Chapter 11 of 17
Curriculum Design, Assessment and Visualization with AI
Shows how teachers can use AI for curriculum design, Bloom-aligned outcomes, question papers, rubrics, feedback and mind maps. It also examines academic integrity and, with numbers, why AI-detection tools cannot be trusted on their own.
In this chapter
11.1 Curriculum design with AI
Curriculum design starts from what learners should be able to do, then plans content, activities and assessment to get them there. In India, outcome-based education (OBE) is required for programmes accredited by the National Board of Accreditation (NBA).
Where AI helps
| Design step | Example prompt |
|---|---|
| Course outcomes | "Write five course outcomes for a 40-hour course on Data Visualisation for BBA students. Use Bloom's action verbs and mark each level K1–K6." |
| Syllabus outline | "Divide these outcomes into five units of eight lectures each, with topics per unit." |
| Lesson plan | "Prepare a 50-minute lesson plan on bar and line charts: objective, warm-up, explanation, activity, assessment, homework." |
| CO–PO mapping | "Map these course outcomes to programme outcomes PO1–PO12 with strength 1, 2 or 3, and justify each." |
| Activity design | "Suggest three active-learning activities for teaching supply and demand to 60 students." |
| Reading list | "Suggest textbooks on this topic." (Every title must be checked; AI may invent books.) |
Bloom's revised taxonomy (Anderson and Krathwohl, 2001) is the standard framework for writing outcomes:
| Level | Code | Sample action verbs |
|---|---|---|
| Remember | K1 | Define, list, recall |
| Understand | K2 | Explain, summarise, classify |
| Apply | K3 | Use, solve, demonstrate |
| Analyze | K4 | Compare, differentiate, examine |
| Evaluate | K5 | Judge, justify, critique |
| Create | K6 | Design, construct, formulate |
This course itself uses K2 to K5, as listed in the syllabus.
11.2 AI in assessment
Generating questions
Create 10 multiple-choice questions on Unit II (Prompt Engineering) of this syllabus.
Mix of levels: 3 at K2, 4 at K3, 3 at K4.
Each question: 4 options, one correct, plausible distractors, the answer, and a one-line explanation.
Avoid "all of the above".
Other assessment tasks
- Question papers with a blueprint: marks per unit and per Bloom level.
- Case-based and scenario questions for higher-order thinking.
- Rubrics for assignments and projects.
- Model answers and marking schemes.
- Viva questions at increasing difficulty.
Example rubric generated with AI (teacher-reviewed)
| Criterion | Excellent (4) | Good (3) | Fair (2) | Poor (1) |
|---|---|---|---|---|
| Prompt design | Role, task, context, constraints and format all clear | Four of five present | Two or three present | Vague, single-line |
| Output evaluation | Checks accuracy against sources, notes errors | Checks most claims | Some checking | No checking |
| Workflow design | Complete, logical, with human checkpoints | Mostly complete | Gaps in logic | Incomplete |
| Presentation | Clear, well-organised, cited | Minor issues | Several issues | Disorganised |
Teacher's checks on AI-made questions: is the marked answer really correct; is there only one correct option; is the Bloom level genuine; is the language clear; is it inside the syllabus?
11.3 AI for feedback
Good feedback is specific, tells the learner what was done well, what to improve, and how. AI can draft such feedback quickly, but the teacher's judgement of the work comes first.
Example prompt
You are a supportive English teacher. Here is a student's 200-word paragraph and the rubric.
Give feedback in this format:
1. Two specific strengths, quoting the student's words.
2. Two specific improvements, with a rewritten example sentence for each.
3. One next step for the next assignment.
Do not give a mark. Keep it under 150 words and encouraging.
Privacy: remove student names and roll numbers before pasting work into a public tool, or use an institution-approved platform.
11.4 Visualisation: mind maps, concept maps and infographics
Visuals help learners see how ideas connect. Generative AI can produce the structure, and tools can draw it.
| Visual | Purpose | How to make it with AI |
|---|---|---|
| Mind map | Central topic with branches of sub-topics | Ask for a hierarchy in Markdown or Mermaid code, then render it in a mind-map tool |
| Concept map | Concepts linked by labelled relationships ("causes", "is part of") | Ask for concept–relationship–concept triples, then draw |
| Flowchart | Steps and decisions in a process | Ask for Mermaid flowchart code |
| Infographic | Key facts and numbers in a visual layout | Draft text and data with AI; design in Canva or similar |
| Timeline | Events in order | Ask for a dated list; place on a timeline template |
Example: a mind map in Mermaid code
Prompt: "Create a mind map in Mermaid syntax for 'Generative AI' with four branches: Models, Platforms, Applications, Risks, each with three sub-points."
mindmap
root((Generative AI))
Models
GAN
Transformer
Diffusion
Platforms
ChatGPT
Gemini
Copilot
Applications
Office
Education
Healthcare
Risks
Hallucination
Bias
Privacy
This code renders as a mind map in tools that support Mermaid, such as the Mermaid Live Editor, many Markdown editors and some note-taking apps.
