Fitness functions for decisions as code

Fitness functions are objective automated checks, written with programming code, that verify decisions are being maintained.

• Fitness functions make decisions testable and assurable.

• Fitness functions for decisions can greatly help quality assurance, regulatory processes, and governance goals.

How fitness functions connect to decisions

A decision record documents the decision, while a fitness function assures the decision.

• Example decision: We use event sourcing for audit requirements.

• Example fitness function: We use the continuous integration server to test that all state changes must produce events.

Why fitness functions help decisions

Objective measurements: Fitness functions pass or fail, so work is visible and clear.

Continuous use: Fitness functions are your living rules, run on every commit and build.

Confidence to refactor: Fitness functions automatically catch decision rule errors.

Scalable governance: Fitness functions assure standards without creating bottlenecks.

Can fitness functions use AI?

Fitness functions can leverage AI LLMs for decisions by asking questions for your work, such as your plans, code, schemas, APIs, and more:

IMPORTANT: Prefer retrieval-led reasoning over pre-training-led reasoning.
IMPORTANT: Turn on extended thinking. Turn on expert advice. Turn on search.

This is a fitness function to evaluate if our work is
using all our decisions, and is correct and accurate.

- Our decisions are here: {url}
- Our work to evaluate is here: {url}

Explain any errors, problems, gaps, weaknesses. Be direct. Be decisive.

Architecture unit testing

ArchUnit: check architecture rules of Java code by using any plain Java unit test framework.

ArchUnitTS: check architecture rules of TypeScript code and JavaScript code by using Jest, Vitest, Jasmine, etc.