A.1: Case Study

An Interpretative Diagram Artefact

Individual assignment: 20% of the course grade
Deadline: 11 November 2026
Submission: One portrait graphic, AI dialogue and process record, generated images, and supporting text submitted through the Google Form.

The Brief

Produce one interpretative diagram based on a focused AI case study.

Think of the artefact as art mixed with a workflow diagram and annotation. It may explain a process, reveal a relationship, expose a tension, or present a critical interpretation.

You may research and interpret an existing documented case, such as an AI application, mechanism, dataset, incident, or controversy. You may also investigate a small case yourself. A new experiment is not required.

Keep the scope manageable: choose one case and one central concept that you can understand and communicate through a single portrait graphic. Your audience is bachelor students from different disciplines, so explain unfamiliar terms and preserve the essential meaning when simplifying.

Choose one of three themes:

The exercise moves through:

Research → Generate + Critique → Document → Submit

Suggested Workflow

The timings below are a guide, not a requirement. Save your dialogue and images as you work.

Stage Suggested time
1. Choose + Research — Select a theme and case, then investigate it using credible sources. 2 hours
2. Generate + Critique — Develop and refine your graphic through critical dialogue with AI. 2 hours
3. Document — Prepare your dialogue record and organise the image files. 2 hours
4. Submit — Upload your files and complete the Google Form. —
Total ~6 hours

1. CHOOSE + RESEARCH

Understand the case before asking AI to represent it.

Choose a specific example within your theme. For instance, investigate how a particular chatbot uses external documents, how responsibility arose in a documented deepfake incident, or how a dataset represents different communities.

Research the case independently using at least three credible sources. These may include course materials, academic publications, official documentation, institutional reports, or reputable journalism.

Identify:

Keep a reference list and use your sources to check the AI’s suggestions. Verify any references suggested by AI before citing them.

Theme-Specific Considerations

Use the relevant row to guide your case study. These factors inform Research and Conceptual Understanding in the common rubric; they do not carry additional marks. Focus on the considerations relevant to your chosen case.

Theme & Summary Example Cases Considerations Assessment Factors
Mechanics of AI: How an AI process works and how its design shapes behaviour. A neural network learning to recognise images; a chatbot finding information in documents; an image generator producing different results from one prompt. What goes in, what happens, and what comes out? Where does the mechanism sit within the system? What does it enable or limit? Accurate mechanism and information flow; connection between design and behaviour; explanation of capabilities and limitations.
Responsibility and AI: Where human choices, risks, and responsibilities arise. A documented deepfake incident; an authorship dispute over AI-generated art; incorrect AI advice that was published or acted upon. What did the AI contribute or assume? Where could people challenge or change the outcome? Who was affected, and who should take responsibility? Distinction between AI contribution and human judgement; risks linked to consequential decisions; reasoned interpretation of authorship or accountability.
Datasets and Risks: How data choices shape AI behaviour and its consequences. A dataset that underrepresents certain communities; images collected online without clear consent; AI performance that differs across languages. How was the data collected, labelled, or filtered? What is included, missing, or distorted? How might this affect outputs and people? Understanding of data quality and representation; evidence connecting data, behaviour, and risk; clear limits on inference; a plausible response, where relevant.
📋 What to Document

Sources: Keep a reference list of the sources used to research and verify your case, including at least two credible sources. Use one referencing style consistently and include links or DOIs where available. Enter this as text in the Google Form.

2. GENERATE + CRITIQUE

Bring your research into the dialogue and actively direct the AI.

Explain your case, intended message, and audience. Ask AI to propose visual components or ways of representing the concept, then evaluate its responses against your research and intentions.

⚠️ Keep to One AI Conversation Develop your graphic within the same conversation so that your research, critiques, image iterations, and decisions remain connected and easy to follow.

Introduce friction and constraints through specific critique:

Area What to check Example direction
Factual accuracy Are the labels, claims, and depicted events supported by your sources? “This label contradicts the source. Correct it and retain the distinction between these two stages.”
Relationships and structure Are the connections, sequences, and groupings correct? “This extra arrow implies a relationship that does not exist. Remove it and show the actual sequence.”
Visual meaning Do colour, scale, position, and line weight convey the intended meaning? “The thicker line makes this look more important. Give equivalent relationships equal emphasis.”
Clarity and coherence Can a non-specialist follow the image? Do its components work together? “These stages are difficult to distinguish. Separate them and use consistent labels.”
Metaphor and representation Does the chosen visual form clarify the concept or distort it? “This metaphor makes the system appear to think like a person. Propose a representation that avoids that implication.”
Following constraints Has the AI followed your instructions? What should stay or change? “Keep the agreed components and colour scheme. Change only the arrangement and do not add new elements.”

Explain why you accept, reject, or revise a suggestion. Use your sources to challenge inaccurate explanations, and your visual judgement to improve how the concept is communicated.

