Education Theory & Learning Technology study guide · Lesson 10 of 10

Essential Vocabulary & Quick Recall

AI & adaptive-systems terms · Generative AI & system infrastructure terms · Pedagogy & instruction terms · Learning-science terms · Assessment, data & standards terms · Quick recall (mixed)

This is the lesson's study guide from the FlashFeed app, plus its key facts and a self-test.

What you'll be able to do

  • Define the adaptive-systems and knowledge-tracing terms (ALS, ITS, Cognitive Tutor, Knowledge State, BKT, DKT, algorithm-based sequencing).
  • Define the generative-AI and infrastructure terms (GenAI, Agentic AI, zero-shot learning, prompt engineering, HITL, affective computing, ontology).
  • Define the pedagogy and instruction terms (metacognition, differentiated vs personalized learning, microlearning, multimodal, PBL, scaffolded curriculum).
  • Define the core learning-science terms (working memory, chunking, transfer, PCK, EdTech) and SDT's three needs.
  • Define the assessment, data, and standards terms (formative vs summative assessment, learner analytics, Kirkpatrick, xAPI, MOOC, syllabus).

AI & adaptive-systems terms

The core vocabulary for software that observes a learner, models what they know, and adjusts what comes next.

An Adaptive Learning System (ALS) is software that dynamically adjusts content difficulty, sequence, and modality based on ongoing learner performance data. At the human-teaching end of the spectrum sits the Intelligent Tutoring System (ITS) — software providing personalized instruction and feedback without a human teacher, using models of the domain, the learner, and pedagogy. The Cognitive Tutor, developed at Carnegie Mellon, is the canonical ITS: it models student cognition and gives personalized hints and feedback within a specific domain, originally mathematics.

Underpinning all of these is the idea of a Knowledge State — a model representing what a specific learner currently knows and doesn't know, updated continuously by the system. Two competing approaches estimate that state. Bayesian Knowledge Tracing (BKT) is a probabilistic model that estimates the probability a learner has mastered a skill, updating after each response. Deep Knowledge Tracing (DKT) replaces the hand-crafted probability tables with a recurrent neural network (LSTM) that models learner knowledge as a hidden state evolving with each interaction (Piech et al., 2015).

Once the knowledge state is estimated, Algorithm-Based Sequencing uses ML models — often BKT or DKT — to predict that state and determine the optimal next item to present. The loop is: observe → update state → select next item → repeat.

Generative AI & system infrastructure terms

The vocabulary for understanding how modern foundation models slot into edtech pipelines — from raw AI capabilities down to curriculum representation.

Generative AI (GenAI) refers to AI systems that produce novel text, images, or audio; in education it enables dynamic content creation, conversational tutoring, and personalized feedback at scale. A step further is Agentic AI — autonomous AI that can plan, take actions, use tools, and pursue multi-step goals with minimal human intervention, capable of orchestrating curriculum planning, content generation, assessment, and feedback in a single loop.

Practical use of these systems requires Prompt Engineering: designing the inputs given to LLMs so they produce the desired instructional outputs — a core skill for edtech builders working with foundation models. A related capability is Zero-Shot Learning (AI), an LLM's ability to perform a task it was not explicitly trained on, generalizing from broad pre-training — useful for generating curriculum on rare or niche subjects.

Responsible deployment calls for Human-in-the-Loop (HITL) design, where humans review, approve, or override AI outputs before delivery — the standard check against errors in edtech. A more experimental layer is Affective Computing: AI that detects and responds to learner emotions — frustration, boredom, curiosity — to adapt instruction accordingly.

At the structural layer, a Curriculum Graph / Knowledge Graph is a representation of learning objectives and their dependencies, used by adaptive systems to find prerequisite relationships and optimal learning paths. Formalizing that structure further is an Ontology (Learning) — a representation of knowledge and the relationships between concepts in a domain, used to organize curriculum graphs and content recommendations.

Pedagogy & instruction terms

The vocabulary for describing how instruction is shaped, sequenced, and matched to individual learners.

The broadest term here is Personalized Learning — an instructional approach that tailors pace, content, sequence, and modality to the individual learner's needs, goals, and learning style. Its classroom precursor is Differentiated Instruction, where a teacher tailors content, process, or product to individual needs within a group setting. Both concepts feed the algorithmic vision of adaptive systems.

