Updated on
July 20, 2026
Copy-paste AI prompts for every level of Bloom's Taxonomy. Practical templates for Remember, Understand, Apply, Analyse, Evaluate and Create, with worked examples across subjects.
Many teachers using AI in the classroom unintentionally stay at the bottom of Bloom's taxonomy (Bloom, 1956). They ask AI to "summarise the key points" or "list the main causes", then wonder why the response feels thin. The tool is not the problem; the prompt is.
For the wider teaching sequence, see our guide to how to develop metacognition. The term describes a clear process for turning evidence into a classroom decision, rather than being a label on its own.
Anderson and Krathwohl (2001) revised Bloom's taxonomy (Bloom, 1956). It has six levels of thinking: Remember, Understand, Apply, Analyse, Evaluate, and Create. Each level needs different thinking skills. AI defaults to Remember or Understand when you do not specify a level.
This guide gives you copy-paste prompt templates for all six levels, worked examples showing the difference in AI output, and teacher tips on when each level is appropriate. Every template is tested across English, maths, science, and humanities so you can see how the structure transfers across subjects.
Our thinking frameworks guide places this approach alongside 41 other models, with advice on when each earns its place.
Sweller (1988) argued that problem solving can place a cognitive load on our limited working memory. Later cognitive load theory explains task complexity through element interactivity, which means how different parts of a task connect and affect one another (Chen, Paas & Sweller, 2023). The same principle applies to AI prompts: a vague instruction leads to a low-demand response because the model takes the easiest path through its training data. A structured prompt for a specific level makes the model produce content at a higher cognitive register.
The mechanism behind this matters for how you frame every AI prompt. Working memory is the cognitive system that holds and works with information in the moment. It can hold roughly four chunks at one time (Cowan, 2001). When a prompt is vague, learners must work out the task, remember the topic, create content and organise their response.
All these steps use the same limited working memory. When it becomes overloaded, learners fall back on the simplest response. Extraneous cognitive load is the mental effort caused by poor task or presentation design, rather than by genuine learning (Chandler & Sweller, 1991). A well-constructed AI prompt provides a scaffold and handles some of the design work. For example, "compare two causes using evidence from our source pack, structured as claim-evidence-analysis" gives learners the task structure, so they can focus on the comparison rather than working out what to do.
EdTechTeacher and similar sites offer generic "AI prompt templates" without cognitive scaffolding. They give teachers starter phrases but no framework for knowing whether those phrases produce the right level of thinking. The result is that teachers get AI outputs calibrated to Year 7 recall tasks even when they are preparing Year 11 evaluation questions.
Understanding Bloom's taxonomy is the first step. Using it to write AI prompts is the practical application that actually saves you time in the classroom.
State the thinking skill clearly in your instructions. For example, do not say, "Write about photosynthesis." Instead, say, "Analyse the relationship between light intensity and photosynthesis rate, noting limiting factors". This gives the learner a clear focus.
In Bloom et al.'s original taxonomy, Knowledge was the lowest cognitive category. In the revised taxonomy, the matching cognitive process is Remember (Bloom et al., 1956; Anderson & Krathwohl, 2001). At this level, you recall facts and definitions from memory. Karpicke (2008) found that retrieval practice helps learners retain knowledge for longer.
| Template | Subject Example |
|---|---|
| List the key events of [topic] in chronological order. | Year 7 History: List the key events of the Norman Conquest in chronological order. |
| Define the term [concept] as it is used in [subject]. | Year 9 Science: Define the term 'osmosis' as it is used in biology. |
| Name the [number] key [facts/terms/dates] associated with [topic]. | Year 8 Geography: Name the five key physical features of a meander. |
| Recall what [person/group] did during [event/period]. | Year 10 History: Recall what the suffragettes did during the campaign for women's votes 1905, 1914. |
AI is very good at Remember-level tasks. The challenge is that many teachers stop here without realising they have done so.
