Seeing Human-AI Interaction

For this project, you will work in groups of three to identify and analyze a real-world human–AI ecosystem. Each group can have at most one design student.

Learning Goals

  1. Recognize actual machine learning (ML) systems despite the ever-shifting colloquial meaning of the term “Artificial Intelligence”;
  2. See the human in ML systems;
  3. Analyze the value flow underlying an ML socio-technical system.

Weekly Tasks

Task 1

Form a team and select one of the topics below as your group’s focus.

  • An AI product that assists with programming (e.g., GitHub Copilot, Cursor, Gemini Code Assist, and various AI assistants available on the Visual Studio Code Marketplace)
  • An AI product that help people find dates (e.g., the models that match people on match.com)
  • An AI product that support 3D tasks (e.g., computer vision systems used in assembly lines, n warehouses, or architecture/construction, and AI features embedded in CAD software)

💭 Avoid selecting general-purpose AI tools such as ChatGPT, Claude, or Google Search. These tools involve too many stakeholders, making it difficult to complete this assignment.

💭 You will explore related aspects of this topic throughout the semester, so please choose one that genuinely excites and motivates you.

🛎️ Submit your choices by completing this form.

Task 2

Identify an ML model in a real-world product or service related to your chosen topic. Illustrate the model’s input data, output data, the data patterns it learns, and its underlying algorithms using the provided stakeholder map template (Figma).

🛎️ One group representative should submit your group’s work and team contract through this link.

🛎️ Every student: Time to complete your first peer evaluation here.

Task 3

Identify the people and organizations that fulfill the following roles for your ML system. Annotate them on the stakeholder map created in the previous week.

  • Data creators (and sometimes labelers)
  • Model designers
  • Model engineers
  • Intended direct users
  • Unintended direct users
  • Intended indirect stakeholders
  • Unintended indirect stakeholders

Start to annotate the financial, practical (e.g., time-saving), emotional value that different stakeholders provide to one another within the ecosystem.

🛎️ One group representative should submit your group’s work through this form.

Final submission

Improve all deliverables from earlier tasks by thoughtfully integrating the feedback you have received.

Afterwards, identify the value flow underlying the ML socio-technical system.

  • On your stakeholder map, annotate the financial, practical (e.g., time-saving), emotional, and other value (e.g., justice, health, human dignity) that different stakeholders provide to one another within the ecosystem.
  • Analyze how design choices in the ML system—–whether intentional or unintentional—–shape these value relationships;
  • Analyze how these value flows contribute to equilibrium, exploitation, or imbalances within the ML ecosystem (i.e., whether they create stable, mutually beneficial exchanges or result in certain populations being disadvantaged or exploited).
  • Each group member individually: Propose one design that improves this current value flow. The design can be modifications of the AI product being analyzed, or an entirely new AI product.

🛎️ One group representative should submit your group’s work through this form. The submission includes a link to the group’s FigJam board, and a link to a 6-page slidedeck:

  • Slide 1: Team name and members
  • Slide 2: Analysis of how ML system design choices shape value relationships
  • Slide 3: Analysis of value flows and ecosystem dynamics
  • Slides 4-6: Individual design improvements proposed by each team member

🛎️ Each group member: Complete your peer evaluation individually via this form.