Human Centered AI

Submit a one-page summary that includes:

• Your cultural background.
• Your background in AI and HCI.
• What you know or have done with artificial intelligence.
• What you hope to get out of this class.

This assignment challenges you to move beyond a purely technical understanding of AI. By analyzing a real-world scenario, you will identify ethical, social, and design flaws in an AI system; explain potential negative consequences; and argue for a broader human-centered approach.

Part 1 — Foundational Videos

Watch at least three of the following before completing the case study:

• Why We Need Responsible AI — Nokia Bell Labs
• Why Do We Need Responsible AI? — IBM Research
• AI Is Dangerous, but Not for the Reasons You Think — TED
• AI Has a Fatal Flaw — Slidebean
• Apple's AI Disaster — ColdFusion
• Gen AI Gone Wild — Fireship

Part 2 — Case Study: "Optimal Hire"

Innovate Inc. has deployed an AI hiring tool called Optimal Hire that automates initial screening of job applications. The system was trained on ten years of historical data and assigns applicants a Success Score of 1–100. HR staff interview only those scoring 85 or higher. The models are opaque — developers cannot explain individual scores. The company celebrates a 400% increase in hiring efficiency.

Part 3 — Essay (1,000–1,200 Words, PDF)

Address the following:

1. Identify at least three major flaws — technical, ethical, organizational, and societal.
2. Explain who is harmed — consider individual applicants, Innovate Inc., and society.
3. Why code is not enough — explain why algorithmic fixes alone are insufficient and argue for human oversight, transparency, diverse teams, inclusive design, and regular auditing.

Watch three foundational videos on HCAI, then read Shneiderman's seminal paper:

• What is Human-Centered AI? — Peter Norvig
• A Call for Human-Centered AI — DLD 23
• What is Human-Centered AI and Why Do You Need It?
• Shneiderman, B. — Human-Centered AI: Reliable, Safe & Trustworthy (paper)

Write a 750–1,000 word essay that: defines HCAI in your own words; explains why HCAI matters with specific examples; connects theory to practice through a real or hypothetical system; and identifies the biggest future challenge for HCAI implementation.

Write and publish a blog post of at least 100 words covering: the heart of the lawsuit; the verdict; generative AI's role (ChatGPT, Gemini, Claude, etc. as context for competition); and what happens next. Cite at least three credible sources.

Format: Medium, LinkedIn Articles, WordPress, or PDF/Word. Must include a References section.

Part 1 — Analytical Essay (50%): Choose One Prompt

Prompt A — The Echo of an AI Winter? Based on your understanding of historical AI Winters, analyze the current generative AI moment. Are we seeing early signs of another winter, or are today's fundamentals different? Predict how the investment and hype cycle may unfold over the next 5–10 years using specific historical parallels.

Prompt B — Old Problems in New Code. Analyze the emergence of AI-AI bias as a modern unforeseen consequence. How does it parallel historical problems like brittleness or embedded bias? Predict long-term social or economic consequences if left unchecked.

Part 2 — Coding: Sentiment Analysis of AI Hype Cycles (50%)

Write a Python script that reads the dataset below and uses TextBlob or NLTK's VADER to classify headlines as positive, negative, or neutral. Aggregate sentiment by era and visualize with a bar chart (Matplotlib or Seaborn).

era,headline
1980s_Boom, The Future is Now: Expert Systems to Revolutionize Every Industry
1980s_Boom, Intelligent Machines Promise a New Era of Economic Prosperity
1980s_Boom, Government Pours Millions into Fifth-Generation Computer Project
1990s_Winter, AI Fails to Deliver: The Unfulfilled Promise of Thinking Machines
1990s_Winter, Funding for AI Research Dries Up Amidst Disappointing Results
1990s_Winter, Expert Systems Deemed Too Brittle for Real-World Use
2020s_GenAI_Boom, Generative AI Set to Add Trillions to the Global Economy
2020s_GenAI_Boom, Breakthroughs in Large Language Models Astonish Researchers
2020s_GenAI_Boom, Is Artificial General Intelligence Finally Within Reach?
2020s_GenAI_Boom, Concerns Mount Over AI Job Displacement and Automation

Submit: Python script + generated chart. Reference your sentiment results in your essay for a data-driven perspective.