Supporting learners as they grow is the job of a Scaffolded Curriculum — a structured sequence that provides appropriate support at each level and fades assistance as competence grows, directly implementing Vygotsky's zone of proximal development. Competency-Based Education (CBE) takes a related stance: progression depends on demonstrating mastery of specific competencies, not time spent in instruction.

Delivery format matters too. Microlearning delivers content in short, focused bursts of two to ten minutes, each built around a single learning objective — well suited to mobile usage and limited attention spans. Multimodal Learning goes broader, engaging multiple sensory channels — visual, auditory, kinesthetic — together to deepen encoding, grounded in dual coding and cognitive load theories.

Problem-Based Learning (PBL) shifts the design entirely: students learn by investigating and solving an authentic, complex, open-ended problem, rooted in Dewey's experiential philosophy. Binding all of these together is Metacognition — thinking about one's own thinking, awareness of one's learning process, knowledge gaps, and strategies. AI tutors like Khanmigo foster it by prompting students to explain their reasoning rather than just supplying answers.

Learning-science terms

Six concepts from cognitive psychology and motivation research that explain why learning succeeds or stalls.

Two terms concern memory and processing. Working Memory is the cognitive system that temporarily holds and manipulates information during active processing — roughly 7±2 items at a time. Cognitive load theory is built on its limits: overload it and learning degrades. Chunking is the direct response — breaking complex information into smaller, more manageable units to reduce that load and improve encoding.

The ultimate measure of learning is Transfer of Learning: applying knowledge or skills from one context to new, different ones. It is the goal that shallow memorization typically fails to reach; deep, connected understanding is what transfers.

On the teaching side, Lee Shulman's 1986 concept of Pedagogical Content Knowledge (PCK) distinguishes knowing a subject from knowing how to teach it. A chemistry expert and a chemistry teacher have different knowledge; PCK names what the teacher has that the expert may lack.

Motivation is addressed by Self-Determination Theory (SDT) (Deci & Ryan, 1985), which names three basic psychological needs for intrinsic motivation: autonomy, competence, and relatedness. It is the theoretical foundation most often cited in gamification and engagement design.

Finally, EdTech is the broad term for any technology used to facilitate or enhance learning — the field in which all the above concepts are applied.

Assessment, data & standards terms

The vocabulary for measuring learning outcomes, tracking learner data, and situating edtech within the broader education landscape.

Assessment splits into two distinct modes. Formative Assessment happens during the learning process — low-stakes quizzes, check-ins, retrieval prompts — providing feedback and allowing instruction to adjust. Summative Assessment evaluates learning after instruction: exams, final projects, grades. The distinction matters because the two serve different purposes and carry different stakes.

Evaluating whether training worked at the organizational level calls for the Kirkpatrick Model, a four-level framework: Reaction (did learners like it?), Learning (did they acquire knowledge?), Behavior (did they apply it?), and Results (did it affect outcomes?). It remains the standard evaluation structure in corporate and institutional training.

Data infrastructure has its own vocabulary. Learner Analytics covers the measurement, collection, analysis, and reporting of data about learners and their contexts, with the aim of optimizing learning. xAPI / Tin Can API is the interoperability standard that captures experience statements across all learning contexts — not just inside an LMS — enabling analytics that follow the learner wherever they learn.

Three category terms round out the field. A MOOC (Massive Open Online Course) is a large-scale online course open to unlimited participation; Coursera, edX, and Udemy established the model. STEM / STEAM names the disciplinary groupings — Science, Technology, Engineering, (Arts), Mathematics — most commonly addressed by edtech platforms. A Syllabus / Curriculum Map is the structured outline of learning objectives, content, sequence, and timeline for a course — the planning artifact that sits behind every adaptive system.