Prompts work well for lesson starters and quick vocabulary checks. They also help learners recall knowledge before new topics. Use prompts for low-stakes quizzes and creating revision resources. Prompts help learners succeed (Hattie, 2008; Black & Wiliam, 1998).
Prompt templates:
Worked example: Prompt: "List the key events of the Norman Conquest in chronological order." AI output: A numbered timeline from the death of Edward the Confessor in January 1066 through Harold's coronation, the battles of Gate Fulford and Stamford Bridge, Hastings on 14 October 1066, and William's coronation on Christmas Day. Clean, factual, appropriately pitched for Year 7.
Teacher tip: Use Remember prompts to generate question banks for formative assessment starters. Ask the AI to produce 10 retrieval questions on the topic, then pick the five that match your lesson objectives.
Learners build meaning rather than simply memorising facts. For example, a learner who lists the causes of WW1 without linking them is only showing recall. Anderson and Krathwohl (2001) say that understanding involves interpreting and comparing. It also includes summarising and explaining.
| Template | Subject Example |
|---|---|
| Explain in your own words how [concept/process] works. | Year 5 Science: Explain in your own words how the water cycle works. |
| Summarise the relationship between [A] and [B]. | Year 9 English: Summarise the relationship between Macbeth and Lady Macbeth at the start of Act 1. |
| Paraphrase [concept] so that a [year group] learner could understand it. | Year 6 Maths: Paraphrase the concept of equivalent fractions so that a Year 6 learner could understand it. |
| Give three examples that illustrate [concept/principle]. | Year 8 RE: Give three examples that illustrate the Buddhist concept of impermanence. |
The cognitive jump from Remember to Understand is where many AI prompts stall. Teachers ask for a "summary" but do not specify that the summary should show how ideas connect.
When to use: After introducing a new concept, when checking comprehension before a more complex task, and when learners need to process information in their own words. Understand prompts are effective for generating model texts that show learners how to explain ideas.
Prompt templates:
Worked example: Prompt: "Explain in your own words how the water cycle works for a Year 5 class." AI output: A clear three-paragraph explanation covering evaporation from oceans and lakes, condensation as water vapour rises and cools to form clouds, and precipitation as rain or snow returning to the surface. The language is appropriate for 9, 10 year olds without being patronising. This is a strong model text for learners to annotate or adapt.
Teacher tip: Pair Understand prompts with metacognitive questioning. After the AI generates an explanation, ask learners to identify which parts they found surprising and which parts they already knew. This activates prior knowledge and shows you where genuine understanding gaps exist.
Application involves using knowledge and procedures to carry out tasks in new situations. This is where abstract concepts become concrete tools. Maths word problems are the classic Apply-level task: the learner knows the formula but must recognise when and how to deploy it in an unfamiliar context.
| Template | Subject Example |
|---|---|
| Use [concept/formula/rule] to solve this problem: [problem context]. | Year 9 Maths: Use Pythagoras' theorem to solve this problem: a builder needs to cut a diagonal support beam across a rectangular doorframe that is 2.1m tall and 0.9m wide. How long should the beam be? |
| Demonstrate how [concept] would be applied in [real-world scenario]. | Year 10 Business: Demonstrate how the concept of supply and demand would apply to a bakery that introduces a new sourdough loaf during a local food festival. |
| Write a worked example that shows how to [process] step by step. | Year 7 Maths: Write a worked example showing how to find the area of a compound shape step by step, using a real-life L-shaped floor plan. |
| Create three practice problems that require learners to apply [concept] in different contexts. | Year 8 Science: Create three practice problems that require learners to apply their knowledge of density (mass ÷ volume) in different real-world contexts. |
Apply-level AI prompts generate worked examples, practice problems, and scenarios that require learners to use what they know. The key prompt move is to specify the context that makes the knowledge application non-trivial.
Apply prompts work best after you teach a concept, before testing. They help make varied practice tasks and real world examples. Prompts fit well in learning sequences. Learners can progress from teacher guidance to more independent work within their zone of proximal development (Vygotsky, 1978).
