Graph Computing Hybrid Reasoning for Explainable AI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AI systems, particularly deep learning and neural networks, lack explainability in their output generation, making it difficult for users to understand the reasoning behind their predictions, which is a critical issue in applications like security, medical decisions, and vehicle control where transparency is essential.
Innovation Solution
The implementation of graph computing using a processor that generates a graph with explicit and implicit nodes, allowing for hybrid reasoning by combining deductive and inductive reasoning, enabling traversals between nodes to provide explainable predictions and improved transparency in AI decision-making processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning or neural networks are used for AI systems, then prediction accuracy is improved, but explainability of the system output deteriorates
Solution Approach 1:
The patent introduces an intermediary layer between the neural network and the output that generates explanations. This intermediary translates the internal representations and decision processes of the neural network into human-understandable formats, allowing both high prediction accuracy and explainability to coexist. The intermediary acts as a mediator that preserves the predictive power of deep learning while making the reasoning process transparent to users.
2Productivity
If AI systems operate as black boxes, then computational efficiency is improved, but trustworthiness in critical applications deteriorates
Solution Approach 1:
The patent segments the AI system into distinct components: the neural network for prediction and the explanation generation module for transparency. This segmentation allows the neural network to operate efficiently as a black box for computational tasks, while the separate explanation module provides trustworthiness by making the decision process visible. The segmentation enables both computational efficiency and reliability to be maintained in different parts of the system.
3Loss of information
If traditional reasoning systems are used, then explainability is improved, but prediction accuracy deteriorates
Solution Approach 1:
The patent merges two previously separate systems: the neural network (providing high prediction accuracy) and the traditional reasoning/explanation system (providing explainability). By combining these systems into a unified architecture where the neural network's internal states are mapped to symbolic representations that can be explained using traditional reasoning frameworks, the patent achieves both high prediction accuracy and explainability simultaneously. The merging allows the strengths of both approaches to complement each other.
Data Source
AI summary
Embodiments for graph computing are provided. A graph including a plurality explicit nodes and at least one implicit node is generated. A first of the plurality of explicit nodes and a second of the plurality of explicit nodes are traversed between utilizing deductive reasoning. A third of the plurality of explicit nodes and a fourth of the plurality of explicit nodes are traversed between through the at least one implicit node utilizing inductive reasoning.