Key terms

Adaptive Learning System (ALS)
Software that dynamically adjusts content difficulty, sequence, and modality based on ongoing learner performance data.
Intelligent Tutoring System (ITS)
Software that provides personalized instruction and feedback without a human teacher in the loop, using models of the domain, the learner, and pedagogy.
Cognitive Tutor
An AI tutoring system developed at Carnegie Mellon that models student cognition and gives personalized hints and feedback within a specific domain, originally mathematics.
Knowledge State
A model representing what a specific learner currently knows and doesn't know within a domain, updated continuously by the system.
Bayesian Knowledge Tracing (BKT)
A probabilistic model used in intelligent tutoring systems to estimate the probability that a learner has mastered a skill, updating that estimate after each response.
Deep Knowledge Tracing (DKT)
A recurrent neural network (LSTM) approach to knowledge tracing that models learner knowledge as a hidden state evolving with each interaction (Piech et al., 2015).
Algorithm-Based Sequencing
Using ML models (often BKT or DKT) to predict learner knowledge states and determine the optimal next item to present.
Generative AI (GenAI)
AI systems that produce novel text, images, or audio. In education, GenAI enables dynamic content creation, conversational tutoring, and personalized feedback at scale.
Agentic AI
Autonomous AI that can plan, take actions, use tools, and pursue goals across multiple steps with minimal human intervention. In education it can orchestrate curriculum planning, content generation, assessment, and feedback in one loop.
Zero-Shot Learning (AI)
An LLM's ability to perform a task it was not explicitly trained on, generalizing from broad pre-training — relevant to generating curriculum for rare or niche subjects.
Prompt Engineering
Designing the inputs given to LLMs so they produce the desired instructional outputs — a practical skill for edtech builders working with foundation models.
Human-in-the-Loop (HITL)
AI system design where humans review, approve, or override AI outputs before delivery — used in responsible edtech to catch errors and ensure quality.
Affective Computing
AI systems that detect and respond to learner emotions — frustration, boredom, curiosity — to adapt instruction accordingly.
Curriculum Graph / Knowledge Graph
A structured representation of learning objectives and their dependencies, used by adaptive systems to find prerequisite relationships and optimal learning paths.
Ontology (Learning)
A formal representation of knowledge and the relationships between concepts in a domain, used to structure curriculum graphs and content recommendations.
Metacognition
Thinking about one's own thinking — awareness of one's learning process, knowledge gaps, and strategies. Tutors like Khanmigo foster it by asking students to explain their reasoning.
Differentiated Instruction
Tailoring content, process, or product to individual learner needs within a classroom setting — the human precursor to algorithmic personalization.
Personalized Learning
An instructional approach that tailors pace, content, sequence, and modality to the individual learner's needs, goals, and learning style.
Microlearning
Delivering content in short, focused bursts of 2–10 minutes, each built for a single learning objective. It exploits attention spans and mobile usage patterns.
Multimodal Learning
Engaging multiple sensory channels — visual, auditory, kinesthetic — together to deepen encoding. Grounded in dual coding and cognitive load theories.
Problem-Based Learning (PBL)
An approach where students learn by investigating and solving an authentic, complex, open-ended problem — rooted in Dewey's experiential philosophy.
Scaffolded Curriculum
A structured learning sequence that provides appropriate support at each level and fades assistance as competence grows — directly implementing Vygotsky's ZPD.
Competency-Based Education (CBE)
An educational model where progression depends on demonstrating mastery of specific competencies, not on time spent in instruction.
Pedagogical Content Knowledge (PCK)
Shulman's 1986 concept: the specific knowledge teachers need about how to teach a particular subject, distinct from knowing the subject matter itself.
Transfer of Learning
Applying knowledge or skills learned in one context to new, different contexts. It is the ultimate goal of deep learning; shallow memorization often fails to transfer.
Working Memory
The cognitive system that temporarily holds and manipulates information during active processing (about 7±2 items). Cognitive load theory focuses on its limits.
Chunking
Breaking complex information into smaller, more manageable units to reduce cognitive load and improve encoding.
Self-Determination Theory (SDT)
Deci & Ryan's theory (1985) naming three basic psychological needs for intrinsic motivation: autonomy, competence, and relatedness. A framework for engagement and gamification design.
EdTech
Education technology — any technology used to facilitate or enhance learning.
Formative Assessment
Assessment during the learning process to provide feedback and adjust instruction. Low-stakes quizzes that trigger retrieval practice are an example.
Summative Assessment
Evaluation of learning after instruction — exams, final projects — to grade outcomes. Contrasted with formative assessment.
Learner Analytics
The measurement, collection, analysis, and reporting of data about learners and their contexts in order to optimize learning.
Kirkpatrick Model
A four-level framework for evaluating training effectiveness: Reaction, then Learning, then Behavior, then Results.
xAPI / Tin Can API
A learning-data interoperability standard that captures experience statements across all learning contexts, not just the LMS.
MOOC (Massive Open Online Course)
A large-scale online course open to unlimited participation. Coursera, edX, and Udemy pioneered the model.
STEM / STEAM
Science, Technology, Engineering, (Arts), and Mathematics — disciplinary categories commonly addressed by edtech platforms.
Syllabus / Curriculum Map
A structured outline of learning objectives, content, sequence, and timeline for a course.