Prompt templates:
Worked example: Prompt: "Use Pythagoras' theorem to solve this: a builder needs to cut a diagonal beam across a doorframe that is 2.1m tall and 0.9m wide. How long should the beam be?" AI output: A fully worked solution with the formula stated, values substituted (2.1² + 0.9² = 4.41 + 0.81 = 5.22), square root calculated (approximately 2.28m), and a sentence contextualising the answer. This is a ready-to-use modelling resource.
Teacher tip: Apply prompts pair well with Rosenshine's (2012) principle of modelling. Use the AI output as your 'I do' demonstration, then give learners a similar problem to try as their 'we do.' The AI has done the heavy lifting of creating a well-structured worked example; you focus on the live explanation.
Learners analyse material by breaking it into parts and finding links between them. They then consider how each part fits within the whole structure (Anderson & Krathwohl, 2001). This process develops advanced thinking skills. Anderson and Krathwohl (2001) name differentiating, organising, and attributing as key skills.
| Understand Level | Analyse Level |
|---|---|
| Summarise the causes of World War One. | Explain how the alliance system transformed a regional dispute into a world war. |
| Explain what a food web shows. | Analyse what happens to the food web if the population of a top predator collapses. |
| Describe the features of a Shakespearean soliloquy. | Identify how Shakespeare uses Hamlet's soliloquy to reveal the contrast between thought and action. |
| Explain how a market economy works. | Break down how the 2008 financial crisis exposed structural weaknesses in deregulated markets. |
Analyse-level AI prompts require you to specify both the object of analysis and the analytical framework. "Analyse this poem" is too vague. "Analyse how Wilfred Owen uses imagery in 'Dulce et Decorum Est' to challenge the idea that war is glorious" is precise enough to produce a substantive response.
Learners need a deeper understanding to complete writing tasks and think within a subject. Teachers can use higher-order thinking tasks with any year group. However, the task difficulty and scaffolding must match what learners already know. Explicit instruction improves outcomes (Hattie, 2008; Rosenshine, 2012).
Prompt templates:
Worked examples help. Consider: "Compare power in 'My Last Duchess' and 'Ozymandias'." AI identified Browning's speaker's power through control (Browning, n.d.). Shelley used irony to critique power's illusion (Shelley, 1818). The response compares by concept, not poem, like GCSE schemes want.
Teacher tip: Webb's (1997) Depth of Knowledge framework defines level 3 as Strategic Thinking, which requires reasoning, planning and evidence. Level 4 is Extended Thinking, but Bloom's Analyse tasks do not always reach DOK level 3. When using AI to create analysis tasks, state the number of steps in your prompt so that the content is demanding enough. See Webb's Depth of Knowledge for a fuller guide to task calibration.
Evaluation means making a judgement based on set criteria and standards. Learners weigh evidence, consider different views and defend their ideas (Anderson & Krathwohl, 2001). Two key processes are "checking" for internal consistency and "critiquing" against external criteria (Anderson & Krathwohl, 2001).
| Template | Subject Example |
|---|---|
| Compare and contrast [A] and [B], focusing on [specific criteria]. | Year 10 English: Compare and contrast the way power is presented in 'My Last Duchess' and 'Ozymandias', focusing on the relationship between the speaker and their subject. |
| What are the causes and effects of [event/phenomenon]? Organise your answer by [short-term/long-term or direct/indirect]. | Year 9 History: What are the causes and effects of the Industrial Revolution? Organise by short-term and long-term effects on working-class life. |
| Break down how [process/system] works by identifying its component parts and the function of each. | Year 10 Science: Break down how the human immune system works by identifying its component parts and the function of each in fighting bacterial infection. |
| Identify the assumptions underlying [argument/policy/text]. Which assumptions are most open to challenge? | Year 11 Economics: Identify the assumptions underlying the case for free trade. Which assumptions are most open to challenge in the context of developing economies? |
Teachers can use Evaluate-level AI prompts when preparing lessons. AI can create arguments, counter-arguments, mark-scheme responses and "to what extent" essay frames. However, the teacher must still check that the subject content is accurate and show learners how to weigh evidence.
