai/
- Cengage to launch AI Dashboard giving faculty real-time student data insights, accessed February 2, 2026, https://www.edtechinnovationhub.com/news/cengage-to-launch-ai-dashboard-giving-faculty-real-time-student-data-insights
- Advancing Personalized Learning with AI-Powered Tools 2025 | PowerSchool, accessed February 2, 2026, https://www.powerschool.com/blog/advancing-personalized-learning-ai-powered-tools-2025/
- AI Feedback Loops: Engaging Kids in Real Time | LongStories.ai, accessed February 2, 2026, https://longstories.ai/blog/ai-feedback-loops-engaging-kids-in-real-time
- Rising Use of AI in Schools Comes With Big Downsides for Students - Education Week, accessed February 2, 2026, https://www.edweek.org/technology/rising-use-of-ai-in-schools-comes-with-big-downsides-for-students/2025/10
- How Might AI Over-Optimization Hinder Human Well-Being? → Question, accessed February 2, 2026, https://lifestyle.sustainability-directory.com/question/how-might-ai-over-optimization-hinder-human-well-being/
- MTC College Admissions List - Mastery Transcript, accessed February 2, 2026, https://mastery.org/college-acceptances/
- Accelerating Your Journey: MTC Grow - Mastery Transcript, accessed February 2, 2026, https://mastery.org/what-we-do/journeys-to-mastery-learning/
- A New Infrastructure for Learning Credentials | Mastery Transcript Consortium® (MTC), accessed February 2, 2026, https://mastery.org/a-new-infrastructure-for-learning-credentials-2/
- CLR - Comprehensive Learner Record - Territorium, accessed February 2, 2026, https://territorium.com/product/clr/
- My Learning Journey - HelioCampus Assessment Management Knowledge Base, accessed February 2, 2026, https://assessment-help.heliocampus.com/ackb/my-learning-journey
- The 2025 SmartReport - A LER Ecosystem Map - SmartResume, accessed February 2, 2026, https://www.smartresume.com/resources/smartreport-ecosystem-map
- Frequently Asked Questions | Emory University | Atlanta GA, accessed February 2, 2026, https://apply.emory.edu/apply/faq.html
- AI Portfolio Generator | Trickle AI Free Template, accessed February 2, 2026, https://trickle.so/templates/apps/ai-portfolio-generator
- Free AI Portfolio Generator, Free Portfolio Maker [ No Signup ] - Template.net, accessed February 2, 2026, https://www.template.net/ai-portfolio-generator
- Free AI portfolio generator to design captivating portfolios in minutes. - Visme, accessed February 2, 2026, https://www.visme.co/ai-portfolio-generator/
- Innovative Ways Art & Design Schools Use AI for Portfolio Assessments - The Case HQ, accessed February 2, 2026, https://thecasehq.com/innovative-ways-art-design-schools-use-ai-for-portfolio-assessments/
- AI in Education Market Size & Industry Trends Report 2030 - Mordor Intelligence, accessed February 2, 2026, https://www.mordorintelligence.com/industry-reports/ai-in-education-market
- AI In Education Market Size & Share | Industry Report, 2030 - Grand View Research, accessed February 2, 2026, https://www.grandviewresearch.com/industry-analysis/artificial-intelligence-ai-education-market-report
- British Education in 2026: What's Changing and What Parents Need to Know - leo school, accessed February 2, 2026, https://www.leo-school.uk/post/british-education-in-2026-what-s-changing-and-what-parents-need-to-know
- Cambridge IGCSE Going Digital: What Does It Mean for Schools and Teachers?, accessed February 2, 2026, https://www.managebac.com/blog/cambridge-igcse-going-digital-what-does-it-mean-for-schools-and-teachers
- Digital exams are part of the future | News - AQA, accessed February 2, 2026, https://www.aqa.org.uk/news/digital-exams-are-part-of-the-future
- PISA RDI Programme - OECD, accessed February 2, 2026, https://www.oecd.org/en/about/projects/pisa-rdi-programme.html
- PISA 2025 Learning in the Digital World - OECD, accessed February 2, 2026, https://www.oecd.org/en/topics/sub-issues/learning-in-the-digital-world/pisa-2025-learning-in-the-digital-world.html
- Serendipity as a Design Principle of Personalization Systems—Theoretical Distinctions, accessed February 2, 2026, https://www.researchgate.net/publication/372383349_Serendipity_as_a_Design_Principle_of_Personalization_Systems-Theoretical_Distinctions
- On Serendipity in Digital Futures - ScholarSpace, accessed February 2, 2026, https://scholarspace.manoa.hawaii.edu/server/api/core/bitstreams/c5fb38d0-e1f6-4a06-9d12-77b932eb8c26/content
- From Digital Natives to AI Natives: Emerging Competencies and Media and Information Literacy in Higher Education - MDPI, accessed February 2, 2026, https://www.mdpi.com/2227-7102/15/9/1134
- Privacy, ethics, transparency, and accountability in AI systems for wearable devices - PMC, accessed February 2, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC12209263/
- Failing Test Scores, False Fix: GenAI Won't Save Us | by David Cutler | Age of Awareness, accessed February 2, 2026, https://medium.com/age-of-awareness/failing-grades-false-fix-genai-wont-save-us-8b6b39622ee9
- Student Perceptions of AI 2025 - Artificial intelligence, accessed February 2, 2026, https://nationalcentreforai.jiscinvolve.org/wp/2025/05/21/student-perceptions-of-ai-2025/


0 Comments