Why trust multiplies

Take a simple model. A task needs a group of people, and each pair in the group must be able to rely on each other. If any two people trust each other with probability t, a group of k people has k×(k−1)/2 pairs, and the whole group holds together only if every pair does. That number of pairs grows with the square of the group’s size, which is why large cooperation is rare.

Now suppose people do not need to trust each other one by one. Each only needs to rely on a shared record of signed commitments, verified deliveries and checkable observations. Then the group holds when each member can rely on the record, and the exponent grows only with the group’s size.

The model is deliberately crude. Real trust is not the same between every pair, groups do not simply hold or break, and records can themselves be forged. It is a hypothesis to test in simulation, not a law. But the direction fits the evidence: countries where people trust each other more have tended to grow faster, and studies of small groups have found a measurable collective intelligence that depends more on how evenly members take part than on how clever each one is.

It does not mean that people’s lives are visible. A community that tried to raise trust by watching its members would destroy the thing it is trying to build, because people would stop being honest with it.

If every member brings their judgement and a share of computing, the problems a group can take on grow with the size of the group that can hold together, and trust sets that size. A community with checkable records can pool more people and more computing on one problem, solve harder problems, and prove its improvements sooner.

The answer is not one rule for everyone. Any community can copy the shared rules, change one, and run. Open-source software shows this working: LibreOffice split from OpenOffice in 2010 and MariaDB from MySQL in 2009 when their communities lost confidence, and both are still maintained. The economist Albert Hirschman argued in 1970 that the freedom to leave keeps institutions honest.

Each community’s results would be recorded in a form others can compare: which rules hold larger groups together, which protect people with the least, which get the most done. Rules that work spread by being copied. Rules that fail are dropped. No one is forced onto a rule, and anyone can move to a community that suits them.

• We want thisLarge groups work together

  Communities with shared, checkable records routinely run projects with dozens of members.

  For example: forty households and their agents design, fund and install a neighbourhood heating network, each checking the others’ commitments on the shared record rather than vetting forty strangers.

• We want to avoid thisA forged record is trusted everywhere

  High reliance on a record that can be falsified multiplies harm instead of output.

  For example: a batch of forged delivery records lets one supplier claim hundreds of rewards before independent observers notice.

• We want to avoid thisTransparency turns into surveillance

  Communities try to raise trust by watching their members.

  For example: a housing co-operative asks members’ agents to share location history “for trust”, and half the members quietly stop using their agents.

• We want thisBetter rules spread

  Communities copy and change the rules, compare results openly, and what works spreads.

  For example: a community that slows how fast recognition for care work fades sees more people take part after a year, and nine others copy the change.

A simulation comparing trust pair by pair with a shared record, for groups of 5 to 100, with noise, forged records and members free to leave.

If a shared record does not at least double the largest group that holds together, we will say so and drop the claim.