If you find this resource useful, or are planning to draw from it in your own teaching, please consider filling out this brief survey to help me better understand how this resource interacts with your teaching: SURVEY LINK
Critical & Creative Machine Learning Teaching Activity Prompt
Joy Buolamwini
Refik Anadol
“One of the weird things about all this machine learning stuff is ... they've done this weird thing with learning where there's not really teaching right? It's like the learning just kind of happens with all this data and all these algorithms and ... they don't ask ‘Oh, wait, who's doing the teaching?’"
Jason Edward Lewis, Hawaiian/Samoan Digital Artist and Scholar
Inspired by Lewis’s comment above, we will explore and discuss contemporary concerns around AI, art, and teaching, and then investigate how we can train our own models on data sets that we curate.
This project can be done in groups of 1-4. Because this is a new and experimental (and somewhat technical!) process, working with a partner/team will be helpful.
Your task will be to:
- Curate a training set of 25-30 images and captions
- Write a creative and ethical curatorial statement (250-500 words) explaining your rationale for the images you have chosen
- In curating your training set, consider the various ethical concerns we have discussed in class with machine learning, including (but not limited to):
- Use of intellectual property from creators without compensation or consent
- Learned racial/gender/other biases
- Issues with accountability/credit of artwork
- Consider how your role as curator or “machine teacher” acknowledges, resists, or poses questions about these norms
- You will then train a custom Stable Diffusion model using that data, and generate/curate a set of 10+ images.
- We will exhibit the source and generated images together, alongside your statements.
- You will also write a summative reflective artist’s statement (250-500 words) exploring the relationship(s) you observe between your source and generated images.
Your final submission will include:
A Google Document containing:- A link to a folder with your source images
- A link to a folder with your generated images
- Your creative and ethical curatorial statement (250-500 words)
- Your summative reflective artist statement (250-500 words)
Creative and Ethical Curatorial Statement Prompt:
- Your statement should clearly communicate how you considered ethical issues (e.g. race, gender, intellectual property, etc.) in your choice of training data.
- Consider how your role as curator or “machine teacher” acknowledges, resists, or poses questions about these norms
- E.g. Are you creating a set with ethically-sourced images (by whatever ethical criteria you decide are valid)? Are you creating a set to introduce a perspective lacking in the default dataset? Or something else?
- Your statement should also clearly communicate your creative intentions in your choice of training data. What are you teaching your machine to create, and why?
Reflective Statement Prompt:
- Your statement should thoughtfully analyze the relationship between your source images and generated images, considering:
- Did the machine ‘learn’ what you intended to teach it? Justify your response with reference to the images.
- Did the machine ‘hallucinate’ things that don’t seem connected to your training data?
- How might you revise your curation of your training data were you to revisit this activity?
- How does training your own GenAI system impact your feelings or understandings about GenAI?
Assessment
| Excellent | Satisfactory | Needs Attention | Unacceptable | |
|---|---|---|---|---|
| Creative + Ethical Curatorial Statement 30 pts |
Statement clearly communicates how the artist considered ethical issues (e.g. race, gender, intellectual property) in their choice of training data. Statement clearly communicates artist’s creative intentions in their choice of training data. What are they teaching their machine to create, and why? Statement is clearly-written, free of grammatical errors, and is within the word count range. 24.1 to 30 pts |
Statement acknowledges ethical issues that were considered in curation of source images. Statement expresses some creative intention in selection of images, though it may not be totally clear. Statement may have some formatting, grammar, or other readability issues. 18.1 to 24 pts |
Statement does not clearly describe ethical or creative intentions in curation of images for machine teaching. Statement may have serious issues that impact readability. 0.1 to 18 pts |
No submission. 0 pts |
| Image Generation + Curation 30 pts |
25-30 source images, and writes accompanying captions. Student trains a generative model using those images. Student generates at least 10 images. 24.1 to 30 pts |
Student successfully trains a generative model, but the amount of training images and/or output images may be below the expected amount. 18.1 to 24 pts |
Student has less than half the required amount of source images or output images. 0.1 to 18 pts |
No submission. 0 pts |
| Summative Reflective Artist Statement 30 pts |
Student’s statement thoughtfully analyzes the relationship between the source images and generated images, considering:
24.1 to 30 pts |
Student’s statement addresses most, but not all reflection prompts, or addresses only some prompts with depth. 18.1 to 24 pts |
Statement does not engage with the reflection prompts, or only does so very superficially, Statement may have serious issues that impact readability. 0.1 to 18 pts |
No submission. 0 pts |
| Submission 10 pts |
Submission contains all required components and is formatted as requested in a single Google doc. 8.1 to 10 pts |
Submission contains all required components but may have formatting issues. 6.1 to 8 pts |
Submission may be missing a component and/or have severe formatting issues. 0.1 to 6 pts |
No submission. 0 pts |
