%%{init: {'theme': 'base', 'themeVariables': {'fontSize': '18px', 'primaryColor': '#4a7c6f', 'primaryTextColor': '#1e293b', 'primaryBorderColor': '#334155', 'lineColor': '#64748b', 'secondaryColor': '#f9fafb', 'tertiaryColor': '#f9fafb'}, 'flowchart': {'useMaxWidth': true, 'nodeSpacing': 30, 'rankSpacing': 45, 'width': 1150, 'height': 650}}}%%
flowchart TD
A["Industrial Robots<br/>(Same Technology)"] --> B["Germany"]
A --> C["United States"]
A --> D["Sweden"]
B --> E["Strong unions<br/>Vocational ed<br/>Codetermination"]
C --> F["Weak unions<br/>Limited training<br/>Market-led"]
D --> G["Strong unions<br/>Active labor<br/>market policy"]
E --> H["Moderate inequality<br/>Job security"]
F --> I["Sharp inequality<br/>Job loss"]
G --> J["Low inequality<br/>Smooth transitions"]
style A fill:#1e293b,color:#f9f9fb,stroke:#334155
style H fill:#b7943a,color:#1e293b,stroke:#334155
style I fill:#b44527,color:#f9fafb,stroke:#334155
style J fill:#4a7c6f,color:#f9fafb,stroke:#334155
Technology and Social Change
Lecture 17: Final Thoughts
One Course, One Question
The Question We Kept Asking
We covered 12,000 years in one semester — from grain silos to large language models.
Underneath every lecture was a single question:
Why does the same technology enrich and liberate some societies — and impoverish or trap others?
Wheat, the printing press, the steam engine, robots, AI: the technology changes, the question does not.
Today: no new material. Just the answer — in five ideas, told through the most memorable slides of the course.
Five Ideas to Take With You
- Technology is not destiny — institutions decide what it does.
- Every technology redistributes power — and the losers fight back.
- Rising productivity does not mean rising wages — who captures the gains is decided by bargaining power and politics.
- Inventions appear where they pay, not where people are smartest — and since each one is built from earlier ones, innovation accelerates.
- You cannot predict the future — but you can analyze it.
If you remember these five sentences in ten years, this course did its job.
Idea 1 — Technology Is Not Destiny
The same machine, opposite outcomes — institutions decide which.
The Best Puzzle in the Course
Lecture 5
From 1450, the movable-type press spread across Europe and the Islamic world.
- Same machine. Same basic inputs. Same potential.
Yet the outcomes diverged radically:
- Dutch Republic: presses multiplied freely, book prices collapsed
- England: printing fueled a scientific revolution
- Ottoman Empire: Arabic-script printing banned for nearly 300 years
Why did identical technology produce such different outcomes?
What 290 Years of Blocking Cost
Lecture 5
| Indicator | Ottoman Empire | Western Europe |
|---|---|---|
| First Arabic-script press | 1729 | 1450s |
| Books published by 1800 | ~500 titles | Millions of titles |
| Literacy rate (c. 1800) | ~2–3% | 30–60% (varies) |
| Universities (c. 1800) | Madrasas (static curriculum) | Research universities emerging |
The calculus: the losers (scribal guilds, the ulema) were organized, powerful, and inside the palace. The winners were diffuse and outside it.
And the evidence is causal — Rubin (2014):
- The press predicted Protestant adoption only where authorities didn’t block it
- Technology is necessary but not sufficient
- Institutions are the binding constraint
The Same Puzzle, 570 Years Later: Robots
Lecture 11
Idea 2 — Every Technology Redistributes Power
…and the losers always fight back.
The Original Case: The Agricultural Paradox
Lectures 1 & 4
“Did we domesticate wheat, or did wheat domesticate us?” — Harari (2015)
Gained
- More total food
- Larger populations
- Permanent settlements
- Occupational specialization
Lost
- Longer work hours
- Worse nutrition (less diverse)
- New diseases (density)
- Social inequality
Paradox: Agriculture spread despite making most lives harder. So someone must have gained. Who?
The Evidence Is Written in Our Bones
Lecture 3
Male skeletal heights declined after the Neolithic transition. Source: Based on Angel (1984); Steckel (2008).