Key concepts

Cognitive Tutor

  • The canonical, textbook example of an AI system that instructs a student one-on-one without any human teacher.
  • Developed at Carnegie Mellon, it models how a student thinks and delivers personalized hints and feedback within a specific domain, originally mathematics.

Bayesian Knowledge Tracing

  • One of two competing methods for estimating what a specific learner currently knows, updating after every response.
  • It is a probabilistic model that computes the probability a learner has mastered a skill, relying on hand-crafted probability tables.

Deep Knowledge Tracing

  • The newer of two competing methods for estimating a learner's evolving mastery after each interaction.
  • It replaces hand-crafted probability tables with a recurrent neural network (an LSTM) that models a learner's mastery as an evolving hidden state (Piech et al., 2015).

Knowledge State

  • A continuously updated model, at the heart of adaptive software, of what a specific learner does and doesn't understand.
  • It is precisely the thing two competing methods — one probabilistic, one a neural network — race to estimate.

Matched pairs

Each item on the left goes with the one on the right.

Passive learning
Observing, watching, reading — weak for retention
Anki / SuperMemo
SRS apps that schedule reviews at the optimal interval
B.F. Skinner
Operant conditioning
Lev Vygotsky
Zone of Proximal Development
Comprehension (1956)
Understand (2001)
Apply
use, execute, implement, solve
Siemens & Downes (2004–05)
Connectivism
Engagement (UDL)
The 'why' — multiple means of motivation and interest
Analyze (ADDIE)
Identify goals, audience, prior knowledge, and constraints
SAM (2012)
Rapid, agile prototyping; explicitly iterative
The Diamond Age (Neal Stephenson)
An AI tutor for a poor girl becomes her greatest educational resource
conmigo
Spanish for "with me" — the root of Khanmigo's name
Research Assistants
Document synthesis, multi-source knowledge (NotebookLM, Perplexity)

Self-test

Answer each question from memory, then check your work against the answer key.

  1. 1. The Cognitive Tutor, the canonical Intelligent Tutoring System, was developed at which university and originally for which subject?

    1. AStanford, originally computer science
    2. BMIT, originally introductory physics
    3. CHarvard, originally reading comprehension
    4. DCarnegie Mellon, originally mathematics
  1. 2. An Adaptive Learning System dynamically adjusts which three aspects of instruction, based on ongoing learner performance data?

    1. ARepresentation, expression, and engagement
    2. BContent difficulty, sequence, and modality
    3. CInstructional content, process, and product
    4. DAutonomy, competence, and relatedness
  2. 3. How does Deep Knowledge Tracing (Piech et al., 2015) differ from Bayesian Knowledge Tracing?

    1. AIt replaces the learner model with explicit models of the domain and the pedagogy (the other two ITS models)
    2. BIt replaces per-response updating with a single mastery estimate (a posterior) computed at the end of each unit
    3. CIt replaces hand-crafted probability tables with a recurrent neural network (LSTM) hidden state
    4. DIt replaces the neural hidden state with hand-crafted probability tables (one per skill)
  3. 4. A system maintains a model of what one specific learner currently knows and doesn't know, updating it continuously. What is that model called?

    1. AKnowledge State
    2. BLearner Analytics
    3. CLearning Ontology
    4. DKnowledge Graph
  4. 5. Algorithm-Based Sequencing runs a repeating loop. Which step is NOT part of that loop?

    1. AUpdate the estimated knowledge state
    2. BObserve the learner's response
    3. CSelect the optimal next item to present
    4. DAward a completion certificate once the unit timer expires
  5. 6. What most distinguishes Agentic AI from ordinary Generative AI in an edtech pipeline?

