Merge pull request #211892 from jcodella/patch-4

Updated transparency bullet point

Author: prmerger-automator[bot]

articles/cognitive-services/personalizer/responsible-guidance-integration.md

@@ -45,7 +45,7 @@ When you get ready to integrate and responsibly use AI-powered products or featu
 
 - **User Study**: Any consent or disclosure recommendations should be framed in a user study. Evaluate the first and continuous-use experience with a representative sample of the community to validate that the design choices lead to effective disclosure. Conduct user research with 10-20 community members (affected stakeholders) to evaluate their comprehension of the information and to determine if their expectations are met.
 
-- **Transparency**: Consider providing users with information about how the content was personalized. For example, you can give your users a button labeled Why These Suggestions? that shows which top features of the user and actions played a role in producing the Personalizer results.  
+- **Transparency & Explainability:** Consider enabling and using Personalizer's [inference explainability](https://learn.microsoft.com/azure/cognitive-services/personalizer/concepts-features?branch=main#inference-explainability) capability to better understand which features play a significant role in Personalizer's decision choice in each Rank call. This capability empowers you to provide your users with transparency regarding how their data played a role in producing the recommended best action. For example, you can give your users a button labeled "Why These Suggestions?" that shows which top features played a role in producing the Personalizer results. This information can also be used to better understand what data attributes about your users, contexts, and actions are working in favor of Personalizer's choice of best action, which are working against it, and which may have little or no effect. This capability can also provide insights about your user segments and help you identify and address potential biases.
 
 - **Adversarial use**: consider establishing a process to detect and act on malicious manipulation. There are actors that will take advantage of machine learning and AI systems' ability to learn from their environment. With coordinated attacks, they can artificially fake patterns of behavior that shift the data and AI models toward their goals. If your use of Personalizer could influence important choices, make sure you have the appropriate means to detect and mitigate these types of attacks in place.