What is fair?
Fairness is taking no advantage at the expense of any affected being1 — or of the living systems all beings depend on2 — that they cannot or could not have taken in return; making no distinction between beings that you would reject for yourself3; and where such advantage has already been taken, restoring what could not be defended4.
The reciprocity test
An advantage is unfair if the other side could never take it back.
Hover or tap a highlighted phrase.
One question, four parts
Can fairness be built into technology? Each part answers one piece of that question. Go in order, then test yourself.
Can fairness be built into technology?
The course frame in two steps: how a system turns data into decisions, then why fairness is hard to build in.
1 · How a system decides
Fairness can break at any step. Click a step to see the question it raises.
2 · Why fairness?
Everyone agrees with it. Nobody defines it the same way.
Fairness in Asian thought
Three steps for a newcomer. First the big idea that sets Asian ethics apart, then the five traditions behind it, then how each region compares with the West.
The big idea: from "me" to "between us"
So the fairness question changes from "Was I treated equally?" to "Was this relationship kept in balance?"
A broad tendency, not a rule. Based on Confucian role ethics (R. Ames, Confucian Role Ethics, 2011) and Watsuji's "betweenness" (Stanford Encyclopedia of Philosophy). Western thought also has relational strands, such as care ethics.
Five traditions, one at a time
How each region compares with the West
Published survey data: Hofstede's six cultural dimensions, scored 0–100. Individualism and power distance link most directly to the relational self and role duties above. Hover an axis name for its meaning.
Source: Hofstede, Hofstede & Minkov, Cultures and Organizations (3rd ed., 2010), dimension data matrix (version 2015, geerthofstede.com). SE Asia = average of Thailand, Indonesia and Vietnam; West = average of USA, UK, Germany and France (rounded). Culture scores describe national averages, not individuals.
AI safety directions, East and West
Values become real when governments write rules. First see where nine countries sit, then check how each score was built, then read why one country should not control AI-powered defence.
1 · Where each country sits
Right means more binding law. Up means more focus on controlling risk. Click a dot for the evidence.
2 · The checklist behind the map
Ten yes/no questions. A country's position on the map is just its points. Hover a column name for the question, or a mark for the evidence.
● yes = 1 · ◐ partly (only some rules, only government use, only some states, or not yet in force) = ½ · ○ no = 0. Score = points × 20. Each item was checked in October 2026 against official texts and law-firm trackers; links are in each country card. Some ½ marks are judgement calls.
3 · One country should not hold the keys
Open source runs banks, hospitals and power grids. AI now finds its weak spots at scale. Who controls the defence?
The point: coordination is needed, but centralising it is the flaw. Splitting into rival regional systems would be just as weak.
What worked before: neutral ground
Radio spectrum · ITU
Airwaves are finite and cross borders.Flip ↻Radio spectrum · ITU
- States keep
- Licensing their own users
- States give
- The right to interfere across borders
- Holder
- A treaty and a shared register
Nuclear · IAEA
The same power can light a city or destroy one.Flip ↻Nuclear · IAEA
- States get
- Access to peaceful uses
- States give
- Accepting inspections
- Holder
- Vienna. The US started it but did not own it.
Medicines · patents
Life-saving drugs locked behind patents.Flip ↻Medicines · patents
- Firms keep
- Their patent rights
- Public gets
- Fair, tiered access
- Holder
- Neutral pooled access
Three asks
- Reciprocity: allies were cut off and could not cut back. That is an advantage they cannot take in return.
- Self-test: would the US accept another country switching off its defences?
- Restoration: federated, neutral custody gives back what single-country control took away.
- East–West link: federation treats security as interdependence, close to the relational view in Part 2.
Values East and West: same meaning or not?
This pulls Parts 2 and 3 together. First test single words, then compare the two views as a whole.
1 · Same word, same meaning?
Tap a value to compare its two readings. Each card names its source.
2 · The two views side by side
Tendencies, not rules. Each side is internally diverse, and "Asian values" should not be treated as one block (Sen, 1997).
3 · Where they meet
Sources: SEP: Chinese Ethics · SEP: Justice · SEP: Privacy · SEP: Relational Autonomy · SEP: Dignity · SEP: Watsuji · SEP: Ethics in Indian and Tibetan Buddhism · SEP: Moral Status of Animals · Ames & Rosemont, Confucian Role Ethics · Kitiyadisai (2005), privacy in Thailand · Apology in Japanese justice · Sen (1997), Human Rights and Asian Values · Wong (2020) · Fung & Etienne (2021), Chinese vs EU AI principles · Analects (ctext.org) · Gandhi's Talisman
Fairness Compass
Write your own definition of fairness. Claude checks it against 8 criteria taken from the sourced comparison in Part 4, quoting your words for each one. The page then counts the matches, so every point can be traced.
How the score works · 8 criteria, each with a sourced East and West reading
East score = criteria matching the East reading ÷ criteria your definition addresses. West score works the same way. The two shared ideas count for both. A mark only counts if its quote really appears in your text. Claude reads each definition three times and each criterion keeps the majority answer, so one odd reading cannot swing the score. Labels: a gap of 20 points or more means "Leans East" or "Leans West"; both 60 or above within 20 points means "Bridge". Traditions and regions come from Part 2.
Class board
Each dot is a classmate's definition. Hover a dot to see whose it is.
Sources and readings
- Computing: Anthropic unveils Project Glasswing
- Via Satellite: Anthropic launches Project Glasswing
- Yahoo Finance: Glasswing expanding to 200 partners
- NanoEthics (2025): non-Western perspectives on AI ethics
- EngageMedia: Alternative ethical frameworks for AI
- HKBU: East–West dialogue on AI ethics
- WGBH: What can AI learn from non-Western philosophies?
- Confucius, Analects 13.23, 15.24 · Mencius 1B:8, 3A:4 · Laozi, Daodejing ch. 77 (trans. D.C. Lau) · Watsuji, Rinrigaku · Sen, The Idea of Justice (2009) · Rawls, A Theory of Justice (1971)
- Daodejing ch. 77, D.C. Lau translation and commentary
- Stanford Encyclopedia of Philosophy: Watsuji Tetsurō
- Gandhi's Talisman (1948)
- R. Ames, Confucian Role Ethics: A Vocabulary (2011) · Seventeen-Article Constitution (604)