Evaluate prompts work well for GCSE and A-level prep. Use them in essay writing, debate, or teaching argument construction. They also help staff develop balanced teaching perspectives before discussions (Gibbs, 1988; Angelo & Cross, 1993).
Prompt templates:
AI produced an answer in four paragraphs. It matched the requirements for a high-band GCSE Geography response. The answer included a thesis and two arguments, one about education and one about policy.
It also noted that economic factors often drive education and policy. The conclusion gave a measured level of agreement with the prompt, making the example useful for modelling evaluation.
Teacher tip: Generate two versions of the same Evaluate prompt: one arguing strongly for the position and one arguing against. Use these as paired texts in a questioning sequence to help learners identify the analytical moves that distinguish a well-supported argument from a weak one.
Learners create by combining elements into a whole, or rearranging them (Anderson & Krathwohl, 2001). This is not just production. It means learners generate, plan and make something showing new thought.
| Template | Subject Example |
|---|---|
| To what extent do you agree that [claim]? Argue both sides, then reach a justified conclusion. | Year 11 Geography: To what extent do you agree that economic development is the most important factor in reducing a country's birth rate? Argue both sides, then reach a justified conclusion. |
| What are the strengths and limitations of [approach/theory/policy] when applied to [context]? | Year 12 Psychology: What are the strengths and limitations of the behaviourist approach when applied to explaining phobias in adults? |
| Judge whether [decision/action/policy] was justified, using [criteria] as your evaluative framework. | Year 10 History: Judge whether Chamberlain's policy of appeasement was justified, using the evidence available to British policymakers in 1938 as your evaluative framework. |
| Critique [text/argument/model] by identifying its strongest claim and its most significant weakness. | Year 11 English Language: Critique this opinion article by identifying its strongest rhetorical technique and its most significant logical weakness. |
AI at the Create level is most useful as a collaborator, not a producer. The best Create prompts use AI to generate constraints, criteria, or starting points that learners then work with. Giving learners the AI's first attempt and asking them to improve it also sits firmly at Create level.
By 2026, Create prompts should go beyond text. Ask learners to combine an image, a data set and an audio explanation. They can then use an agentic workflow, which is a series of AI-supported steps, to plan, create and critique an artefact.
This keeps the taxonomy's focus on synthesis (Bloom et al., 1956). Current multimodal systems can work with text, images, audio, video and code (Google DeepMind, 2025). Set subject criteria, provenance requirements and review points. Learners should explain each revision, while the system should not make the final creative decisions.
Prompt templates:
AI created an opening paragraph for a Gothic story. It includes pathetic fallacy, foreshadowing and an unreliable narrator. Teachers can use this AI text as a model. Learners can annotate or improve it as a starting point before producing their own writing.
Teacher tip: The most effective use of AI at Create level is generating the brief, not the final product. Ask AI to produce three different design briefs for a product, three different essay titles at different levels of difficulty, or three alternative starting points for a creative piece. Then learners choose and execute one. This keeps the creative decision-making with the learner.
Most AI prompting mistakes in classrooms come down to a mismatch between what the teacher wants and the cognitive level implied by the prompt wording. Use it as a starting point for professional discussion: identify the learner's current need, record evidence from more than one lesson, and agree the next classroom adjustment with the SENCO or family.
