Week 20: AI Literacy Capstone, Portfolio, and Action Plan

 

Welcome to Week 20 of the AI Literacy Course.

Imagine that a learner presents a polished AI-assisted guide. It looks professional, but the learner cannot explain:

• where its claims came from;

• whether the sources are reliable;

• what information was entered into the AI tool;

• whether images may be reused;

• what the learner changed;

• who checked the final result.

A polished product does not automatically demonstrate AI literacy.

Another learner may investigate the same task and decide not to use AI because the available tool is unapproved, the information is sensitive, or the output cannot be verified. That may demonstrate excellent judgment.

The central rule is:

AI literacy is shown by the quality of decisions, not the frequency of AI use.

Learning goals

By the end of this lesson, you should be able to:

• define a useful and manageable capstone project;

• decide whether and how AI should be used;

• select relevant methods from the course;

• demonstrate evidence, verification, privacy, rights, and human judgment;

• document one meaningful revision;

• give and respond to safe peer feedback;

• create a realistic three-month action plan.

1. AI literacy is a continuing practice

Completing this course does not mean that you have mastered a fixed technology.

AI tools, laws, licences, workplace rules, and social practices will continue to change.

Durable AI literacy includes:

• understanding what a system can and cannot do;

• defining a task before choosing a tool;

• asking good questions;

• checking decisive claims;

• protecting people and data;

• recognising bias and exclusion;

• respecting sources and rights;

• documenting important decisions;

• asking for help;

• knowing when to stop.

A responsible decision may be to:

• use AI;

• use AI with clear limits;

• choose a simpler tool;

• ask a person for support;

• delay the task;

• decline AI entirely.

2. Begin with one capstone question

Your project should answer:

Should and how should AI be used for this task?

Choose a useful task connected to:

• study;

• work;

• creativity;

• everyday life;

• community participation.

Examples include:

• creating a study guide;

• improving an information resource;

• comparing ways to complete a workplace task;

• designing an accessible presentation;

• checking a public claim;

• creating a rights-aware media product;

• planning a fictional workplace AI pilot;

• designing a fictional public consultation;

• explaining why AI should not be used for a particular task.

Keep the project small enough to complete, explain, and revise.

3. Define the purpose and boundaries

Before selecting a tool, describe:

• the purpose;

• intended audience;

• desired outcome;

• people affected;

• level of risk;

• time and resource limits;

• what is included;

• what is excluded.

Also define your boundaries:

• What information will you not enter?

• Which decisions must remain human?

• What will require verification?

• What kind of output would be unacceptable?

• When will you pause or stop?

A clear boundary protects both the project and the people involved.

4. Compare AI with other options

Do not assume that AI is necessary.

Compare at least two options:

• completing the task without AI;

• using an ordinary search engine;

• using a template;

• using simpler software;

• asking a teacher, colleague, librarian, or specialist;

• using AI for one bounded part;

• using AI for several parts with review.

Ask:

• Which option best serves the purpose?

• Which produces the least unnecessary risk?

• Which can I explain and verify?

• Which protects people and data?

• Which is accessible?

• Which is allowed in this context?

• Which creates the least hidden checking work?

Choose AI because it is suitable—not because the course is about AI.

5. Select the course tools you need

You do not need to use every framework.

Select the methods that fit your project.

Prompt and dialogue methods

Use these when interacting with a generative AI system:

• prompting fundamentals;

• RUKF: Role, Task, Context, and Format;

• follow-up questions;

• iterative improvement;

• Socratic dialogue.

Evidence methods

Use these when factual claims matter:

• calibrated trust;

• hallucination checks;

• SIFT source investigation;

• PAUSE for suspicious or manipulated media.

People, data, and fairness methods

Use these when information or decisions affect people:

• A–P–K for personal data;

• FAIR for possible representation patterns;

• meaningful human review;

• accessibility and contestability checks.

Creation methods

Use these when creating or publishing material:

• the Lesson 13 production workflow;

• RIGHT for sources, licences, permissions, disclosure, and traceability.

Organisational and civic methods

Use these when evaluating wider systems:

• WORK for workplace AI;

• VOICE for civic participation and public-sector accountability.

Explain which methods you selected and why. You may also state why another framework was not relevant.

6. Use COMPASS to guide the whole project

COMPASS brings the course together. It does not replace the specialist methods.

C — Context

Ask:

• What is the task?

• What human need does it address?

• Who is the audience?

• Who may be affected?

• What are the stakes and limits?

O — Options and outcome

Ask:

• What result do I want?

• Is AI necessary?

• What non-AI options exist?

• Should AI be used for the entire task or only one part?

• What would a successful outcome look like?

M — Model and method

Ask:

• What does the system do?