Read all three papers, then create two analytical diagrams:

• A Brief History of Artificial Intelligence
• A Brief History of AI: How to Prevent Another Winter
• Human-Centered AI: Reliable, Safe & Trustworthy

Diagram 1 — AI Paradigm Evolution Timeline: Identify at least six critical junctures; explain assumptions, limitations, and breakthroughs; connect to current trends.

Diagram 2 — Human-Centered AI Implementation Framework: Show relationships between trustworthiness, reliability, safety, transparency, and ethics. Include concrete examples and identify design tensions.

Watch the Meta Connect 2025 keynote and write a 500–750 word reflective essay covering:

1. Two or three most significant announcements.
2. The single most groundbreaking announcement and its potential impact.
3. How announcements advance or challenge Meta's metaverse vision.
4. Evaluation of one announcement through a Human-Centered AI lens.

Format: PDF, 12pt Times New Roman, double-spaced, name and date at top.

Watch all three videos, then read the papers in the Mockups/Storyboards folder on Canvas:

• What is a Mockup? — Envato Tuts+
• How to Create a UX Storyboard — Nielsen Norman Group
• Learn UX Storyboarding in 5 Minutes — UXtweak

Deliverable: Create a diagram illustrating the relationship between mockups, storyboards, and the HCAI design process. Include a 1–2 paragraph explanation making explicit connections to the source materials. Submit as a single PDF, Word, or PowerPoint document.

You are part of the product team for FlowState, a web-based project management platform. Analyze one month of user interaction logs, identify a meaningful user need, and design an AI-powered feature to address it.

Dataset

timestamp,userID,sessionID,action,details,session_duration_minutes
2025-08-21 09:05:12,user_101,sess_abc,login,success,45
2025-08-21 09:06:20,user_101,sess_abc,view_dashboard,project_alpha,45
2025-08-21 09:07:31,user_101,sess_abc,create_task,{"task_name": "Draft Q3 Report"},45
2025-08-21 09:07:55,user_101,sess_abc,create_subtask,{"parent_task": "Draft Q3 Report", "subtask_name": "Gather sales data"},45
2025-08-21 09:08:21,user_101,sess_abc,create_subtask,{"parent_task": "Draft Q3 Report", "subtask_name": "Outline report structure"},45
2025-08-21 09:15:45,user_101,sess_abc,assign_user,{"task_name": "Gather sales data", "assigned_to": "user_102"},45
2025-08-21 09:25:11,user_101,sess_abc,set_due_date,{"task_name": "Draft Q3 Report", "date": "2025-09-15"},45
2025-08-21 09:40:30,user_101,sess_abc,change_view,gantt_chart_view,45
...
2025-08-21 11:30:50,user_108,sess_jkl,login,success,5
2025-08-21 11:31:20,user_108,sess_jkl,view_dashboard,project_gamma,5
2025-08-21 11:32:00,user_108,sess_jkl,change_view,kanban_view,5
2025-08-21 11:33:00,user_108,sess_jkl,change_view,list_view,5
2025-08-21 11:34:00,user_108,sess_jkl,change_view,kanban_view,5
2025-08-21 11:35:50,user_108,sess_jkl,logout,success,5

Part 1 — Data Analysis (50 pts)

Using Python + Pandas (or Google Colab), answer: What are the five most frequent actions? What is average session duration? What pattern follows create_task? What does user_108's behavior suggest about the view-switching experience? Write a 2–3 paragraph summary identifying one clear user problem.

Part 2 — AI Feature Mockup (50 pts)

Name your feature (e.g., Smart Subtask Suggestions), write a short description of what it does and how it addresses the problem, create a mockup (wireframe acceptable), and provide bullet-point design justifications linking each choice to your analysis.

Watch five UX videos and read papers on Contextual Inquiry, Participatory Design, and User-Centered Design. Then write a 2-page paper (12pt, double-spaced, 1-inch margins) addressing:

1. The Role of Context — Why is context critical when designing AI systems?
2. Applying a Human-Centered Process — Outline a user-needs discovery plan for an AI tool that helps project managers track team progress. Identify methodologies, explain selections, and describe expected insights.

