Causal Bayesian Network for Bias Reduction in ML Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models often suffer from algorithmic bias, leading to discrimination in prediction outcomes, particularly when using data that correlates with protected attributes like race or gender, even when these attributes are not directly included in the model, resulting in disparate treatment and impact.
Innovation Solution
The approach models the supervised machine learning process as a causal Bayesian network, allowing for the use of new data sources while reducing bias by omitting or substituting protected attributes during the inference phase, and employing causal analysis to address discrimination across various dimensions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If data including information that could identify protected attributes is used to train the model, then predictive quality is improved, but algorithmic bias is introduced
Solution Approach 1:
The patent segments the feature set into protected attributes (which are excluded) and non-protected attributes (which are used for prediction). This segmentation allows the model to leverage predictive information from non-protected features while deliberately excluding features that could identify protected groups, thus resolving the contradiction between predictive quality and bias reduction.
Solution Approach 2:
The patent extracts and removes protected attribute information from the training data and model features. By taking out the harmful elements (protected attribute identifiers) while retaining the beneficial predictive features, the model achieves both good predictive performance and reduced algorithmic bias.
2Object-affected harmful factors
If protected attributes are omitted from the model, then algorithmic bias is reduced, but predictive quality may deteriorate
Solution Approach 1:
The patent introduces non-protected attributes as intermediary features that serve as proxies for the predictive information previously carried by protected attributes. These intermediary features maintain the model's predictive quality while avoiding direct reliance on protected attribute information, thus preventing algorithmic bias.
Solution Approach 2:
The patent changes the parameter selection criteria by shifting from using protected attributes to using non-protected attributes that correlate with the target outcome. This parameter substitution maintains predictive power while eliminating the harmful association with protected groups.
3Reliability
If causal analysis is employed to address discrimination, then fairness is improved, but model complexity increases
Solution Approach 1:
The patent performs preliminary causal analysis during the feature selection and model design phase to identify and exclude features that would lead to discrimination. By addressing fairness concerns upfront rather than attempting to correct them after model training, the approach maintains simplicity while ensuring fairness.
Data Source
AI summary
One or more default protected attribute values may be determined for a prediction model trained based on training data including a plurality of training observations. Each of the plurality of training observations may include a respective plurality of training data values corresponding with a plurality of features. Each of the plurality of training observations may also include a respective target value. Each of the training observations may include a respective protected attribute value corresponding with a protected attribute feature. A request to determine a designated predicted target value for a designated inference observation may be received after determining the one or more default protected attribute values. The predicted target value may be selected from two or more target values determined by applying the prediction model to an inference observation and potentially one or more default protected attribute values.