    1. AIt can produce novel text, images, or audio from a single prompt, generating content and feedback at scale
    2. BIt can detect a learner's frustration, boredom, or curiosity and adapt the next instruction to match
    3. CIt can autonomously plan, use tools, and pursue multi-step goals with minimal human intervention
    4. DIt can route every output through a human reviewer, who approves or overrides it before delivery
  6. 7. Zero-Shot Learning (AI) describes an LLM's ability to do what?

    1. AMaster a task after a few labelled examples appear in the prompt, generalizing from those demonstrations
    2. BPerform a task it was not explicitly trained on, generalizing from broad pre-training
    3. CTrain on data that carries no human-supplied labels, discovering structure in the raw corpus
    4. DRetain every prior learner interaction across sessions, drawing on that history at inference
  7. 8. Prompt Engineering, a core skill for edtech builders working with foundation models, is best described as which activity?

    1. ADesigning the inputs given to LLMs so they produce the desired instructional outputs
    2. BWriting the integration code so a model's outputs are served inside the school's LMS
    3. CRanking pairs of a model's outputs so a separate reward model learns instructional preferences
    4. DFine-tuning a model's weights on curriculum data so it produces domain-specific outputs
  8. 9. An edtech team requires a teacher to review and approve every AI-generated quiz before students see it. Which design principle is this?

    1. AAdaptive Learning System (ALS)
    2. BCompetency-Based Education (CBE)
    3. CHuman-in-the-Loop (HITL)
    4. DProblem-Based Learning (PBL)
  9. 10. How does an Ontology (Learning) relate to a Curriculum Graph?

    1. AIt logs each learner's interactions with the graph, scoring its prerequisite edges from that evidence
    2. BIt represents domain concepts and their relationships, organizing curriculum graphs and recommendations
    3. CIt is the per-learner layer of the graph, tracking what one student has mastered up to now
    4. DIt is the interoperability standard for the graph, carrying its objectives between LMS platforms
  10. 11. Microlearning delivers content in bursts of what length, and each burst is built around how many learning objectives?

    1. A30–60 seconds, built around a single learning objective
    2. B45–60 minutes, built around one unit's learning objectives
    3. C20–30 minutes, built around several learning objectives
    4. D2–10 minutes, built around a single learning objective
  11. 12. AI tutors like Khanmigo prompt students to explain their reasoning rather than just supplying answers. Which capacity is this designed to foster?

    1. AMetacognition
    2. BTransfer of learning
    3. CIntrinsic motivation
    4. DWorking memory
  12. 13. Which statement describes Differentiated Instruction rather than Personalized Learning?

    1. AAn adaptive system tailors pace, sequence, and modality to each learner's goals and learning style
    2. BA tutor tailors content, process, and product to a single learner in private tutoring
    3. CA teacher tailors each learner's rate of progression to demonstrated mastery rather than time spent
    4. DA teacher tailors content, process, or product to individual needs within a group setting
  13. 14. A Scaffolded Curriculum gives support at each level and fades assistance as competence grows, directly implementing whose idea?

    1. ACsikszentmihalyi's theory of flow
    2. BBloom's taxonomy of cognitive objectives
    3. CVygotsky's zone of proximal development
    4. DPiaget's stages of cognitive development
  14. 15. Which statement about these pedagogy and instruction terms is NOT accurate?

    1. AMultimodal Learning delivers one modality at a time so working memory is never overloaded
    2. BMicrolearning builds every short, focused burst around one learning objective only
    3. CCompetency-Based Education counts demonstrated mastery, never time spent, toward progression
    4. DProblem-Based Learning has students learn by investigating an authentic, open-ended problem
  15. 16. Self-Determination Theory names which three basic psychological needs that drive intrinsic motivation?

    1. AChallenge, fantasy, and curiosity
    2. BAutonomy, competence, and relatedness
    3. CContext, culture, and collaboration
    4. DAttention, retention, and reproduction
  16. 17. Working memory temporarily holds and manipulates information during active processing. In Miller's classic estimate, roughly how many items can it handle at once?

    1. AAbout 7±2
    2. BAbout 4±1
    3. CAbout 10±2
    4. DAbout 12±2
  17. 18. A designer splits a 12-digit code into three groups of four so learners can hold it more easily. Which technique is this?

    1. AScaffolding
    2. BChunking
    3. CMultimodal learning
    4. DTransfer of learning
  18. 19. Which statement about these learning-science terms is NOT accurate?