Read the report Inclusive and Secure Artificial Intelligence: A Global Perspective and listen to the accompanying podcast. Adapt the AI assistant built during bootcamp using the principles of inclusion, cultural awareness, and fairness.

Part 1 — Reading and Analysis (30 pts)

Answer five questions (2–4 sentences each) on: AI neutrality and facial recognition; digital colonialism; content moderation failures in non-Western languages; cultural impact assessments; and participatory design.

Part 2 — Coding Challenge (40 pts)

Enhance your assistant with: culturally aware greetings by region/language; diversified knowledge base with global cultural examples; audited and gender-neutral language; and improved handling of culturally ambiguous concepts through clarifying questions.

Part 3 — Reflection Essay (30 pts, 400–600 words)

Discuss: the biggest challenge of retrofitting AI for cultural awareness; two policy recommendations from Section 4 that your modifications implement; and one additional feature you would add.

Same assignment as Week 5, continued with the additional context of AI risk frameworks covered this week. Submit a single PDF with all three parts.

Read "We Are AGI Researchers, and We Are Asking for Caution" in TIME. Write:

Part 1 — Summary (200–250 words): Who wrote it, what AGI/superintelligence risks do they highlight, and what actions do they advocate for?

Part 2 — Critical Analysis (700–800 words): Whose perspective does the letter prioritize? How do proposed solutions align with HCAI principles? What immediate risks does the existential focus overlook (bias, job displacement, privacy, etc.)? If you rewrote the conclusion from a strictly HCAI viewpoint, what would you recommend?

Transform user research findings into a coherent vision for a Human-Centered AI system. A central requirement is traceability: reviewers must follow a clear path from raw data → affinity clusters → personas → design decisions.

Part 0 — Data-Driven Personas (30 pts)

Describe data sources; summarize 5–8 affinity clusters with representative quotes; create 2–3 personas with name, background, goals, needs, frustrations, context, constraints, and technology habits; justify each persona with explicit links to clusters and quotes.

Part 1 — Mockup (40 pts)

List major AI features; create a high-fidelity mockup showing layout, screens, navigation, and key interactions; annotate with design rationale callouts connecting each element to persona needs.

Part 2 — User Storyboard (30 pts)

Choose one persona and one concrete goal. Show the complete journey: problem → interaction → resolution. Include user actions, thoughts/emotions, AI actions, and captions.

Read both papers:

• The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation
• Fairness and Abstraction in Sociotechnical Systems

Watch the three videos on AI risk management (NIST framework, risk mitigation, conversational AI risk).

Part 1: For each paper, write a 1–2 paragraph summary of main risk/harm concepts, highlighting examples relevant to your final project.

Part 2: For each video, write a 1–2 paragraph analysis of key takeaways and how they connect to the papers.

Critically evaluate your proposed AI tool for risks, harms, and unintended consequences. May be completed individually or as a team (all members submit separately).

Part 1 — Risk Identification and Analysis (40 pts)

Identify at least three risks. For each, specify: the category (fairness/bias, privacy, security, malicious use, reliability), a detailed description of how it could emerge and worst-case scenarios, and the affected populations.

Part 2 — Mitigation Plan (40 pts)

For each risk, describe mitigations at three stages: Design (UI changes, transparency mechanisms, inclusive design), Development (fairness metrics, bias testing, human-in-the-loop), and Deployment (audit cycles, reporting mechanisms, usage restrictions).

Part 3 — Ethical Reflection (20 pts, 2–3 paragraphs)

Reflect on: how this exercise changed your perspective; why proactive risk consideration matters; and the biggest ethical challenge specific to your AI tool.

What to Submit

• Slides for your 10-minute presentation.
• A short demo video (1–3 minutes).
• Code with sufficient structure and comments to be understood and run.

Presentation Format

Each team/individual has 10 minutes, followed by a ~1-minute Q&A. Structure your presentation around: Problem → Approach → Demo → Reflection, or any flow that best tells your story.

Evaluation Criteria

1. Integration of Human-Centered Design (40%) — Show how HCAI principles are woven throughout the project, with clear attention to user needs, context, and values.

2. Narrative and Communication (60%)