MediaArtTutorials

Option 2: Glitching tools - AI data poisoning

Groups of 3–4 students · Research-Creation Project

Project Overview

In this project, you will critically and creatively explore AI data poisoning as a glitch art methodology. Through adversarial prompting, data poisoning (using Nightshade), and image-to-image remixing (via Pollo.ai), you will generate a series of speculative images that challenge machine perception and aesthetic logic. This process allows you to examine the ethical, technical, and artistic stakes of manipulating AI-generated media.

Artistic Production Guidelines

As a group, you must create a cohesive series of images that reflect a shared theme, aesthetic, and conceptual direction.
Each student is responsible for producing 3–4 final images that align with the group’s agreed-upon visual and thematic logic and saving all of the prompts used to create them.
Your group must also decide how the images will be arranged and visually assembled as a cohesive series in the digital portfolio.
Your arrangement strategy should enhance the impact of the series, amplifying the glitch aesthetics and critical themes you’re engaging with.

Index

W8: Conduct Individual Research

Software

Microsoft Copilot – For generating base AI images using adversarial prompts

Access with McMaster email credentials
Nightshade (data poisoning tool) – For poisoning image data and altering AI model perception
Pollo.ai (Image-To-Image AI Generator) – For remixing poisoned images with mislabeled or contradictory prompts
Photopea or Adobe Express – Free collage generator – For image editing.
Adobe Express – Free collage generator or Microsoft PowerPoint – For moodboard creation.

Project Components

1. Conduct Individual Research

Each student will choose one theme:

Glitch Aesthetics and Machine Error as Art
Algorithmic Bias and the Ethics of Data Corruption
Poetics of Prompting: Language, Contradiction, and AI Interpretation
Speculative Media and Machine Perception

Write a 300–400 word report grounded in:

1 academic source
2 public sources (e.g., article, talk, podcast, artwork)
Connect research to identity, technology, and speculative design

2. AI Data Poisoning Workflow

Individually and collectively, create a cohesive series of images by following this structured workflow:

Generate base images using adversarial prompting in Microsoft CoPilot and slightlyh modify them using any image editing software.
Apply Nightshade to poison selected images, distorting or confusing their machine-learned features.
Remix poisoned images in Pollo.ai, using mislabelled or contradictory prompts to create unexpected visual outcomes.
Document each step in your individual Process Journal, including prompts, screenshots, and observations.

3. Assemblage and Final Image Arrangement

As a group:

Select, sequence, and present your strongest images in a curated digital portfolio (PDF).
Consider:
- Layout: grid, visual narrative, collage, or progression
- Supporting text: image titles, captions, or poetic fragments
- Consistent visual formatting: aspect ratio, color treatment, and tone
The final format can include visual sequences, speculative layouts, or layered image-text compositions that amplify your artistic intent.

4. Group Self-Assessment

Complete the Self-Assessment Form as a group.
Ensure your responses are thoughtful and critically reflect on your final outcome.

Project Breakdown

Activity	Focus	Points
Week 8 Activities	Conduct Individual Research	5
Week 9 Activities	Final Research Report + AI Data Poisoning Workflow	5
Week 10 Activities	Group Brainstorming, Project Planning, and Image Testing	5
Week 11 Activities	Complete final images	5
Week 12	Class showcase: work-in-progress projects	5
Final Submission	Final project + deliverables	13
Total Points		38