The height decline is consistent with the emergence of extractive hierarchies.
The Mechanism: Surplus → Hierarchy
Lecture 4
%%{init: {'theme': 'base', 'themeVariables': {'fontSize': '22px', 'primaryColor': '#4a7c6f', 'primaryTextColor': '#1e293b', 'primaryBorderColor': '#334155', 'lineColor': '#334155', 'secondaryColor': '#f9fafb', 'tertiaryColor': '#f9fafb'}, 'flowchart': {'useMaxWidth': true, 'nodeSpacing': 50, 'rankSpacing': 70, 'curve': 'basis'}}}%%
flowchart LR
A["<b>SURPLUS</b><br/>Storage &<br/>accumulation"] --> B["<b>SPECIALIZATION</b><br/>Division of labor<br/>& new roles"]
B --> C["<b>HIERARCHY</b><br/>Elites emerge<br/>& unequal power"]
C --> D["<b>INSTITUTIONS</b><br/>Property rights<br/>& political order"]
D -->|"Feedback: institutions<br/>stabilize extraction"| A
style A fill:#4a7c6f,color:#f9fafb,stroke:#334155
style B fill:#4a7c6f,color:#f9fafb,stroke:#334155
style C fill:#b44527,color:#f9fafb,stroke:#334155
style D fill:#b44527,color:#f9fafb,stroke:#334155
Source: Author’s illustration based on Harari (2015) and Acemoglu & Robinson (2012).
Replace “grain” with data and attention, and this is a diagram of the platform economy. Who controls the surplus controls the rest.
The Losers Fight Back: Polanyi’s Pattern
Lecture 2
When markets commodify things not produced for sale — land, labor, attention — society pushes back. Every era replays the cycle:
| Era | Technology | Fictitious Commodity | Counter-Movement |
|---|---|---|---|
| Agricultural | Enclosure | Land | Poor Laws |
| Industrial | Factory system | Labor | Labor rights |
| Digital | Platforms | Attention | Privacy/safety regulation |
| Emerging | AI | Cognition/identity | ? |
Why Do the Powerful Ever Concede?
Lecture 8
Elites didn’t reform out of goodwill — the costs of not reforming became too high.
- Threat: Luddism, Chartism, food riots — revolution was not hypothetical (France 1789, 1830, 1848)
- Organization: Unions gave workers collective power that individual bargaining never could
- Elections: Reform Acts (1832, 1867) forced politicians to answer to working-class voters
Idea 3 — Rising Productivity ≠ Rising Wages
Who captures the gains is a political question, not a technical one.
Engels’ Pause: Who Got the Growth
Lecture 8
Figure 1: Real wages lagged GDP per capita for decades
After 10,000 flat years, growth finally exploded — and for fifty years workers got none of it. Now look at the next slide carefully.
The Same Chart, Two Centuries Later
Lecture 15
Why Sharing Fails: The Bandwagon’s Two Conditions
Lecture 15
Wages rise with productivity only if two things are true (Acemoglu & Johnson, 2023):
- The technology still needs workers — it helps them, not replaces them
- Workers have the power to demand a share — unions, tight labor markets, a vote
If either fails, productivity rises but wages don’t. That is the chart you just saw.
History’s Warning — and Its Hope
Lecture 15
Why the bandwagon broke (Britain, 1790–1840):
- Machinery substituted for artisans (Condition 1 failed)
- Unions illegal, no vote, surplus rural labor (Condition 2 failed)
- Result: productivity gains went to industrialists for half a century
Why it eventually worked — deliberate institutional change:
- Trade unions legalized (1824, 1871)
- Mass public schooling (Forster Act 1870)
- Suffrage extended (1867, 1884)
Idea 4 — Inventions Appear Where They Pay
…not where people are smartest — and each one seeds the next.
What Is a Technology, Anyway?
Lecture 9
- A stick on the ground is just a stick
- Use it to lever a rock and it becomes a technology
- The stick didn’t change — so what did?
Arthur’s answer: a technology is nature put to a purpose. The invention isn’t the object — it’s the use.
And new technologies are built from old ones — so the more we have, the faster new ones appear.
