Learning Objectives by Activity
Note: This page is largely AI-generated. Please report any errors or omissions to the instructor.
This table organizes the course activities by date and identifies the key learning objectives for each session.
| Date | Week | Activity | Learning Objectives |
|---|---|---|---|
| Week 1 | |||
| Wed Sep 3 | W1D1 | Course Introduction | • Articulate personal goals and interests in human-centered AI • Understand course structure and collaborative approach • Analyze the current AI landscape critically • Establish community norms for AI use in learning • Explore project interests and find potential collaborators |
| Fri Sep 5 | W1D2 | Vibe Prototyping | • Create rapid AI-powered prototypes using vibe-coding tools • Design specific user tasks to test prototypes with peers • Conduct informal usability testing and observe user interactions • Compare predicted vs. actual user behavior • Reflect on low-fidelity prototyping as a design exploration method • Build community connections through collaborative testing |
| Week 2 | |||
| Mon Sep 8 | W2D1 | Prompting | • Understand how LLMs work (prompts, responses, system messages, tokens) • Design effective prompts by treating LLMs as contractors needing clear instructions • Structure prompts to avoid prompt injection using delimiters • Implement LLM functionality via API calls in Python • Evaluate variability in LLM outputs due to sampling temperature |
| Wed Sep 10 | W2D1 (cont.) | Prompting | • Apply context engineering techniques to provide relevant information • Use function calling (tools) to allow LLMs to request additional information • Identify best practices for prompt engineering • Understand when and how to use context engineering for improved performance |
| Fri Sep 12 | W2D2 | Course Advisor Bot | • Implement Agentic RAG (Retrieval-Augmented Generation) techniques • Use structured output (Pydantic models) to constrain LLM responses • Design and implement search systems over structured data • Own control flow in agentic systems rather than letting LLM fully drive interaction • Test and measure AI performance across multiple dimensions (failure rate, latency, relevance) |
| Week 3 | |||
| Mon Sep 15 | W2D3 | Testing Email Feedback Bots | • Measure inter-rater reliability in human evaluation of AI outputs • Design evaluation prompts for LLMs to serve as judges • Compare rating-first vs. reasoning-first evaluation approaches • Quantify variability in both human and LLM evaluations • Design regression tests for ongoing performance monitoring • Understand the role of guidelines and rubrics in evaluation |
| Wed Sep 18 | W3D2 | Project Work | • Develop specific, narrow use cases for human-centered AI • Apply human-AI interaction guidelines • Iterate on project proposals based on feedback • Submit detailed project proposals |
| Fri Sep 19 | W3D3 | Testing Course Advisor Bots | • Conduct stakeholder analysis to identify who is affected beyond direct users • Recognize how technology embeds values through design choices • Perform red teaming to uncover technical flaws and potential social harms • Test with edge cases, adversarial inputs, and diverse user scenarios • Identify systematic bias patterns that create predictable unfairness • Apply the “ripple effect” method to map broader AI system impacts |
| Week 4 | |||
| Mon Sep 22 | W4D1 | Local LLMs | • Install and run local LLMs using Ollama • Understand resource requirements (compute, memory, storage, disk space) • Compare local vs. cloud LLM deployment trade-offs (cost, privacy, latency, performance, control) • Measure LLM performance metrics (time to first token, total generation time) • Examine model specifications (parameters, context length, licensing) • Understand caching mechanisms (model loading, KV cache) |
| Wed Sep 24 | W4D2 | Course Progress & Student Input | • Reflect on most helpful activities and interesting discussions so far • Provide input on future course direction (activities and readings) • Synthesize learning from prototyping, LLM programming, and evaluation |
| Fri Sep 26 | W4D3 | Design Norms & System Analysis | • Apply normative design lenses (transparency, justice, trust, caring) to AI systems • Critically analyze real AI systems (Perusall’s comment quality scoring) • Envision and evaluate alternative design choices • Reflect on personal growth in AI literacy • Consider how course projects contribute to professional portfolios |
| Week 5 | |||
| Mon Oct 1 | W5D1 | Recommender Systems Analysis | • Examine design decisions in recommendation algorithms • Understand how success metrics affect user behavior • Apply “Guidelines for Human-AI Interaction” to projects • Learn about Model Context Protocol (MCP) architecture • Update project proposals with refined scope |