Tools often used: Napkin AI (turns text into diagrams), Canva's AI features (infographics and presentations), Whimsical and Miro (mind maps and boards with AI), and Google's NotebookLM, which can generate mind maps from uploaded sources.
11.5 Academic integrity
Generative AI makes it easy to produce essays and code, which challenges traditional assignments.
Approaches institutions are taking
- Clear policy: state where AI use is allowed, required to be declared, or banned.
- Disclosure: students state what AI tool they used and how.
- Assessment redesign: in-class writing, oral viva, personal reflections, local data, drafts with version history, and process-based marking.
- Teaching AI literacy: treat AI as a tool to be used well, with checking and citation.
Limits of AI detectors. In January 2023 OpenAI released a classifier to detect AI-written text. It withdrew the tool in July 2023 because of its low accuracy: in OpenAI's own tests it correctly identified only 26% of AI-written text and wrongly labelled 9% of human-written text as AI. Detection scores alone should never be the basis for a misconduct finding.
11.6 Worked numericals
Numerical 11.1: Item analysis of an AI-generated MCQ.
An AI-generated question is given to 60 students. The top 27% (16 students) and bottom 27% (16 students) by total score are compared. In the upper group 14 answer correctly; in the lower group 6.
P = 0.625 is moderate difficulty. D = 0.50 is excellent; the usual guideline treats D ≥ 0.40 as very good and D < 0.20 as poor. So this AI-made item can be kept. An item with D near zero or negative should be rewritten, because strong and weak students do equally well or the weak do better.
Numerical 11.2: Why AI detectors mislead (using OpenAI's published rates).
A class submits 200 essays; in reality 40 are AI-written and 160 are the students' own. A detector catches 26% of AI text and falsely flags 9% of human text.
More than half of the flagged essays (about 58%) would be honest students wrongly accused, and about 30 of the 40 AI-written essays would go unflagged.
Numerical 11.3: Building a question paper blueprint.
A 70-mark paper must follow the weightage K2 20%, K3 30%, K4 30%, K5 20%.
The teacher can prompt the AI with this blueprint so that it generates the right number of questions at each level.
11.7 Industry use cases
- Lesson planning in Indian government schools. The Shiksha copilot pilot in Karnataka (Chapter 10) generated lesson plans, activities and assessments aligned to the state curriculum in English and Kannada.
- AI tutors and teacher assistants. Khan Academy's Khanmigo helps teachers draft lesson plans, rubrics and exit tickets, and guides students with questions instead of answers.
- AI-assisted grading. Platforms such as Gradescope group similar student answers so a teacher can grade a whole group at once, which speeds up consistent marking while keeping the teacher in control.
11.8 Review questions
- Write three course outcomes for a course on Business Communication at K2, K3 and K4 levels, using Bloom's verbs.
- Analyse the checks a teacher must apply to AI-generated MCQs.
- Prepare a prompt for AI-generated feedback on a lab report. Why must the teacher review it?
- Create a prompt that produces Mermaid code for a mind map on "Prompt Engineering".
- An item is answered correctly by 10 of 20 students in the upper group and 12 of 20 in the lower group. Compute P and D and decide whether to keep it. (Answer: P = 0.55, D = −0.10; reject or rewrite.)
- Why should AI-detection scores not be used alone to judge misconduct? Support your answer with numbers.
Knowledge sources
- Anderson, L. W. and Krathwohl, D. R. (eds.), A Taxonomy for Learning, Teaching, and Assessing: A Revision of Bloom's Taxonomy of Educational Objectives, Longman, 2001.
- National Board of Accreditation (India), Outcome Based Education resources.
- OpenAI, New AI classifier for indicating AI-written text, January 2023, updated 20 July 2023.
- UNESCO, Guidance for Generative AI in Education and Research, 2023.
- Ebel, R. L. and Frisbie, D. A., Essentials of Educational Measurement, 5th ed., Prentice Hall, 1991 (item analysis guidelines).
- Mermaid, Mindmap syntax documentation.
Independent study notes for learning purposes. Not affiliated with any university or course provider. Facts and case-study figures are linked to their original sources. ChatGPT, Gemini, Copilot, Jasper, Synthesia, Runway, Lumen5 and other product names are trademarks of their respective owners.