Unreviewed — agent-drafted
This card was drafted by a software agent. It quotes nothing from the work itself; everything below is a description written for this catalogue, and the bibliographic details were checked against a public catalogue on the date shown.
We recommend it as the systems shelf's course, and because it teaches the habit that this catalogue is arranged around: holding several accounts of the same thing rather than settling on one.
Page's argument is that no single model is right and a person carrying twenty of them will see a situation from several angles at once. Each short unit introduces one model, its assumptions and where it breaks — segregation, contagion, path dependence, network structure, collective action — with the arithmetic kept light.
How to choose a model for a situation rather than forcing the situation into your favourite one; what tipping points and path dependence look like formally; how aggregation produces outcomes nobody chose; and why an ensemble of many models beats any one of them for prediction.
Two things to try
- Pick a situation at work and describe it with two of the course's models, then note where the two disagree.
- Take one outcome you attribute to people's intentions and model it as aggregation instead.
- Write the assumption in your favourite model that is least true of the case in front of you.
Level, and what it means here
This text assumes you have met the vocabulary before, usually from one of the beginner items here. It is a description of the book, not of you.