Experiment
Synthetic Policy Lab
Use evidence-grounded synthetic actors to expose likely questions, tensions and hypotheses before research with real people.
01
Problem to address
Policy teams may not see how a proposal could land across very different actor groups until late consultation or delivery. Real research remains essential but can be better focused.
02
What the evidence tells us
Farmers, field officers, regulators, analysts and policy teams have genuinely different concerns. Those differences can be used to challenge a proposition before choosing what to research.
03
The proposition to explore
The idea
Use evidence-grounded synthetic perspectives as a pre-research challenge tool to expose tensions and generate better hypotheses for research with real people.
04
Hypothesis
If synthetic actors grounded in real evidence are used as a pre-research challenge tool, teams will identify more relevant questions and recruit real participants around the highest-risk assumptions.
05
How it could work
Create constrained personas from verified research themes. Present the same proposal to several synthetic perspectives, compare tensions and extract hypotheses. Label every output as synthetic and trace it to the source evidence used.
06
Example: bring it to life
A proposed long-term land commitment prompts a synthetic tenant farmer to question authority to commit, while a field officer persona raises verification burden. Those questions become hypotheses for real interviews, not findings.
07
AI opportunity
LLMs role-play constrained perspectives using supplied research. Outputs are hypothesis generators only and must never be represented as participant evidence.
08
Potential value
- Broader early challenge
- Faster research planning
- Reduced blind spots in policy design
- Better use of existing qualitative evidence
09
Reduce uncertainty without overcommitting
Smallest meaningful experiment
Use the lab to plan one upcoming research round, then compare the synthetic hypotheses with what real participants actually say. Measure which hypotheses were useful, wrong or missing.
10
Open questions and risks
What we still need to learn
- Can create false confidence or stereotypes
- Must not replace user research
- Source evidence can be too narrow to support a persona