| Template | Subject Example |
|---|---|
| Design a [product/system/solution] that addresses [problem], specifying [constraints]. | Year 8 DT: Design a packaging solution for a fragile product that uses only recycled materials, must protect the item during postal delivery, and must be assembled without tools or adhesives. |
| Write a [genre] piece that demonstrates [concept/technique], aimed at [audience]. | Year 9 English: Write the opening of a Gothic short story that demonstrates the use of pathetic fallacy, foreshadowing, and an unreliable narrator. Aimed at a Year 9 reading level. |
| Propose a solution to [problem] that integrates knowledge from [subject area 1] and [subject area 2]. | Year 10 cross-curricular: Propose a solution to food insecurity in sub-Saharan Africa that integrates knowledge from geography (climate, water access) and science (crop modification, soil chemistry). |
| Construct a [argument/model/experiment/plan] that [achieves goal], explaining the reasoning behind each decision. | Year 12 Biology: Construct an experimental design to test whether increasing CO2 concentration affects the rate of photosynthesis in pondweed, explaining the reasoning behind each methodological decision. |
Mistake 1: Asking for lists when you need analysis. "List the effects of deforestation" produces a list of facts. In contrast, "Analyse how deforestation creates feedback loops that accelerate climate change" produces a reasoned chain of cause and effect. The subject matter is the same, but the cognitive demand is completely different.
Learners need clear frameworks. 'Discuss' is vague. For example, "Discuss the causes of the French Revolution" lacks focus. Instead, specify: "Discuss the relative importance of economic, social, and political causes, reaching a conclusion".
Mistake 3: Forgetting to specify audience and constraints. A Create prompt without constraints produces generic output. The constraint is what forces specificity. "Write a lesson plan" produces a mediocre template. "Write a 50-minute lesson plan on fractions for a Year 6 class where six learners have dyscalculia, using concrete resources before abstract notation" produces something genuinely useful.
Mistake 4: Accepting the first output. AI first drafts at higher Bloom's levels often begin well but turn into lists in later paragraphs. Read the whole response. If it slips back into Remember-level content, add this instruction to your prompt: "Do not use bullet points or numbered lists. Maintain analytical prose throughout."
Do not treat Bloom's taxonomy as a staircase. Paul (1993) argued that recall is not separable from knowledge in the neat way the hierarchy suggests, while Case (2013) showed how verb-led levels can misrepresent task difficulty. Genuine thinking loops: learners retrieve evidence, test an interpretation, revise it and explain the change. Use Bloom's taxonomy as a planning scaffold, then judge complexity through the knowledge used, the reasoning demanded and the quality criteria, not the verb alone.
The Thinking Framework covers eight ways of thinking: Remembering, Understanding, Applying, Analysing, Evaluating, Creating, Knowing about Knowing, and Knowing about Doing (Anderson & Krathwohl, 2001). It turns an abstract taxonomy into clear actions. This can increase learner engagement and attainment. It also provides a scaffold for higher order thinking.
The framework gives learners more control over their learning. It helps you move from teacher-centred to learner-centred instruction. By mirroring Bloom's taxonomy, it links theory with classroom lessons and names eight operations that match Bloom's levels (Anderson & Krathwohl, 2001). In this way, it makes abstract ideas practical, supports higher-order thinking and shifts the focus to learners.
| Level | Prompt | What Learners Produce |
|---|---|---|
| Remember | List the features of a river's upper course. | A bulleted list of geographical terms. |
| Analyse | Explain how the processes of erosion and deposition change as a river moves from its upper to lower course. | A causal chain connecting gradient, velocity, energy, and landform change. |
| Evaluate | To what extent is human intervention the primary cause of flooding in a river's lower course? Use geographical evidence to justify your answer. | A balanced argument weighing human versus physical factors, with a supported conclusion. |
This mapping means that when you plan a lesson using the Thinking Framework, you already know which Bloom's level the cognitive work sits at, and you can write your AI prompts to match. The AI for Teachers article includes a live AI Prompt Builder widget that generates prompts aligned to each of these operations across ten subjects and five year groups.
The significance of this alignment extends beyond lesson planning. Cognitive load theory tells us that learners can only process a limited amount of new information at once (Sweller, 1988). When your AI prompts are calibrated to the right Bloom's level for where your learners are in their learning, you reduce the risk of generating material that either under-challenges or overwhelms them.
