AI Model Performance vs Explainability
Apply risk-based AI governance requiring explainability and mandatory human review for high-impact citizen decisions while permitting greater automation elsewhere.
CyberTRIZ analysis · EGovernment contradiction TDC011 · one of 8,235 worked contradictions published by CyberTRIZ.AI
Regulations
Business Context
Advanced artificial intelligence models improve fraud detection, predictive analytics, document processing, policy analysis, and citizen service automation. Highly sophisticated models often deliver greater predictive accuracy and operational efficiency.
As model complexity increases, however, decision-making becomes more difficult to explain. Government organizations must demonstrate transparency, fairness, and accountability, particularly when AI influences decisions affecting citizens' rights or access to public services.
The Contradiction
More sophisticated AI models improve predictive performance.
Simpler AI models improve explainability and public accountability.
Why the Contradiction Exists
Complex machine learning algorithms frequently achieve greater accuracy while reducing the ability of humans to understand how decisions are produced.
e-GovernmentTRIZ Analysis
Governments should balance predictive performance with explainability according to decision impact. High-risk decisions require explainable AI, transparent governance, and human review, while lower-risk operational activities may benefit from more advanced automation.
Recommended e-GovernmentTRIZ Principles
Principle 23 – Feedback
Principle 24 – Intermediary
Principle 28 – Mechanics Substitution
Principle 35 – Parameter Changes
Practical Resolution
Deploy explainable AI frameworks, model documentation, human review processes, continuous monitoring, and risk-based governance for all government AI systems.
Expected Benefits
Greater public trust
Better regulatory compliance
Improved AI governance
Higher decision quality
Responsible innovation
Reduced ethical risk