Need help in fixing ethical bias simulation
Hi,
I was able to understand Ethical bias simulation testing. Now after running the bias simulation it was confirmed that there is a bias. But how do I fix the bias? Is this feature only limited to track the bias?
For example: There are two scenarios in my solution.
Scenario 1 (Direct bias): I have a bias field called Gender which is used in the adaptive model. After running the simulation, I found that there is a bias shift on this. How do I fix this issue? Is this feature only limited to generate a simulation report? or Do we have any way to fix this one in business strategy or adaptive models?
Scenario 2 (Indirect bias): I have another field called Age which is not used in the adaptive model. Bias simulation predicts there is a bias shift based on the report. As it was not used in my adaptive model how do i fix this issues?
Do we have any ways to fix the above issues? Please let me know. Thanks in advance.
***Edited by Moderator Marissa to update Platform Capability tags****
Accepted Solution
Updated: 29 Apr 2020 3:00 EDT
Pegasystems Inc.
NL
Hi, Measuring bias is a good first step. If there is direct bias as a result of using "protected" fields (fields you shouldnt be using according to business or legal policies) the steps are obvious - remove that field from your strategy logic and models (note bias testing is not just about models). If the bias is indirect it is harder to influence - a lot of things will be correlated with Age or Gender so just removing those won't be a guarantee there is no bias. Some bias may be acceptable - and perhaps inevitable - but if you need to reduce it, you will have to change the way the arbitration is done in your strategy logic, e.g. by boosting the priority of (some) offers for some groups.
Pegasystems Inc.
US
Please see this link for possible steps: https://community.pega.com/knowledgebase/articles/decision-management/84/complying-policies-or-regulations-detecting-unwanted-bias. You would have to follow this link for details: https://community.pega.com/knowledgebase/documents/pega-marketing-84-implementation-guide
Tech Mahindra
IN
Thanks for the reply. I was able to do this.
"Understand how to properly measure unwanted bias in offering products to your customers to comply with your company policies and regulations."
But how to fix the bias issue? Should I change the adaptive models or remove predictors and run it again? Or is there any way to fix the bias issues in the adaptive models? Thanks in advance.
Accepted Solution
Updated: 29 Apr 2020 3:00 EDT
Pegasystems Inc.
NL
Hi, Measuring bias is a good first step. If there is direct bias as a result of using "protected" fields (fields you shouldnt be using according to business or legal policies) the steps are obvious - remove that field from your strategy logic and models (note bias testing is not just about models). If the bias is indirect it is harder to influence - a lot of things will be correlated with Age or Gender so just removing those won't be a guarantee there is no bias. Some bias may be acceptable - and perhaps inevitable - but if you need to reduce it, you will have to change the way the arbitration is done in your strategy logic, e.g. by boosting the priority of (some) offers for some groups.