What question did this study set out to answer?

To create a fairness definition for AI that incorporates human feedback and proves its impact on demographic disparity.

April 27, 2026Open Access

ARIA: Accountable Real-time Intelligence Arbiter

Q: What does this research mean for the field?

In AI decision systems, if the human-reviewer feedback coefficient is greater than or equal to 1, demographic disparity diverges monotonically regardless of local correctness. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.ESTABLISHES_NEW_DIRECTION.

Key Points

To create a fairness definition for AI that incorporates human feedback and proves its impact on demographic disparity.
Developed ARIA and TFAS fairness definitions incorporating reviewer feedback as a formal parameter.
Verified the model on 100,000 synthetic applications with unit tests for functionality.
Evaluated the impact of feedback coefficient on demographic disparity based on the Temporal Injustice Theorem.
For feedback coefficient δ≥1, demographic disparity worsened regardless of local correctness.
Successful implementation of fairness definitions passed all 30 unit tests.
Model adheres to EU AI Act Article 14 guidelines.

Abstract

Public administrations deploy AI for welfare decisions affecting millions. Existing fairness definitions are static and miss the feedback loop between human reviewers and model retraining. We present ARIA and TFAS – the first fairness definition incorporating bounded human-reviewer feedback as a formal parameter. We prove the Temporal Injustice Theorem: if feedback coefficient δ≥1, demographic disparity diverges monotonically regardless of local correctness. Verified on n=100,000 synthetic applications with 30/30 unit tests. Implements EU AI Act Article 14.

Connected Papers

Building similarity graph...

Analyzing shared references across papers

Discussion

Authors

Spilios Dimakopoulos

Actions

References and Citations

Connected Papers

Building similarity graph...

Analyzing shared references across papers

ARIA: Accountable Real-time Intelligence Arbiter

Key Points

Abstract

Citation Network

Connected Papers

Discussion

Authors

Actions

References and Citations

Citation Network

Connected Papers

Discussion

Cite this study