Predictive Model Bias Analysis Using Node Scoring and Exclusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Predictive performance models often generate biased outcomes despite excluding protected data, as unethical or illegal biases can emerge during machine processing, particularly in neural networks or decision trees, leading to undesirable differences in user classifications.
Innovation Solution
A system and process that utilize a hardware processor to generate and refine artificial intelligence neural networks by analyzing input data, determining predictive biases, scoring nodes, and excluding biased compartments to produce a bias-free predictive model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If protected data is excluded from input to predictive models, then legal and ethical compliance is improved, but bias in model outputs still occurs due to unprotected data combinations
Solution Approach 1:
The patent segments the predictive model into multiple components including input data, processing layers, and output predictions. It separately analyzes each component for bias contributions, allowing identification of biased unprotected variables without including protected data in the model input
Solution Approach 2:
The patent introduces an intermediary bias analysis mechanism that examines the relationship between unprotected input variables and model outputs. This intermediary analysis identifies which unprotected variables are indirectly causing bias toward protected groups, enabling their removal or adjustment
2Productivity
If machine learning processes are used to generate predictions, then productivity and scalability are improved, but unethical differences and biases are introduced during processing
Solution Approach 1:
The patent performs preliminary bias analysis on input variables before they are processed by the machine learning model. By identifying and removing biased unprotected variables in advance, the system prevents bias introduction during the high-productivity machine processing phase
Solution Approach 2:
The patent implements a feedback mechanism where model predictions are analyzed for bias patterns, and this information feeds back into the variable selection process. The system continuously identifies and removes unprotected variables that contribute to biased outcomes, maintaining ethical standards while preserving productivity
3Measurement precision
If complex variable combinations are analyzed in neural networks, then prediction accuracy is improved, but hidden biases emerge in specific nodes and layers
Solution Approach 1:
The patent segments the neural network into individual layers and nodes, analyzing each component's contribution to bias. This segmentation allows detection of hidden biases in specific nodes that would be difficult to identify in the overall complex model
Solution Approach 2:
The patent introduces intermediary analysis variables that track the flow of information through neural network layers. These intermediaries help identify which unprotected variables are causing bias in hidden layers, making the detection process feasible despite model complexity
Data Source
AI summary
A hardware processor can receive a set of input data individually describing a particular asset associated with an entity. The hardware processor can receive a set of inputs individually responsive to a respective subset of a plurality of queries for a particular user. The hardware processor can generate a predictive model based on the set of input data. The hardware processor can calculate a predictive outcome for the particular user by applying the predictive model to the set of inputs. The hardware processor can identify a target score impacting the predictive outcome for the particular user. The hardware processor can assign a training program to the particular user corresponding to the target score.