Machine Learning Model for Detecting Erroneous AI Decisions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Detecting erroneous artificial intelligence (AI) decisions is technically difficult due to the 'black box' nature of AI systems, where inputs and outputs are known but the logic used to achieve outputs is unknown, leading to inefficient use of computing resources in understanding or reverse engineering AI decision-making processes.
Innovation Solution
A system utilizing machine learning models trained on first information related to AI usage and second information from historical decisions to determine if an AI decision is erroneous, with the ability to add complaint information to a blockchain for data-driven improvement of AI accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are used to detect erroneous AI decisions, then detection accuracy is improved, but computing resource consumption increases
Solution Approach 1:
The system pre-trains machine learning models offline using historical AI decision data and complaint information stored on blockchain. This preliminary training allows the models to be ready for deployment without consuming computing resources during runtime detection, as the heavy computational work has already been performed in advance.
Solution Approach 2:
The system creates simplified copies of AI decision patterns by training machine learning models to replicate the decision-making behavior. Instead of reverse-engineering complex original AI systems, the patent uses trained models that copy the essential decision patterns, enabling efficient error detection without requiring equivalent computational resources to the original AI systems.
2Reliability
If computing resources are allocated to understand AI decision-making logic, then error detection capability is improved, but resource efficiency deteriorates
Solution Approach 1:
The patent replaces the mechanical approach of reverse-engineering and understanding AI decision-making logic with a data-driven machine learning approach. Instead of allocating resources to analyze and comprehend complex AI logic, the system trains models on input-output pairs from historical decisions, substituting computational analysis with pattern recognition that requires fewer resources.
Solution Approach 2:
The system enables machine learning models to automatically learn error patterns from historical data stored on blockchain without requiring continuous human expertise or complex analytical resources. The models self-improve by processing complaint information and historical decisions, reducing the need for ongoing computational resources dedicated to understanding AI logic.
3Reliability
If machine learning models are trained on historical data, then detection reliability is improved, but data processing time increases
Solution Approach 1:
The system performs comprehensive training of machine learning models on historical AI decision data and complaint information stored on blockchain before deployment. This preliminary action ensures models are fully trained and ready for immediate use, eliminating the need for real-time data processing during error detection operations.
Solution Approach 2:
The patent trains machine learning models on specific local patterns from historical data relevant to particular AI decision contexts. By focusing training on relevant local patterns rather than processing all historical data continuously, the system achieves high detection reliability while minimizing data processing time during operational phases.
Data Source
AI summary
In some implementations, a device may identify a use of artificial intelligence by an entity to reach a decision in connection with a user. The device may determine, using a machine learning model, that the decision in connection with the user is erroneous. The machine learning model may be trained to determine whether the decision is erroneous based on first information relating to the use of artificial intelligence by the entity and second information relating to one or more historical decisions in connection with the user or one or more other users. The device may provide a notification indicating that the decision in connection with the user is erroneous.