Experiment
Adaptive Environmental Services
Move from administering static scheme menus toward interventions that continuously learn from place, participation and outcomes.
01
Problem to address
National option menus can be easy to administer but poorly matched to local conditions. Evidence from delivery and outcomes does not always feed back quickly enough to improve the next intervention.
02
What the evidence tells us
Blanket approaches work in some places and not others. Menu-style choices can reduce environmental impact. Strategy teams still need to know whether interventions achieved what they intended.
03
The proposition to explore
The idea
Move from static scheme menus toward services that use local context and outcome learning to help people choose interventions more likely to work in their circumstances.
04
Hypothesis
If services continuously combine local context with evidence about what works, they can shift from offering static options toward helping people choose interventions more likely to contribute to outcomes.
05
How it could work
Connect land context, current agreements, environmental need and outcome evidence. Present relevant interventions with explanation and uncertainty. As outcomes are observed, update the evidence that informs future recommendations.
06
Example: bring it to life
Imagine that Defra understands:
- CATCHMENT / LANDSCAPE
- environmental outcome required
- current condition
- farms / parcels involved
- possible interventions
- local suitability
- collective intervention
- monitor
- adapt
Instead of asking:
“Which options from the national menu would you like?”
Defra and the farmer could eventually explore:
“These are the outcomes this landscape needs. Given what we collectively know about your land, here are the interventions most likely to contribute.”
That changes the organising principle from:
administering schemes
to:
achieving outcomes.
07
AI opportunity
Contextual recommendation, optimisation, evidence synthesis and adaptive targeting. AI should explain the basis for recommendations and surface uncertainty, not make opaque eligibility or funding decisions.
08
Potential value
- More place-sensitive interventions
- Potentially higher environmental return per pound
- Better user confidence in choices
- A service that improves as evidence accumulates
09
Reduce uncertainty without overcommitting
Smallest meaningful experiment
Start with one intervention where site-suitability evidence is reasonably understood. Compare generic guidance with an evidence-informed contextual view; measure choice quality and trust.
10
Open questions and risks
What we still need to learn
- Opaque targeting could undermine fairness
- Outcome evidence may be weak
- Policy may intentionally preserve user choice for reasons beyond optimisation