| Wed Oct 3 | W5D2 | Agents & Tool Calling | • Define tools (functions) that LLMs can call to extend capabilities • Implement agentic loops where LLMs chain multiple tool calls • Separate tool definitions from agent logic using MCP • Build MCP servers that expose tools via standard protocol • Build MCP clients that connect to tool servers • Maintain conversational context across multiple user inputs • Design system prompts to guide agent behavior in tool usage |
| Fri Oct 5 | W5D3 | Deep Research & Goodhart’s Law | • Understand Goodhart’s Law: when metrics become targets they lose validity • Analyze system prompts and agentic loop design patterns • Examine deep research agent architectures • Discuss what we desire vs. what we optimize for • Reflect on reductionism and perverse incentives in AI systems |
| Week 6 | |||
| Mon Oct 6 | W6D1 | Multimodal AI & Self-Assessment | • Discuss training data sources for image, video, and audio models • Evaluate challenges in assessing multimodal AI quality • Conduct self-assessment across all course learning outcomes • Identify strongest and weakest areas of growth • Continue Course Advisor v2 work |
| Wed Oct 8 | W6D2 | Peer Feedback | • Conduct structured peer review using project rubric • Provide constructive feedback to other teams • Receive and integrate feedback on own project • Reflect on feedback process and implementation plans |
| Fri Oct 10 | W6D3 | Debate | • Distinguish strong arguments from weak arguments based on evidence • Support claims with specific citations from course readings • Challenge existing arguments with clarifying questions and counterpoints • Recognize complexity in controversial questions about AI • Build arguments that extend rather than repeat existing points • Synthesize multiple perspectives on contentious AI issues • Practice evidence-based argumentation |
| Week 7 | |||
| Mon Oct 13 | W7D1 | Project Presentations | • Present final projects demonstrating technical implementation • Explain critical analysis of design decisions and alternatives • Showcase user-centered design process and iterations • Provide constructive feedback on peer presentations |
| Wed Oct 15 | W7D2 | Project Presentations & Reflection | • Complete project presentations • Synthesize learning across technical, ethical, and design dimensions • Reflect on personal growth throughout the course • Identify how course concepts apply to ongoing work and career development |
Key Learning Themes
Throughout these activities, students develop competencies in:
- Technical Implementation: From basic prompting to complex RAG systems, structured outputs, tool calling, local models, and agentic systems
- Human-Centered Design: Rapid prototyping, user testing, stakeholder analysis, and peer feedback
- Critical Evaluation: Systematic testing, bias analysis, reliability measurement, and understanding evaluation limitations (Goodhart’s Law)
- System Analysis: Analyzing existing AI systems (recommender systems, Perusall, deep research agents) through multiple lenses
- Normative Design Frameworks: Applying design norms (transparency, justice, trust, caring) and considering human flourishing
- Ethical Considerations: Understanding embedded values, fairness concerns, broader social impacts, and debates about privacy, education, and work
- Collaborative Problem-Solving: Peer testing, team projects, iterative improvement, and constructive debate
Assessment Alignment
These learning objectives align with the course’s project-based assessment approach, where students apply these skills to develop and critically analyze their own human-centered AI system throughout the semester.
Additional Activities
Throughout the course, students also complete:
Course Advisor Bot 2.0
An extended project to: - Architect conversational AI systems that maintain context across multiple turns - Design and implement MCP tools with well-defined interfaces for agents - Evaluate AI system performance using automated testing frameworks - Analyze security and trust implications of tool-based AI systems - Apply stakeholder-centered design to identify and address systematic biases
Project Component Learning Objectives
Throughout weeks 4-7, students complete a three-part project that integrates:
Critical Analysis Component: - Analyze design decisions’ impact on human flourishing - Envision and evaluate alternative design approaches - Consider agency, capability building, privacy, and relationship to work/learning
Technical Evaluation Component: - Build functional toy models of AI systems - Design and conduct quantitative evaluations - Document performance analysis and system comparisons
User-Centered Design Component: - Develop testable prototypes incorporating user feedback - Iterate based on user testing results - Reflect on how building reveals blind spots in original analysis