GEM: Inside the Machine—How GenAI Systems Work

Activity: Tracing the Machine's Mind

Purpose

This assessment helps you trace how generative AI systems shape the information you receive. By mapping the path from your own prompt to the AI’s response, you’ll visualize where human choices, training data, and algorithmic rules influence what knowledge gets produced and shared.

Download this activity template if you'd like to work on this activity digitally.

Part 1: Discover information using GenAI

• Select a question or topic that interests you—something you might genuinely want to learn more about. Examples:
  • How is climate change affecting Las Vegas?
  • Should college use of ChatGPT be encouraged or banned?
  • How do social-media algorithms affect mental health?
• Pick one generative AI tool (ChatGPT, Gemini, Copilot, Claude, etc.).
• Enter your question exactly as you would for an assignment.
• Copy or briefly summarize the full AI output.

Part 2: Identify influences

• Read the output carefully, then list at least three influencing factors that may have shaped it. Use these prompts for clues:
  • What kinds of sources might this AI have been trained on?
  • Does it include or exclude particular perspectives or examples?
  • Are there any patterns of tone, language, or region?
  • What platform rules or moderation filters might be at play?
  • Does it include linked sources?

• Write short notes beside each influence.

Part 3: Describe the output

• Make note of the following
  • Type of response: explanation, list, persuasive text, image, etc.
  • Evidence use: Did it cite sources or make claims without them?
  • Perspective: Whose voices or values seem most represented?
  • Limitations: What’s missing or uncertain?

Part 4: Make a map

• Create a simple three-column chart on paper or digitally:

• Then, look closely at your prompt and output:
  • Highlight or underline words in your prompt that guide the AI’s focus (e.g., should, best, in the U.S.).
  • In the AI System & Influences column, note how the system might have processed those cues (e.g., narrowed scope, mirrored tone, filtered sensitive content).
  • Draw lines or arrows showing how your prompt language and system factors together shaped specific parts of the output.
• Your goal: show the chain of influence — how your question, the AI’s design, and its training data combined to produce the final response.

Submission

If you would like credit for your work to earn gems towards the Mapping the Information Landscape badge, do either of the following:

• Snap a photo of the map you created for part 4 of this activity and submit that photo to this GEM WebCampus assignment.
• OR, make a copy of this document, write your responses and submit the document to this GEM WebCampus assignment.