The Unverifiable AI Source
An HL candidate’s comparative study has 12 screens and compares three works: Utagawa Hiroshige’s Plum Estate, Kameido (1857), a 2023 AI-generated image made with Midjourney that imitates ukiyo-e, and a 2019 photograph by a named contemporary artist. The AI image is cited as ‘anonymous digital artist, midjourney.com’. The candidate does not record the prompt, model version, date or who generated it. The candidate argues that the work’s function is ‘to question authorship’ and that its cultural significance is ‘the democratisation of art’. The teacher must decide whether to approve the selection before the file is locked. The candidate is unwilling to remove it because it is the only contemporary work that connects to the Hiroshige print. Deadline is five days away. The school requires accurate citation of all sources.
1. Is the AI-generated image academically defensible as one of the three comparative works in its current form? Explain the judgement.
Read the worked answer
Not in its current form. The comparative study depends on identifiable works and contextual evidence. ‘Anonymous digital artist’ is inaccurate when the candidate knows the platform and may know the prompt author; the platform is not the artist. Without the prompt, model version, date and author, the candidate cannot reliably interpret function and purpose or evaluate cultural significance. If the candidate generated the image, it is the candidate’s own work, not a third-party work for comparison, and should not occupy one of the three comparative slots. If someone else generated it, provenance is still unverified. An AI image can be studied as a cultural product, but the study must be transparent about its conditions of production. In five days, the safer judgement is to replace it with a documented work by a named artist; otherwise the file risks failing on research and interpretation rather than on the quality of the formal analysis.
2. If the candidate keeps the image, what must be cited and how should the analysis shift? If it is replaced, what kind of work should replace it?
Read the worked answer
If kept, cite the platform, model version, generation date, prompt author, prompt text, any post-processing, and any known source images. Label it clearly as AI-generated. Shift the analysis away from assumed artist intention and toward conditions of production: training data, mimicry of ukiyo-e conventions, platform labour, copyright, circulation and reception. Compare its function with Hiroshige’s print as a mass-produced commercial image, not only as an aesthetic object. That can be rigorous, but it still may not satisfy a comparison of works by artists if authorship remains unclear.
If replaced, choose a named contemporary artist whose work visibly engages ukiyo-e, image circulation or authorship, with verified title, date, medium and source. The replacement should let the candidate compare function and cultural significance, not just visual similarity. Three works by at least two named artists is stronger than three images with one unverifiable origin.
3. How should the candidate handle the claim that the work ‘questions authorship’ without overstating?
Read the worked answer
The candidate should separate four roles: the prompter, the platform/model, the training data and the viewer. Avoid saying ‘the AI intended’ or ‘the artist intended’ unless intention is documented. Instead write: the work’s production and circulation raise questions about authorship because the prompt, selection and platform are all authored in different ways. Support that with a cited source about AI image generation or copyright. Then connect it to the candidate’s own practice-for example, whether the candidate uses generated source material and how they credit it. The claim becomes defensible when it is specific about who did what and when it is supported by evidence, not when it is used as a slogan.