    1. ASelf-Determination Theory was put forward by Deci & Ryan in 1985
    2. BEdTech is any technology used to facilitate or enhance learning
    3. CWorking memory temporarily holds and manipulates information during active processing
    4. DTransfer of Learning means repeating material until it can be recalled verbatim
  19. 20. Lee Shulman's 1986 concept of Pedagogical Content Knowledge names which kind of knowledge?

    1. ACommand of a grade's curriculum standards, distinct from mastery of the subject itself
    2. BDeep expertise in a subject itself, distinct from any skill at teaching that subject
    3. CKnowing how to teach a specific subject, distinct from knowing the subject itself
    4. DGeneral skill at managing a classroom, distinct from expertise in any one subject
  20. 21. In the Kirkpatrick Model of training evaluation, which of the four levels asks whether learners actually applied what they learned back on the job?

    1. ABehavior
    2. BLearning
    3. CResults
    4. DReaction
  21. 22. Which of these is an example of Formative Assessment?

    1. AA final project graded once the course ends that fixes each student's overall mark
    2. BA low-stakes retrieval quiz during a unit that lets instruction adjust
    3. CA high-stakes midterm exam partway through a unit that counts toward the final grade
    4. DA certification test at the close of training that attests to a candidate's competence
  22. 23. The xAPI / Tin Can API standard exists to do what?

    1. AGenerate quiz items from a source document, so assessments can be produced at scale
    2. BRepresent learning objectives and their prerequisites, so optimal paths can be found
    3. CCapture experience statements across all learning contexts, so analytics follow the learner
    4. DEstimate the probability a learner has mastered a skill, so the next item can be chosen
  23. 24. Learner Analytics is defined as which of the following?

    1. ACapturing, storing, and exchanging experience statements across systems, to keep learning records portable
    2. BOutlining objectives, content, sequence, and timeline for a course, to plan what is taught when
    3. CRating reaction, learning, behavior, and results after training, to judge whether it worked
    4. DMeasuring, collecting, analyzing, and reporting data about learners and their contexts, to optimize learning
  24. 25. Which statement about these assessment, data, and standards terms is NOT accurate?

    1. ACoursera, edX, and Udemy established the MOOC model
    2. BSTEAM adds Assessment to the STEM disciplines
    3. CA syllabus / curriculum map outlines objectives, content, sequence, and timeline
    4. DA MOOC is a large-scale online course open to unlimited participation