• Does it generate, search, classify, recommend, predict, or combine several functions?

• What sources or data influence its output?

• What limitations follow from the method?

• Which course frameworks should I use?

These functions may overlap. A search tool can include generated summaries, and a recommendation system may use machine learning. Focus on how the system works in your task rather than forcing it into one category.

P — People, privacy, and permissions

Ask:

• Who could benefit or be harmed?

• Does the task involve personal or confidential information?

• Could someone be identified indirectly?

• Is the tool approved?

• Are sources, images, music, code, or other materials permitted?

• Are accessibility and representation considered?

Do not assume that removing a name makes information anonymous. Use fictional or synthetic cases where possible.

A — Assess evidence and impact

Ask:

• Which claims are decisive?

• Which errors could cause harm?

• What evidence supports the conclusion?

• Have I traced important claims to suitable sources?

• Could the output reproduce bias or stereotypes?

• What remains uncertain?

• Does the result meet the original purpose?

Do not check every sentence equally. Give the strongest verification to the claims that matter most.

S — Supervision and stop

Ask:

• Who reviews the work?

• Does that person have enough knowledge, time, evidence, and authority?

• Who makes the final decision?

• What can be corrected?

• What finding would pause or stop the project?

• Is there a safe alternative?

Human oversight means more than clicking “approve.”

S — Show the process and next step

Record:

• why you selected or rejected AI;

• relevant prompts and changes;

• decisive sources;

• data and rights checks;

• corrections;

• uncertainty;

• AI disclosure;

• final responsibility;

• what you learned;

• what you will do next.

Keep enough evidence to explain the process without storing unnecessary personal or sensitive information.

7. Verify what matters most

Verification should match the stakes.

Lower-stakes example

AI suggests possible headings for a fictional study guide.

A quick relevance and quality check may be enough.

Higher-stakes example

AI produces information about:

• health;

• legal rights;

• employment;

• public services;

• safety;

• a real person;

• a public claim.

This requires stronger evidence, appropriate expertise, and possibly a decision not to rely on AI.

Identify three to five decisive claims or evidence items.

For each one, record:

• the claim;

• why it matters;

• original or authoritative source;

• date;

• supporting or conflicting evidence;

• conclusion;

• remaining uncertainty.

Several websites repeating the same statement do not automatically provide independent confirmation.

8. Protect people, data, and rights

Before entering, uploading, or publishing anything, check:

• personal data;

• sensitive information;

• confidential material;

• workplace or school information;

• copyrighted material;

• licences and permissions;

• faces, voices, or likenesses;

• trademarks or brands;

• third-party information.

Apply A–P–K and RIGHT where relevant.

Use fictional information instead of real cases whenever possible.

If the project cannot be completed safely with the available tools and permissions, redesign it or do not use AI.

That is a valid capstone result.

9. Show your human contribution

Human contribution is not measured by the number of prompts.

Show how you contributed through:

• defining the purpose;

• selecting methods;

• researching;

• comparing sources;

• interpreting evidence;

• rejecting weak outputs;

• correcting errors;

• arranging material;

• writing or revising;

• considering affected people;

• making the final decision.

Be able to explain:

• what AI contributed;

• what you contributed;

• what changed through review;

• what you would not delegate.

10. Keep a selective decision log

A process log should make your important decisions visible.

It may include:

Stage	Decision or action	Evidence	Risk or uncertainty	Result
Purpose	Defined a study guide for adult learners	Learner needs	Language level	Limited guide to five topics
Tool choice	Used AI only for draft structure	AI/non-AI comparison	Generic output	Human-designed final structure
Verification	Checked three decisive claims	Original sources	One date unclear	Removed unsupported claim
Data	Used a fictional example	A–P–K	No real learner data	Approved to continue
Rights	Used one CC BY image	RIGHT log	Attribution required	Added TASL credit
Review	Peer identified unclear wording	Feedback notes	Possible confusion	Rewrote explanation

Do not include every experimental prompt if it did not influence the final decision.

Do not store personal, confidential, or harmful information merely to make the log look detailed.

11. Include one before-and-after example

Your portfolio must show one meaningful revision.

Possible examples include:

• an unsafe prompt and a safer fictional version;

• an initial answer and a corrected answer;

• a weak source and a better source;

• a vague prompt and an improved RUKF prompt;

• an inaccessible design and an accessible revision;

• a stereotyped output and a revised process;

• an unclear licence and a replacement component;

• an AI plan and a reasoned non-AI alternative.

