Generative AI, Authorship, and Academic Integrity in Digital Photography
You teach a high school digital photography class of 22 students. The assignment asks for a series of three images exploring 'a sense of place,' with evidence of planning, shooting, and editing. One student, Priya, submits a strikingly polished image with impossible sunset lighting, a building removed from the skyline, and no camera metadata. When asked, she says she used a generative AI tool to 'fix the sky' and to 'clean up' the composition, and that the building removal was 'just editing.' Your school's acceptable use policy allows AI for brainstorming and feedback but prohibits AI-generated imagery as the final submitted artwork unless the assignment explicitly permits it. The class has just begun a unit on copyright, fair use, and image licensing. You must respond to the integrity issue, assess the work, and use the moment instructionally without turning the class into a punishment session.
1. How would you distinguish AI-assisted editing from AI-generated imagery in this case, and what evidence would you request?
Read the worked answer
The distinction should be based on authorship of the meaningful visual content, not on whether a tool was used. Removing a building, replacing the sky with a generated sunset, and producing a final image with no camera metadata goes beyond retouching into synthesis. If the sky and the building removal were generated from a model rather than from Priya's own captures, the final image is substantially AI-generated even if the original photograph is underneath. If she used a healing brush on her own file, that is ordinary digital editing and should be judged by the assignment's editing limits.
The evidence request should be concrete and routine: the original RAW or camera file, the contact sheet or shoot log, the editing history or layers, and a short process statement. I would ask for the prompts used and the source images. I would not demand a confession or search her devices. The absence of metadata is not proof by itself, because metadata can be stripped by export, but it is a reason to ask for the original.
I would also check the assignment sheet. If it did not state an AI policy, that is a curriculum gap I need to fix, not solely a student failure. The policy should define permitted uses, disclosure, and consequences.
2. What instructional response would you make to Priya and to the class, and how would you grade the submitted work?
Read the worked answer
I would separate the integrity conversation from the art critique. With Priya, I would state the policy, ask for the process evidence, and explain the difference between editing a photograph and generating content. The consequence should be proportionate and restorative: a redo of the assignment using her own captures, with a required process portfolio and disclosure statement. I would not fail her for the semester over one assignment unless the policy specifies that penalty.
For grading, I would not assign a high grade to an image whose central visual decisions were generated by a model and presented as her photography. That would misrepresent the learning target. I would record the original submission as incomplete or non-compliant, then grade the resubmission against the same standards. If she discloses which parts were AI-assisted and reworks the image from her own sources, the new work can be assessed fairly.
For the class, I would use the case anonymously. I would teach the four fair-use factors, transformative use, licensing, model releases, and disclosure norms. I would have students annotate a famous photograph and an AI-generated image, asking: who authored the composition, the lighting, the subject, and the final file? The point is not that AI is forbidden; it is that authorship must be truthful.
3. How would you redesign the assignment and assessment to make authorship visible and reduce future ambiguity?
Read the worked answer
The assignment needs process artifacts that are hard to generate after the fact. I would require a shoot plan with location scouting, a minimum number of original captures, a contact sheet with selected frames, an editing log, and a short artist statement that explains at least two decisions and one rejected alternative. Students would submit the original file alongside the final export. I would also require a disclosure line: 'Tools used and what they did.' That normalizes transparency rather than treating AI as contraband.
I would revise the rubric to weight concept, photographic seeing, and editing decisions, not just polish. A strikingly polished but generated image should not outscore a less polished but authentically seen series. I would include a criterion for ethical and transparent use of sources.
Finally, I would align the assignment with the copyright unit. Students could compare a documentary photograph, a licensed stock image, and an AI-generated image, then write about transformation and credit. That makes the policy educative rather than punitive and gives me evidence if a future case is disputed.