These tools and metrics are designed to help AI actors develop and use trustworthy AI systems and applications that respect human rights and are fair, transparent, explainable, robust, secure and safe.
Robust and Reliable Algorithmic Recourse (ROAR)
ROAR is a technical framework for generating algorithmic recourse, actionable recommendations that tell an individual what changes would reverse an unfavourable automated decision (e.g. denied loan or rejected application) that remain valid even when the underlying predictive model is later updated or retained. It addresses a known reliability gap in existing counterfactual explanation methods, which typically assume that deployed model stays fixed and can therefore produce recourse that is silently invalidated by routine updates (e.g. due to data corrections or geospatial validation).
The framework formulates recourse generation as a minimax optimisation problem. Instead of minimising the cost of a counterfactual change under a single fixed model, it minimises the worst-case loss over a bounded set of plausible future model shifts (perturbations to model parameters). It solves this using an adversarial-training-inspired iterative algorithm. The method is designed for linear models and extends to non-linenar models (e.g. deep neural networks) via local linear approximation techniques.
Auto-discovered on 2026-08-26 by OECD Catalogue Automation
About the tool
You can click on the links to see the associated tools
Tool type(s):
Objective(s):
Purpose(s):
Target sector(s):
Lifecycle stage(s):
Type of approach:
Maturity:
Usage rights:
License:
Target groups:
Target users:
Stakeholder group:
Geographical scope:
Risk management stage(s):
Technology platforms:
Use Cases
Would you like to submit a use case for this tool?
If you have used this tool, we would love to know more about your experience.
Add use case
Partnership on AI