Go deeper

Answer key

  1. 1. D Carnegie Mellon, originally mathematicsThe Cognitive Tutor was built at Carnegie Mellon. It models student cognition and delivers personalized hints and feedback within one domain — originally mathematics — making it the textbook example of an ITS.
  2. 2. B Content difficulty, sequence, and modalityAn ALS is defined by adjusting content difficulty, sequence, and modality from ongoing performance data. Content, process and product are differentiated instruction's levers; representation, expression and engagement are UDL; autonomy, competence and relatedness are SDT.
  3. 3. C It replaces hand-crafted probability tables with a recurrent neural network (LSTM) hidden stateDKT models learner knowledge as an LSTM hidden state evolving with each interaction, replacing BKT's hand-crafted probability tables. Both model the learner and update after every response, so the domain-and-pedagogy and end-of-unit options fail; replacing the network with tables is backwards.
  4. 4. A Knowledge StateThe Knowledge State is the per-learner model of what is and isn't known, updated continuously. A knowledge graph maps objectives and dependencies, a learning ontology formalizes a domain's concepts, and learner analytics measures and reports data — none is one learner's standing.
  5. 5. D Award a completion certificate once the unit timer expiresThe loop is observe, update state, select next item, repeat. Timer-based certificates are the opposite of algorithmic sequencing, which selects items from the predicted knowledge state.
  6. 6. C It can autonomously plan, use tools, and pursue multi-step goals with minimal human interventionGenerative AI produces content; Agentic AI goes further, autonomously planning, using tools, and pursuing multi-step goals — able to orchestrate curriculum planning, content generation, assessment, and feedback in a single loop.
  7. 7. B Perform a task it was not explicitly trained on, generalizing from broad pre-trainingZero-shot means performing an untrained task by generalizing from pre-training — useful for curriculum on rare subjects. In-prompt examples make it few-shot, learning from unlabelled data is unsupervised training, and remembering past sessions is a memory feature, not zero-shot.
  8. 8. A Designing the inputs given to LLMs so they produce the desired instructional outputsPrompt engineering is designing the inputs to LLMs so they produce the desired instructional outputs. Fine-tuning changes the model's weights — a different intervention entirely.
  9. 9. C Human-in-the-Loop (HITL)HITL design has humans review, approve, or override AI outputs before delivery — the standard check against errors in edtech. An ITS is defined by teaching without a human teacher, UDL is a framework for building flexible curricula, and GenAI is the technology whose quizzes the teacher is checking.
  10. 10. B It represents domain concepts and their relationships, organizing curriculum graphs and recommendationsAn ontology formalizes knowledge and concept relationships in a domain, and is used to organize curriculum graphs and content recommendations. The per-learner record of mastery is the Knowledge State, and the interoperability standard for learner data is xAPI, not an ontology.
  11. 11. D 2–10 minutes, built around a single learning objectiveMicrolearning delivers content in short, focused bursts of two to ten minutes, each built around a single learning objective — a format well suited to mobile usage and limited attention spans.
  12. 12. A MetacognitionMetacognition is thinking about one's own thinking — awareness of one's learning process, gaps, and strategies. Prompting learners to explain their reasoning targets exactly that awareness.
  13. 13. D A teacher tailors content, process, or product to individual needs within a group settingDifferentiated Instruction is the classroom precursor: teacher-led tailoring within a group. Tailoring pace, content, sequence, and modality to the individual is Personalized Learning, the broader term.
  14. 14. C Vygotsky's zone of proximal developmentScaffolding that fades as competence grows is a direct implementation of Vygotsky's zone of proximal development — the gap between solo and supported performance.
  15. 15. A Multimodal Learning delivers one modality at a time so working memory is never overloadedMultimodal Learning engages several sensory channels — visual, auditory, kinesthetic — together to deepen encoding; it does not isolate them one at a time. The other three statements are accurate.
  16. 16. B Autonomy, competence, and relatednessDeci & Ryan's Self-Determination Theory (1985) names autonomy, competence, and relatedness as the three needs behind intrinsic motivation — the framework cited most often in gamification and engagement design.
  17. 17. A About 7±2Miller's classic estimate is roughly 7±2 items, and cognitive load theory is built on working memory's limit: overload it and learning degrades. Cowan (2001) later argued the limit is nearer 4.
  18. 18. B ChunkingChunking breaks complex information into smaller, manageable units to reduce cognitive load and improve encoding — the direct response to working memory's limits.
  19. 19. D Transfer of Learning means repeating material until it can be recalled verbatimTransfer of Learning is applying knowledge or skills from one context to new, different ones — the goal shallow memorization fails to reach. Verbatim repetition is the opposite of transfer. The other three are accurate.
  20. 20. C Knowing how to teach a specific subject, distinct from knowing the subject itselfPCK distinguishes knowing a subject from knowing how to teach it: a chemistry expert and a chemistry teacher have different knowledge, and PCK names what the teacher has that the expert may lack.
  21. 21. A BehaviorKirkpatrick's four levels run Reaction (did they like it?), Learning (did they gain knowledge?), Behavior (did they apply it on the job?), and Results (did it affect outcomes?). Behavior is the application level.
  22. 22. B A low-stakes retrieval quiz during a unit that lets instruction adjustFormative assessment happens during learning — low-stakes quizzes, check-ins, retrieval prompts — giving feedback so instruction can adjust.
  23. 23. C Capture experience statements across all learning contexts, so analytics follow the learnerxAPI is the interoperability standard that captures experience statements wherever learning happens, not just inside an LMS. Mastery probability estimation is BKT's job; dependencies are a curriculum graph's.
  24. 24. D Measuring, collecting, analyzing, and reporting data about learners and their contexts, to optimize learningLearner Analytics covers measuring, collecting, analyzing, and reporting data about learners and their contexts to optimize learning. The exchange standard is xAPI; the four-level framework is Kirkpatrick.
  25. 25. B STEAM adds Assessment to the STEM disciplinesSTEAM adds Arts — Science, Technology, Engineering, Arts, Mathematics — not Assessment. The other three statements match the definitions exactly.

Sources

Download all 10 lessons: PDF · EPUB