Explainable Reasoning Decisions With Certainty-Based Action Checks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning systems lack explainability, making it difficult to understand why specific outcomes are predicted, which is problematic when used for decision-making as there is no visibility into the reasoning process, and no way to trace back incorrect predictions.
Innovation Solution
The implementation of explainable and automated decision-making processes in computer-based reasoning systems, which determine suggested actions and provide explanation data, including certainty scores, similarity measures, and other metrics, to enhance transparency and accountability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning systems are used to predict outcomes based on input data, then prediction capability is improved, but explainability deteriorates as the systems provide no indication of why particular outcomes are predicted
Solution Approach 1:
The patent introduces an intermediary explanation module that acts as a mediator between the machine learning prediction system and the user. This module generates human-understandable explanations (such as feature importance scores, decision paths, or natural language interpretations) that bridge the gap between the black-box prediction mechanism and the user's need for understanding, without altering the core prediction algorithm.
Solution Approach 2:
The system segments the prediction process into distinct components: the core prediction engine and the explanation generation module. By dividing the overall system into these separate functional parts, the patent enables the prediction component to maintain high accuracy while the explanation component provides interpretability, thus resolving the contradiction between prediction capability and explainability.
2Productivity
If machine learning systems make automated decisions without explanation data, then decision speed is improved, but reliability deteriorates as there is no way to trace back incorrect predictions
Solution Approach 1:
The patent implements a feedback mechanism where explanation data is generated and provided alongside automated decisions. This feedback loop allows users to understand the reasoning behind predictions, trace back incorrect decisions through explanation analysis, and potentially provide corrective input, thereby maintaining fast automated decision-making while improving reliability and accountability through transparent reasoning trails.
3Loss of information
If explanation data is provided alongside predicted outcomes, then transparency is improved, but system complexity increases due to additional processing requirements
Solution Approach 1:
The patent enables the machine learning system to generate its own explanation data autonomously as part of the prediction process. The explanation module utilizes the same input data and internal model structures already present in the prediction system, allowing the system to self-document its decision-making process without requiring external explanation tools or significantly increasing overall system complexity.
Data Source
AI summary
The techniques herein include using an input context to determine a suggested action. One or more explanations may also be determined and returned along with the suggested action. The one or more explanations may include (i) one or more most similar cases to the suggested case (e.g., the case associated with the suggested action) and, optionally, a conviction score for each nearby cases; (ii) action probabilities, (iii) excluding cases and distances, (iv) archetype and/or counterfactual cases for the suggested action; (v) feature residuals; (vi) regional model complexity; (vii) fractional dimensionality; (viii) prediction conviction; (ix) feature prediction contribution; (x) conviction ratio; (xi) contribution ratio; and/or other measures such as the ones discussed herein, including certainty. In some embodiments, the explanation data may be used to determine whether to perform a suggested action.