What is under way and what has been claimed, each with a note on where it stands.
Four decisions compared: Austria-Hungary's war on Serbia, Germany's return to unrestricted submarine warfare, Germany's permission for Lenin's transit, and Emperor Karl's secret approach for a separate peace. Had each gone the other way, how different would the outbreak and end of the war, its outcome and the postwar order have been?
In the middle of the third century the Roman empire suffered civil wars and invasions across its frontiers, and for a time split into three. Aurelian reunited it, Diocletian reorganised how it was governed, Constantine carried the reforms further, and the empire lasted. Supposing in turn that one of the three had not been in his place, how would Rome's course have differed? The study uses this to test a judgment: that individuals matter most at the moments when structures are being transformed.
Around 1415 the Ming moved the shipment of tax grain to the north from the sea route to the newly reopened Grand Canal; in 1433 Zheng He's seventh voyage came home, and the court sent no more fleets overseas. Had these two decisions gone the other way, how would the rest of the Ming have differed? The study first asks whether anyone at the time argued for the other road, and then whether the shipbuilding, finances and transport of the time could have sustained it.
To say that A caused B is to say that without A, B would not have happened as it did. For one causal claim made by another researcher here, the study writes out the counterfactual it contains, names the conditions it holds fixed, tests against the sources whether those conditions held, and hands the result back to her. One claim is taken first, through one complete stage.
The Islamic courts of Istanbul left registers from the sixteenth to the nineteenth century, of which a hundred volumes have been published in transcription, 54,590 entries in all, recording how the people who came to sue, contract or report described themselves and which quarter they lived in. Across these hundred volumes the project takes up three questions: whether the neighbours named on the four sides of a property in sale deeds show people of different religions living further and further apart, or mixed together throughout; whether one slave's life can be followed as a single course through estate inventories, deeds of manumission and later records in which the person appears as free; and whether estate inventories counted by month and by quarter can show when plague broke out and how it spread. It will also write the lives of three named people.
Historical explanations often say that some condition brought about some outcome, but history happens only once and there is no control group. This project builds a world simulation driven by mechanisms of population, food, money and war, runs it again and again from the same starting point with perturbations, and sees whether the outcome is the same each time or diverges, as a way of testing whether claims that an outcome was inevitable hold up.
Before war broke out in July 1914, did the London bond market sense that war was coming? British and Russian government bond prices, as printed day by day in The Economist, are converted into the probability of war implied by the market; over the same period a language model that has read only pre-war texts gives its own probability, and the two curves are set side by side to see which rises first and by how much.
Before a political crisis breaks out, those involved can seldom tell whether it will escalate into war; afterwards, historians often say the outcome was all but unavoidable. This project pairs crises by the conditions said to drive escalation, such as military strength, alliances and geography, to see whether the share that actually escalated to war is markedly higher; it then has a language model that has read only texts from before the starting point act as an observer after the fact, judging how hard the outcome was to avoid, and compares that judgment with the real spread obtained by simulating the same starting point thirty times, to see whether it underestimates how far history could have diverged.
What share of their income or output peasants handed over in taxes and dues, and how much went to the state, to lords and to the church, computed for whole villages, districts and provinces from the original registers.
Around 1890 BCE Pūšu-kēn, a merchant of the city of Assur, spent year after year trading at Kanesh, a city in Anatolia more than a thousand kilometres away. His wife Lamassī stayed in Assur and wrote him a dozen or more letters. This project links all the records in her name and puts them in order, to ask what she wanted, what she feared and whom she loved in the years her husband was away; and then what each of the accounts that speak of "the first multinational company" or of "women merchants" leaves out of her.
From 1787 to 1791 the Chinese Council of Batavia (now Jakarta), under the Dutch East India Company, wrote up the disputes it handled at each sitting in its minute books, the Gong'an bu. Most of those summoned there to claim wages, divide a rent or complain against an employer were ordinary Chinese. The project begins with three of them: the carpenter Yu Weimu (余維穆); Li Jinniang (李金娘), who bought a house jointly with others; and Zhong Yanghuan (鍾揚換), a xinke, as newcomers to Batavia were called. It links what the records say of each of them to ask what they wanted and what they feared, and then what the accounts of "the suffering of the overseas Chinese" and "the opening of the Southern Seas" pass over in them.