Even ChatGPT is a recombination
ChatGPT feels brand-new, but its parts already existed:
- the text of the internet
- chips built for video games
- a decades-old trick — guessing the next word
The novelty was the combination, not the parts.
So Why Didn’t Rome Industrialize?
Lecture 3
Massive slavery → very cheap labor
Allen: cheap labor → low returns to labor-saving machinery
Scheidel estimates ~30-40% of Italy’s population enslaved at peak
Why build a water mill when slaves grind grain cheaply?
Rome had the engineering talent — aqueducts, concrete, roads. What it lacked was the incentive.
…And Why Britain Did
Lecture 8
Empire → high wages → incentive to mechanize
- Same spinning jenny: 2–3% return elsewhere, 30–40% in Britain
- The knowledge existed everywhere — only Britain’s wages made it pay
“The Industrial Revolution was the knock-on effect of British imperialism.” — Robert C. Allen
Rome and Britain: one theory, opposite directions — cheap labor kills mechanization, expensive labor demands it.
Idea 5 — You Can’t Predict the Future, But You Can Analyze It
The tools you leave with work on technologies that don’t exist yet.
The Tool: Think in Tasks, Not Jobs
Lecture 11
Key Insight
Jobs are bundles of tasks. Technology affects tasks, not jobs directly.
┌────────────────────────────────────────────────────────┐
│ JOB DECOMPOSITION │
│ │
│ "Accountant" = { data entry, analysis, client │
│ communication, judgment calls } │
│ │
│ Automation affects these tasks differentially: │
│ • Data entry: HIGH substitution potential │
│ • Analysis: PARTIAL augmentation │
│ • Communication: LOW substitution potential │
│ • Judgment: VERY LOW substitution potential │
└────────────────────────────────────────────────────────┘Three forces follow: displacement (−), productivity (+), reinstatement (+). The net effect is not predetermined (Acemoglu & Restrepo, 2019).
Why This Tool Matters: 47% → 9%
Lecture 11
Frey & Osborne (2017): “47% of U.S. jobs at high risk of automation.”
- They scored whole jobs — but a nurse gives injections (automatable) and comforts patients (not). One task isn’t the job.
- Redo it task-by-task and 47% collapses to 9% (OECD; Arntz et al., 2016).
One tool — task decomposition — deflated the scariest headline of the decade. That is what this course trained you to do.
The Direction of AI Is a Choice
Lecture 15
AI can be built to replace people — or to help them do more. Same technology, two directions (Acemoglu & Johnson, 2023).
- Today’s incentives push toward replacement — it’s cheaper to automate a worker than to augment one
- But that’s a policy choice, not a law of nature — taxes, research funding, and worker voice can change it
Will AI share the gains — or repeat the Engels Pause?
What to Take With You
Five Ideas — Now With Receipts
| Idea | Proof you saw today |
|---|---|
| 1. Technology is not destiny | Same press: Dutch Republic vs. Ottomans. Same robots: Germany vs. US. |
| 2. Technology redistributes power | Farming grew the food supply — yet farmers’ skeletons shrank. The surplus went to elites. |
| 3. Rising productivity ≠ rising wages | Engels’ Pause, 1820 — and the same chart again after 1979. |
| 4. Inventions appear where they pay | Rome’s slaves vs. Britain’s wages: one theory, opposite outcomes. |
| 5. You can analyze the future | One tool (tasks, not jobs) turned “47% of jobs at risk” into 9%. |
Three Mistakes You Will Now Catch Others Making
Lecture 1
1. “Tech is destiny” — “AI will eliminate jobs.” No: institutions decide.
2. Correlation ≠ causation — “Social media and depression rose together.” Find the mechanism.
3. Ignoring distribution — “Technology benefits society.” Which part of society?
You’ll hear all three for the rest of your life. Now you can answer back.
Thank you.
Grading
The course in relation to grading:
- Contributions to Class 33%
- Mid-term 33%
- Final exam 33%
Contributions to Class (Reminder)
- Presentation, physical presence, and class participation
- Submit two questions based on the class readings via Canvas each week
- Quality of questions matters, not just submission
Popescu (TEC) Technology & Social Change Lecture 17: Final Thoughts