Rosenshine (2012) found ten effective teaching principles in his research. Some directly link to Bloom's prompting, as described in this guide. Use it as a starting point for professional discussion: identify the learner's current need, record evidence from more than one lesson, and agree the next classroom adjustment with the SENCO or family.
Daily review helps learners remember past lessons. Use AI for quick "remember" recall questions, as Rosenshine (2012) suggests. You can create a week's worth in just minutes.
Principle 2 is presenting new material in small steps. Understand-level prompts help you generate clear, well-sequenced explanations of new concepts that you can annotate, adapt, or use as models. See Rosenshine's Principles for a fuller account of how each principle applies to lesson design.
| Thinking Framework Operation | Bloom's Level | Example AI Prompt Verb |
|---|---|---|
| Part-Whole | Understand | Identify the components of... |
| Sequence | Apply | Order the stages of... explaining how each leads to the next |
| Compare | Analyse | Compare and contrast... focusing on [criteria] |
| Classify | Analyse | Categorise these examples into groups, explaining your criteria |
| Cause and Effect | Analyse | Trace the chain of causes leading to... |
| Analogy | Understand / Analyse | Explain [concept] by analogy with something familiar to [year group] |
| Perspective | Evaluate | From the perspective of [stakeholder], evaluate the decision to... |
| Systems Thinking | Evaluate | Explain how [system] would respond to a change in [variable], including feedback effects |
Principle 6 is checking for learner understanding regularly. Apply and Analyse prompts generate the kind of practice problems and discussion questions that give you real-time evidence of whether learners have moved beyond surface knowledge. Combine these with the AI in lesson planning strategies to build a coherent sequence.
Pick one topic you are teaching this week. Write prompts at three different Bloom's levels using the templates above. Run all three prompts and compare the outputs side by side.
You will notice three things. First, the AI outputs differ substantially in depth and complexity. Second, the higher-level prompts produce content that is harder to generate yourself from scratch but that your learners genuinely need. Third, the outputs give you an immediate sense of which level your current lesson activities are actually sitting at.
If most of your current activities generate Remember-level AI outputs, your lessons may be spending too much time on recall and not enough on the thinking skills that build long-term understanding. That is not a failing; it is useful diagnostic information.
For whole-school use, move beyond one-off prompts. Leaders can map approved AI tools for teachers and agreed Bloom's taxonomy prompt types to curriculum keystones, subject assessment objectives and the school's AI policy. Sample learner work, model prompts and teacher review notes should show what changed in curriculum and teaching, the area inspected under Ofsted's current framework (Ofsted, 2025). Use this evidence for CPD and quality assurance, while keeping assessment validity and learner authorship explicit.
Open a free account and help organise learners' thinking with evidence-based graphic organisers. Reduce cognitive load and guide schema building dynamically.
Related reading: A Diamond 9 CPD lesson for ranking classroom AI uses
Bloom's taxonomy is not a validated model of how cognition unfolds. It began as a classification of educational objectives, yet classroom diagrams often present its categories as a fixed staircase. Case (2013) argues that this ordering can restrict access to demanding work because teachers may postpone analysis and evaluation until factual knowledge appears complete. The categories also overlap: explaining a complex idea may demand more expertise than evaluating a simple claim.
Verb choice is therefore a weak proxy for cognitive demand. Paul (1993) criticised Bloom's separation of recall from knowledge and rejected the taxonomy's claim to value neutrality. Webb's alignment work makes a related methodological point: depth depends on the reasoning, knowledge and time a task requires, not on its command verb alone (Webb, 1997). A prompt containing "analyse" may still produce shallow pattern matching.
The taxonomy also arose within a Western assessment tradition. Used without adaptation, it can privilege individual written performance over collective, relational and oral ways of knowing. Battiste's critique of Eurocentric education (Battiste, 2013) supports a practical response: invite community knowledge, oral explanation, collaborative judgement and locally meaningful criteria into both prompts and assessment.