Explain:

1. What was wrong or incomplete?

2. What evidence or feedback revealed the problem?

3. What did you change?

4. Why is the new version better?

5. What uncertainty remains?

12. Build a portfolio that shows judgment

Your capstone portfolio should contain:

1. Project question and scope

2. Artifact or reasoned decision not to use AI

3. Selected course frameworks

4. Selective decision log

5. Three to five decisive sources or evidence items

6. People, data, rights, and risk check

7. One before-and-after revision

8. Reflection

9. Three-month action plan

The artifact may be:

• a guide;

• a plan;

• an analysis;

• a presentation;

• an information resource;

• a media product;

• a prototype;

• a documented decision not to use AI.

A responsible refusal or redesign can receive the highest assessment when it is well supported.

13. Reflect in an accessible format

Choose one reflection format:

• approximately 400–600 written words;

• a four-to-six-minute audio response;

• a short presentation with explanatory notes;

• a structured discussion with your teacher.

Address:

• What was the purpose?

• Why did you use, limit, or reject AI?

• Which course tools did you select?

• Which evidence changed your thinking?

• What risk required the most attention?

• What did you revise?

• What remains uncertain?

• What would you do differently next time?

• Who is responsible for the final result?

Assessment should focus on judgment and evidence—not advanced English, presentation polish, or access to a paid AI tool.

14. Give and receive safe peer feedback

Share only material that is safe and permitted to share.

Do not share:

• personal data;

• confidential work or school documents;

• private account information;

• protected material that cannot be redistributed;

• harmful media;

• sensitive reflections.

Use a redacted project summary when necessary.

Give feedback in three parts:

1. Strength:
   “One strong decision supported by evidence is…”

2. Question:
   “One question I still have is…”

3. Next step:
   “One realistic improvement could be…”

After receiving feedback:

• make one meaningful revision; or

• explain why you did not adopt the suggestion.

Responsible judgment does not mean accepting every recommendation.

15. Use the portfolio rubric

Assess each criterion using evidence from the portfolio.

Criterion	Evidence expected
Purpose and scope	Clear need, audience, stakes, boundaries, and non-AI option
Tool and method choice	Reasoned choice of AI, limited AI, another tool, or no AI
Evidence and verification	Decisive claims traced to suitable evidence; uncertainty stated
People, data, and rights	Privacy, bias, accessibility, licences, and permissions considered where relevant
Human judgment	Meaningful selection, correction, supervision, and final responsibility
Documentation and disclosure	Selective process trail, AI use, sources, changes, and limitations visible
Revision and reflection	Concrete before-and-after evidence and realistic learning
Action plan	Specific practice, boundary, support, measure, trigger, and review date

Use four performance levels:

• 4 — Independent and well-supported

• 3 — Responsible with minor gaps

• 2 — Partly supported; important revision needed

• 1 — Insufficient evidence or unresolved risk

For each criterion:

1. select a level;

2. point to evidence;

3. name one possible improvement.

16. Create a three-month action plan

A useful action plan is specific enough to practise and review.

Include:

One goal

Example:

“Improve my ability to verify AI-generated study information.”

One context

Example:

“My weekly adult-education assignments.”

One routine

Example:

“For each assignment, identify the two most important factual claims and trace them to suitable sources.”

One boundary

Example:

“I will not enter information that identifies classmates or colleagues.”

One support person or institution

Examples include:

• teacher;

• colleague;

• librarian;

• manager;

• data-protection contact;

• union representative;

• specialist organisation;

• public authority.

One measure

Example:

“I will keep a short verification log for four assignments.”

One review date

Set a date approximately three months from now.

Early-review triggers

Review the plan sooner if:

• the tool changes significantly;

• a new school or workplace policy appears;

• the intended use changes;

• repeated errors occur;

• a privacy or security incident happens;

• new evidence changes your judgment;

• a licence or legal rule changes.

Capstone assignment: My responsible AI project

Create or improve one useful artifact—or make a reasoned decision not to use AI.

Submit

1. Project question and scope

2. Artifact or non-use decision

3. Selected course frameworks

4. Selective decision log

5. Three to five decisive sources or evidence items

6. People, data, rights, and risk check

7. One before-and-after example

8. Reflection in your chosen format

9. Three-month action plan

AI-use guidance

AI may support bounded parts of the project if its use is allowed and documented.

Do not enter:

• sensitive personal data;

• confidential material;

• identifiable third-party information;

• protected material you are not permitted to upload.

Your final judgment, verification, reflection, and explanation must be your own.

Final reflection

Complete these sentences:

• One responsible practice I will continue is…

• One boundary I will not cross is…

• One sign that I should pause is…

• One person or institution I can ask for help is…

• One decision I now feel more prepared to make is…

Responsible AI literacy includes being able to say:

• “I do not know.”

• “I need a better source.”

• “I need permission.”

• “This is outside my role.”

• “I need help.”

• “We should stop.”

