Blockchain Verification for Machine Learning Anti-Discrimination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning algorithms are susceptible to discrimination and bias, making it difficult to verify compliance with nondiscrimination policies, especially since these systems are considered 'black box' and lack direct influence mechanisms for correcting biases.
Innovation Solution
Implementing a corrective module within machine learning models that uses a blockchain network to record and verify compliance by generating a blockchain transaction data structure including the model's state, input data, and correction indications, ensuring that decisions adhere to nondiscrimination policies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If machine learning models are used to make automated decisions, then productivity and efficiency are improved, but discrimination and bias may occur making verification of compliance difficult
Solution Approach 1:
The patent introduces blockchain technology as an intermediary layer between the machine learning model and the verification process. The blockchain records immutable proofs of model behavior, training data characteristics, and compliance verification results, allowing third parties to verify nondiscrimination compliance without accessing or understanding the complex internal workings of the ML model. This mediator resolves the contradiction by enabling verification while preserving the model's operational autonomy and efficiency.
Solution Approach 2:
The patent creates cryptographic copies and proofs of the machine learning model's state, training data characteristics, and decision-making process. These copies are stored on the blockchain as verifiable records that allow external verification of compliance without requiring access to the actual model or its internal logic. The copying principle enables transparency and verification while maintaining the model's operational integrity and efficiency.
2Reliability
If corrective algorithms are applied to remove discrimination, then nondiscrimination compliance is improved, but system complexity increases
Solution Approach 1:
The patent segments the compliance verification system into distinct modular components: the machine learning model itself, the corrective algorithms, the blockchain infrastructure, and the verification interface. Each component has a specific function and can be independently developed, tested, and maintained. This segmentation reduces system complexity by allowing each part to be optimized separately while working together to achieve overall compliance.
Solution Approach 2:
The blockchain serves as an intermediary layer that manages the complexity of coordinating corrective algorithms and verification processes. It provides a standardized interface for recording and verifying compliance information, reducing the complexity burden on both the ML model and the verification system. The mediator absorbs and manages the computational and architectural complexity of the compliance ecosystem.
3Reliability
If blockchain technology is used to record model state and corrections, then transparency and accountability are improved, but computational overhead and processing time increase
Solution Approach 1:
The patent implements preliminary action by recording proofs of model state, training data characteristics, and compliance verification results on the blockchain in advance, before actual decisions are challenged or audited. This proactive recording eliminates the need for time-consuming real-time verification computations. When verification is needed, the pre-recorded proofs can be quickly retrieved and validated, significantly reducing processing time while maintaining full transparency.
Solution Approach 2:
The system creates lightweight cryptographic copies and proofs of the model's state and behavior that can be quickly verified on the blockchain without requiring access to or computation on the actual model. These copies are designed for efficient verification, allowing rapid validation of compliance claims while maintaining full transparency. The copying approach reduces computational overhead compared to re-running or re-analyzing the actual model.
Data Source
AI summary
Disclosed are systems and method for machine learning and blockchain-based anti-discrimination validation. The described techniques uses a machine learning model to generate a numerical determination associated with a first person based on an input data set associated with the first person. The numerical determination is further based on a corrective module configured to compensate for prohibited discrimination by the machine learning model. The technique generates a blockchain transaction data structure comprising a state of the machine learning model at the time of generating the numerical determination, a copy of the input data set associated with the person, and an indication of a correction by the machine learning model. The blockchain transaction data structure is recorded or published in a blockchain network.