Applying a human taxonomy to an AI system adds another limitation. Fluent Evaluate or Create output does not prove that a language model understands, reasons or creates as a learner does; Bender et al. (2021) warn against inferring understanding from plausible language. Teachers should assess the learner's decisions, evidence and revisions, not the apparent sophistication of the machine's prose. Despite these limits, Bloom's taxonomy retains value as a shared planning vocabulary when teachers use it flexibly, alongside disciplinary knowledge, cultural context and direct evidence of learning.
Authoritative guidance on AI in education: DfE guidance on generative AI in education, DfE support materials for using AI in education settings, Ofsted findings from AI early adopters in schools and FE, EEF Teacher Choices trial on ChatGPT in lesson preparation.
Bloom, B. (1956). Taxonomy of educational objectives.
Karpicke, J. (2008). The critical importance of retrieval for learning.
Rosenshine, B. (2012). Principles of instruction.
Sweller, J. (1988). Cognitive load during problem solving.
Webb, N. (1997). Criteria for alignment of expectations and assessments.
These five studies provide the evidence base for using Bloom's taxonomy to design higher-order AI prompts in classroom contexts.
Bloom's Taxonomy, revised by Anderson and Krathwohl (2001), helps plan lessons. It offers educators a framework for learning objectives. This aids in assessment design, say Anderson and Krathwohl (2001). Use it to support each learner's progress, note Anderson and Krathwohl (2001).
Anderson, L. W., & Krathwohl, D. R. (2001). Longman.
Anderson and Krathwohl (2001) revised Bloom's Taxonomy, switching nouns to verbs. They reordered the six levels, placing Create above Evaluate. This updated framework informs all six levels used here. It remains a standard reference for cognitive task sorting in research.
Bloom's taxonomy (1956) came from a 1948 conference. Bloom chaired a team creating a shared assessment system. It wasn't a value hierarchy; recall enables thinking. Anderson and Krathwohl (2001) revised it with verb categories. They added a Knowledge Dimension: Factual, Conceptual, Procedural, and Metacognitive. Use AI to target specific areas, not just single levels.
Cognitive Load Theory View study ↗
Sweller, 1988
Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257, 285.
Sweller (date not provided) showed instruction needs to consider working memory. Prompts above a learner's schema add extra load, hindering learning. Use prompts at the right Bloom's level to lower this risk.
Principles of Instruction: Research-Based Strategies That All Teachers Should Know View study ↗
Rosenshine, 2012
Rosenshine, B. (2012). Principles of instruction. American Educator, 36(1), 12, 19.
Rosenshine (2012) distilled classroom research into ten key principles. He stressed daily review, small steps, and checking understanding. These link to Bloom's (1956) Remember, Understand, Apply levels. This gives teachers a practical link for planning (Rosenshine, 2012; Bloom, 1956).
Depth of Knowledge View study ↗
Webb, 1997
Webb (1997) created criteria for maths and science. The research aligns expectations with assessments. This helps learners in both subjects. The CCSSO published Webb's research in Washington.
Webb's (2002) Depth of Knowledge helps alongside Bloom's (1956) taxonomy. It pinpoints if AI tasks need real extended thinking (DOK 3, 4). The frameworks together improve task difficulty for higher attaining learners.
Retrieval Practice and Test-Enhanced Learning View study ↗
Roediger & Karpicke, 2006
Roediger and Karpicke (2006) found that retrieval practice supports long-term retention more effectively than repeated study in delayed tests. Karpicke (2012) also frames active retrieval as a meaningful learning process. This supports Remember-level AI prompts when they are used for low-stakes recall before learners move towards analysis, evaluation and creation.
Karpicke, J. D., & Blunt, J. R. (2011). Retrieval practice produces more learning than elaborative studying with concept mapping. Science, 331(6018), 772-775.
For the wider picture, explore our AI and EdTech tools hub, our home for evidence-based AI guidance across policy, lesson planning, and classroom practice.