💬 Make Your Thinking Visible

• Question the representation.
Does the image actually explain your concept? Identify what is inaccurate, missing, unclear, or misleading.

• Give specific directions.
Explain what the next version should preserve, remove, or change—and why.

• Focus on meaningful progress.
There is no required number of prompts or images. We assess the quality of your critique and how your decisions improve the work.

Developing the Composition

You may generate and refine the whole graphic with AI, or build it from separate components. For example, you can generate components first, then ask AI to bring them together within the same conversation, specifying their arrangement and relationships.

You may also select, edit, annotate, and assemble AI-generated components yourself using PowerPoint, Photoshop, Illustrator, Canva, or another suitable tool. This can give you better control over relationships, arrows, labels, layout, and visual hierarchy. Manual composition is optional.

Whichever approach you choose, check the reading order, consistency, legibility, and accuracy of the final image. Retain any AI-assisted composition in your dialogue record and briefly note any editing completed outside it.

Your Final Graphic

Produce one graphic in a 4:5 portrait aspect ratio (width:height). The graphic itself is the final artefact.

📋 What to Document

Core Files
  1. Final graphic: Your completed image, saved as PNG or JPG.
  2. Process conversation log: The relevant AI dialogue, critiques, and image iterations, saved as PDF.
  3. Process images: The generated images saved individually and collected in a ZIP file.
Additional Supporting Text (to be entered in the Google Form)
  1. Title: A clear title for your graphic.
  2. Caption: Up to 100 words identifying the case, explaining the central idea, and stating the main takeaway.
  3. Software: The AI models or platforms and editing tools used, with a brief note on what you used each for.

Assessment Rubric

Criterion Weight What we are looking for
Research and Conceptual Understanding 20% Accurate understanding of the case, grounded in credible references, including its key relationships, implications, and limitations.
Critical Thinking and Reflection 20% Research-informed questioning of AI responses. Recognition of inaccuracies, assumptions, and weak representations, with reflection on your contribution and what remains unresolved.
Direction and Visual Deliberateness 20% Purposeful constraints and reasoned decisions about what to retain, reject, combine, or change. Deliberate control of composition, hierarchy, relationships, and visual language.
Progressive Development 20% Traceable evolution through meaningful iterations that respond to identified conceptual, factual, or visual problems.
Communication 20% The graphic, title, and caption convey a clear central idea to a non-specialist audience. The case is understandable without reading the process record.
🔎 A.1: What Are We Looking For?

Show us how your research informed your critique, how you directed the AI, and how your decisions shaped the final artefact. Visual polish alone does not demonstrate understanding.

Do your best. We will be understanding about minor formatting or organisational shortcomings and focus primarily on the substance of your work.

3. DOCUMENT

Here is a quick recap on your deliverables. Please replace UID and NAME in the filenames with your student number and name.

📝 Supporting Text
Enter the following directly in the Google Form:
  • Title: A clear title for your graphic.
  • Caption: Up to 100 words identifying the case, explaining the central idea, and stating the main takeaway.
  • Sources: At least three credible sources
  • Software: The AI models or platforms and editing tools used, with a brief note on what you used each for.
🖼️ Final Graphic
  • Format: One PNG or JPG.
  • Filename: A1_UID_NAME_graphic.png or A1_UID_NAME_graphic.jpg.
💬 Process Conversation Log
  • Content: Include your prompts, AI responses, critiques, directions, AI-assisted composition, and meaningful unsuccessful attempts. Briefly note any editing or assembly completed outside the conversation.
  • Filename: Consolidate into one PDF where possible, named A1_UID_NAME_log.pdf.
    (For multiple logs, number them chronologically: A1_UID_NAME_log01.pdf, A1_UID_NAME_log02.pdf, etc.)
  • Export: Use Print → Save as PDF or a conversation-export tool. Check that the full relevant dialogue is visible. Maximum 100 MB per PDF.
  • Missing images: Images may not appear in the exported PDF. This is fine: submit them separately as process images, named using the page containing the corresponding generation prompt or response.
🗂️ Process Images
  • Package: Download the generated images and collect them in A1_UID_NAME_images.zip.
  • Image filenames: Match each image to its page in the final conversation log.
    A1_UID_NAME_P10-01.png = first image associated with page 10.
    A1_UID_NAME_P10-02.png = second image associated with page 10.
    (For multiple logs, include the log number: A1_UID_NAME_log01_P10-01.png.)
  • Page reference: Use the page where the image first appears. If it is missing from the PDF, use the page containing its generation prompt or response.
  • File type: Keep the actual extension, such as .png, .jpg, or .webp.

4. SUBMIT

📤 Final Submission — 11 November 2026

Complete the [Google Form]:
  • Upload: Your final graphic, AI dialogue and process record, and ZIP of generated images.
  • Enter: Your abstract, title and caption, reference list, and software details.