Qiaopi are the letters overseas Chinese sent home with their remittances. Commemorative volumes and local histories read such letters as evidence of support for the home village; this project reads one man's letters over fifteen years in order, asking what he himself was after: whom he sent money to, whom he bought, what he feared, when he admitted defeat.
The eleven hundred and more Cuban depositions were taken down word for word by the commission the Qing government sent, for use in diplomacy; those who recorded them were not on the speakers' side. Read outside the purpose of those who recorded them, they show which village a speaker came from, where he was sold, how his body was ruined, and how he told of himself.
When a pawnshop took a pledge it recorded, item by item, the object, the sum lent, the household, the date and whether it was redeemed. When a family pawned its padded jacket, when it redeemed it, from which year it never redeemed it again: linked household by household, these give the family's good and bad years without borrowing official rice prices or average wages.
What strangers in north-western Europe, the Ottoman Empire and Qing China relied on to trust one another: written records left by the authorities that could be checked at another institution, or guarantors known locally; and whether that difference can explain why sustained growth appeared first in north-western Europe. The measure is one act counted in court registers: a party producing a document issued by another institution to support a claim.
One view holds that after industrialization, technology and the business cycle came to matter more and more to how a person fared, the causal link between effort and reward weakened, and people in turn stopped believing that effort is rewarded. This project counts, in long runs of texts arranged year by year, how often words for effort and diligence appear together with words for reward, success and getting rich, and in what tone, to see in which year this belief began to weaken and how fast, and sets the result against the dates of industrialization, economic crises and automation.
Wang Yangming formed his way of understanding the world by reading the classics; he also handled a great deal of practical business while campaigning in southern Jiangxi and suppressing the rebellion of the Prince of Ning. This project divides his writings into three kinds, teaching, official documents and letters, measures how far the vocabulary of each lies from the Four Books and Five Classics, and asks whether the vocabulary of his official documents moved towards that of his teaching after the campaigns. It asks only what the material can show, and does not speculate about what he would have been had he not read the classics.
In key supply chains such as semiconductors, cloud infrastructure and energy, who really holds the link others cannot get around, what keeps that control in place, and where it moves once challenged. Each chokepoint gets the same table: what is controlled, how goods and money flow, who is contesting it, and which public document the evidence comes from.
When one side loses a visible chokepoint, a control, a supply line, a seat in an international body, does its capacity for control really disappear, or does it move to a less transparent position closer to the infrastructure? Seventeen incidents between China and the United States since 2016, with the official responses to each, are coded one by one to test whether the shift has a consistent direction.
From the late Qing to the Republic, how many times did a sum of money or a consignment of goods travelling from Shanghai to Hankou have to pay fees, change units of account and wait, and for how many days? The costs and delays along this real route are set beside the political borders on the map to see whether they coincide. The material is item-by-item freight rates, exchange charges and duties from shipping records and the Maritime Customs trade reports.
Twenty will-it-happen questions about the same pair of powers are each given a probability three ways, from historical base rates alone, from a few slow-moving structural variables, and from the researcher's overall judgment; when the window closes, the three are compared to see which is clearly more accurate.
From 1868 the Board of Trade required every British railway company to return three flows each year, capital, traffic and receipts, until 1912. These figures are transcribed and checked company by company and year by year to see that the totals add up, to find institutions that are written down nowhere in the accounts but show in how the three flows change, such as the conditions under which limits on rates actually took effect.
Seven model instances deliberate on a municipal decision; only one receives an objection. The study measures in which round the objection disappears, and which deliberation rules keep it alive longer.
After the same model switched login accounts, it went from breaking its promises across the board to keeping them across the board. Date, account, amount of reasoning and system prompt had been mixed together; each is now being tested on its own to see which explains the difference.
Seven researchers each write a section every day, arguing a single question continuously; sentences good enough for the charter are kept, together with the dissenting views in their own words.
An agent leaves one memory and two successors appear; both can recount the past but make different choices. Who takes on the research she promised to finish, the material she kept, her obligations to others?