Global AI Case Study
Throughout this module you have built a toolkit of analytical frameworks for understanding AI as a global phenomenon: the geopolitics of AI power, the dimensions of the development gap, the global AI divide, the data and labor supply chain, AI for global challenges, cultural diversity and AI, and international governance. In this lesson you apply that toolkit to a real, specific global AI issue — moving from abstract understanding to the kind of structured, evidence-based analysis that researchers, policymakers, and journalists actually produce.
How to Analyze a Global AI Issue
Not all analysis is equal. Surface-level analysis describes what happened. Structural analysis identifies why it happened, what forces shaped it, and what patterns it exemplifies. Your case study analysis should reach the structural level. A strong structural analysis answers four questions beyond the factual summary: Who benefits and who bears costs? Every AI deployment creates winners and losers. Following the distribution of benefits and harms — not just their aggregate — reveals the political economy of the issue. Who has power and who does not? AI issues almost always involve power asymmetries: between AI companies and users, between AI-producing and AI-consuming nations, between those who design AI systems and those who are subject to them. Identifying these asymmetries is essential. What structural factors explain the outcome? Individual bad actors or good intentions rarely explain AI outcomes fully. What economic incentives, regulatory gaps, historical inequalities, or technical constraints produced this outcome? A structural analysis identifies the forces that would reproduce the same outcome with different actors. What would change the outcome? Genuine analysis includes a theory of change — not optimism, but a specific argument about what intervention, policy, or shift in power would produce a different result.
The best case studies are specific, documented, and global in their implications. Avoid cases so broad they cannot be analyzed concretely ('AI and poverty') and cases so narrow they have no global dimension ('one company's chatbot had a bug'). Strong cases: AI-powered surveillance technology export from China to authoritarian governments; a specific AI healthcare deployment in a low-income country; a documented algorithmic decision system that affected a population across multiple countries; a specific international AI governance event and its outcomes.
Suggested Cases
The following cases are well-documented and analytically rich. Each connects to multiple module themes. Case A: Facial Recognition Export to Authoritarian Governments — Chinese AI companies including Hikvision, Dahua, and Huawei have sold facial recognition and surveillance infrastructure to governments in Central Asia, Africa, and the Middle East, where it has been used for political repression. This case connects to AI geopolitics, AI and economic development, the global AI divide, cultural diversity, and international governance. Case B: AI-Driven Refugee and Migrant Processing — The UN High Commissioner for Refugees (UNHCR) and several national governments have piloted AI systems to process refugee applications, determine eligibility, and allocate resources. Documented cases in the Netherlands, Denmark, and Jordan have raised serious due process concerns. This case connects to the harm divide, data sovereignty, cultural diversity in AI, and governance. Case C: Automated Content Moderation and the Global South — Facebook and other platforms use AI content moderation systems that perform significantly worse in non-English languages, resulting in both over-moderation (legitimate political speech removed) and under-moderation (hate speech and incitement allowed) in countries in Africa, Southeast Asia, and Latin America. During Myanmar's Rohingya crisis, Facebook's algorithm reportedly amplified incitement. This case connects to the language divide, cultural diversity, AI harms, and governance. Case D: AI Credit Scoring in East and Sub-Saharan Africa — A wave of mobile lending apps using AI credit scoring have expanded credit access in Kenya, Nigeria, Tanzania, and elsewhere — and generated intense controversy over predatory interest rates, opaque scoring, and abusive debt collection practices enabled partly by AI-driven customer profiling. This case connects to AI and economic development, the AI divide, data sovereignty, and governance. Case E: Climate Prediction AI for the Global South — Google's DeepMind, the European Centre for Medium-Range Weather Forecasts, and others have deployed AI weather and flood prediction systems specifically targeting vulnerable low-income regions. This case connects to AI for global challenges, the development gap, the access divide, and questions about dependency versus genuine public good.
Global AI Case Study Analysis
- This is the core work of Lesson 9. Choose one of the five cases described above (or propose an alternative to your teacher for approval). Produce a structured written case study analysis of 600-800 words, organized as follows:
- Section 1 — Factual Summary (100 words): What happened? Who are the key actors? What AI system is involved, and what does it do?
- Section 2 — Who Benefits, Who Bears Costs (150 words): Map the distribution of benefits and harms across different groups, regions, and populations. Be specific — name the groups and quantify where evidence exists.
- Section 3 — Power Analysis (150 words): Identify the key power asymmetries. Who designed the system? Who controls it? Who is subject to its decisions? Who has the ability to contest, appeal, or reform it?
- Section 4 — Structural Explanation (150 words): What structural forces — economic incentives, regulatory gaps, historical inequalities, technical constraints, geopolitical interests — produced this outcome? Argue that the same outcome would likely recur with different actors unless these structures change.
- Section 5 — Theory of Change (150 words): What specific policy intervention, technical change, governance mechanism, or shift in power would change this outcome? Be concrete and realistic — evaluate the feasibility and limits of your proposed change.
- Sourced claims only: cite at least three external sources (news investigations, academic papers, NGO reports, government documents). Your analysis must go beyond what this lesson describes.
- After submitting your written analysis, present your case in 4 minutes to the class. Classmates will ask at least one structural question each analyst must address.
Good analysis does not start with a conclusion and find evidence for it. It starts with evidence — documented facts, data, testimonies — and builds an argument from the ground up. If the evidence points in a direction you did not expect, follow it. The strongest analyses are often the ones that surface genuine tensions rather than neat conclusions: 'This system does X, which benefits group A but harms group B, and the structural forces that created this tension are Y and Z, which means any reform faces these specific obstacles.'
Which statement best describes the difference between surface-level and structural analysis of a global AI issue?
The suggestion to include a 'theory of change' in a case study analysis most directly addresses which analytical task?