Key vocabulary

Capstone:
A final project that brings together knowledge, skills, judgment, and reflection.

Portfolio:
A selected collection showing both the result and the process behind it.

Decision log:
A concise record of important choices, evidence, checks, revisions, and responsibility.

Decisive claim:
A claim that strongly affects the conclusion or could cause harm if wrong.

Human oversight:
Meaningful human ability to understand, question, correct, pause, or stop an AI-supported process.

Boundary:
A limit defining what information, task, decision, or risk will not be delegated or accepted.

Review trigger:
An event requiring an action plan or decision to be reconsidered earlier than scheduled.

COMPASS:
Context; Options and outcome; Model and method; People, privacy, and permissions; Assess evidence and impact; Supervision and stop; Show the process and next step.

Summary

In Week 20, we brought the course together through a responsible AI capstone.

Begin with a human purpose and a bounded question. Compare AI with other options. Select only the course methods relevant to the task.

Use COMPASS:

• Context;

• Options and outcome;

• Model and method;

• People, privacy, and permissions;

• Assess evidence and impact;

• Supervision and stop;

• Show the process and next step.

Your portfolio should show:

• why you made each important decision;

• what evidence you checked;

• how you protected people and rights;

• what you changed;

• what remains uncertain;

• what you will practise next.

A polished artifact is not the only successful outcome. A well-supported decision to limit, redesign, or reject AI can demonstrate excellent literacy.

AI literacy is shown by the quality of decisions, not the frequency of AI use.

Lesson 20 Interactive Quiz: AI Literacy Capstone and Action Plan

Choose one answer for each question. Then select Check my answers. This practice quiz does not collect names or scores.

1. A learner begins a capstone by asking, “Which AI tool should I use?” What should come first?

A. Selecting the tool with the largest number of features B. Buying a paid subscription before defining the task C. Defining the purpose, people, stakes, boundaries, and non-AI options D. Writing the longest possible prompt

2. A project requires confidential workplace records, but no approved AI tool is available. What is the strongest decision?

A. Use another method and document why AI was rejected B. Upload the records to a personal account and delete them later C. Remove names but retain every distinctive workplace detail D. Use AI because every capstone must contain generated output

3. A learner uses AI to create a sourced guide containing a Creative Commons image but no real personal data. Which framework selection is best?

A. Use every course framework whether relevant or not B. Use only RUKF because a good prompt guarantees reliable sources C. Use VOICE only because every guide is a public consultation D. Use RUKF, SIFT, and RIGHT, while explaining why A–P–K requires only a brief check

4. A capstone contains twenty factual statements. Which claims need the strongest verification?

A. Only the shortest statements B. Claims that affect the conclusion or could cause harm if wrong C. Only claims written without AI assistance D. None, if the final text sounds confident

5. What belongs in a selective decision log?

A. Every experimental prompt, including those unrelated to the result B. Personal information added only to make the log more detailed C. Important choices, decisive evidence, checks, revisions, uncertainty, and responsibility D. Only the name of the AI tool

6. Which example best demonstrates meaningful revision?

A. Replacing an unsupported claim after tracing it to an unreliable source and documenting the correction B. Changing the heading colour without explaining why C. Generating the same answer again and selecting the longer version D. Removing the process log so the final product appears more polished

7. A capstone is based on confidential workplace material. How should peer feedback be handled?

A. Share the complete confidential file because the class is private B. Upload the file to a public AI tool for automatic peer feedback C. Skip all reflection because the original cannot be shared D. Share a fictionalised or redacted summary and protect restricted information

8. A learner asks a classmate to approve an AI-generated health guide, but the classmate has no sources or relevant knowledge. Is this meaningful oversight?

A. Yes, because any human approval is sufficient B. No, because the reviewer lacks the evidence and competence needed to challenge it C. Yes, if the guide uses professional-looking formatting D. Yes, if the AI produced the same answer twice

9. Which sequence correctly represents the revised COMPASS framework?

A. Copy, Output, Memorise, Prompt, Automate, Save, Submit B. Context, Ownership, Marketing, Popularity, Appearance, Speed, Score C. Context; Options and outcome; Model and method; People, privacy, and permissions; Assess evidence and impact; Supervision and stop; Show the process and next step D. Choose a tool, generate an answer, and publish it

10. Which three-month action plan is most useful?

A. “For four assignments, I will verify two decisive claims, avoid identifiable peer data, ask the librarian when uncertain, and review the log on 15 October or after a policy change.” B. “I will try to use AI more often.” C. “I will trust better tools when they become available.” D. “I will make no boundaries because future tasks are unpredictable.”

Final reflection: Name one practice you will continue, one boundary you will keep, and one person or institution you can ask for help.