Table 1 Comparison of three logics of experimentation
From: Understanding how city networks are leveraging climate action: experimentation through C40
| Controlled experimentation | Darwinian experimentation | Generative experimentation | |
|---|---|---|---|
| Characteristics |
-Search for valid inferences about cause and effect -Setting controlled as much as possible -Findings aim for external validity -Deductive |
-Oriented towards variation through many trials -Identifies ‘best practices’ but also expects many failures -Variation more important than control -Inductive |
-Iterative refinement of prototype with goal of ‘success’ -Discovery and design of new solutions -‘Success’ often depends on meeting stakeholder expectations -Abductive |
| Allowance for failure | High (researcher should not influence outcome) | Very high (few variations will be successful) | Low (researcher should strive for success) |
| Innovations vs. routine | Both | Both | Innovations |
| Observational vs. interventional | Intervention at the beginning | More observational than interventional | Continuous improvement of intervention |
| Examples |
-Randomised control trials -Natural and quasi-experiments |
-Parallel experimentation and benchmarking -Rapid experimentation -Simulation experiments |
-Design experiments -Exploratory pilot projects -Problem-driven iterative adaptation |