Distributed Ledger Tracking for Explainable AI Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI models, particularly deep learning models, operate in an opaque manner, making it difficult for human programmers and systems to understand how they produce specific results, especially when unexpected outcomes occur, and there is a need to explain their processes and biases.
Innovation Solution
An XAI platform integrates with a distributed ledger to store records of each iteration of a deep learning model, including clustering and rationale information, allowing identification of anomalous outputs and initiating response actions to prevent further anomalies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If deep learning models are used to solve complex business problems, then problem-solving capability is improved, but transparency and understandability of model processes deteriorate
Solution Approach 1:
The patent introduces an intermediary system comprising a distributed ledger and explanation generation module that mediates between the opaque deep learning model and human users. The intermediary captures model states, generates human-readable explanations, and stores them on the distributed ledger, thereby making the model's decision-making process transparent without altering the model's core functionality.
Solution Approach 2:
The patent creates copies of the model's internal states and decision pathways and stores them on the distributed ledger. These copies serve as transparent representations that can be examined without affecting the original model's operation, allowing users to understand and verify model behavior while maintaining the model's problem-solving capabilities.
2Productivity
If AI models operate automatically without human intervention, then productivity is improved, but ability to identify and correct anomalies deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where the distributed ledger stores model states and explanations that can be analyzed to detect anomalies. The system continuously monitors model behavior, compares it against stored historical states, and generates alerts when deviations are detected, enabling automatic anomaly identification while maintaining high automation levels.
Solution Approach 2:
The patent performs preliminary actions by storing model states and explanations on the distributed ledger before anomalies occur. This creates a baseline of normal model behavior that can be used for comparison, enabling proactive anomaly detection and correction before problems impact business operations.
3Measurement precision
If iterative training processes are used to improve model accuracy, then model performance is improved, but complexity of tracking and auditing model changes deteriorates
Solution Approach 1:
The patent merges the model training process with the distributed ledger system, where each training iteration's state and explanation are automatically captured and stored on the ledger. This integration creates a unified system that tracks all model changes without requiring separate auditing mechanisms, reducing the complexity of tracking and auditing iterative training processes.
Data Source
AI summary
Aspects related to providing explainable artificial intelligence (XAI) using distributed ledger technology are provided. An XAI platform may train a first iteration of a deep learning model. The platform may store a record of the model to a distributed ledger. The platform may automatically perform an iterative recording process to generate additional iterations of the model and store records of the additional iterations of the model. During the process, the platform may receive outputs of the model. Based on the outputs, the platform may identify an anomalous output. The platform may identify one or more source iterations of the model that correspond to the anomalous output. The platform may initiate one or more response actions based on identifying the one or more source iterations of the model.