Responsible AI FAQ

  • What is the Responsible Business Practices bot?

    1. The Responsible Business Practices bot is a chatbot system that helps users find and discover information related to Microsoft’s efforts to promote responsible business practices in our value chain. It is a type of natural language processing system that can understand and respond to user queries in plain English. At a high level, the system takes a user query related to responsible business practices as input, such as “How does Microsoft conduct human rights due diligence?” or “In which countries are Microsoft’s suppliers located?”. The system then searches for relevant information from various publicly available sources, such as Microsoft’s Global Human Rights Statement and Supply Chain Integrity Statement. The system produces a natural language answer that provides the requested information as output. The system also provides links to more detailed resources if the user wants to learn more.

  • What can the Responsible Business Practices bot do? 

    1. It can help users find and discover information related to responsible sourcing in Microsoft’s supply chain and Microsoft’s human rights commitments.

    2. It can provide users with links to more detailed resources, such as publicly available reports, policies, and codes of conduct.

    3. It can answer common questions about responsible sourcing, such as “How many companies are in Microsoft’s supply chain?” or “From how many countries does Microsoft source materials?”.

  • What are the Responsible Business Practices bot’s intended use(s)?

    1. The bot is intended to provide users with easy and convenient access to information related to responsible sourcing in Microsoft’s supply chain and Microsoft’s human rights commitments.  

  • How was the Responsible Business Practices bot evaluated? What metrics are used to measure performance?

    1. Our technical and subject matter experts used manual red teaming to evaluate the performance of the chatbot as well as to mitigate harm.

    2. The results show that the bot performs well across different types of user queries, such as how-to questions, feature questions, and problem questions. The system is able to provide accurate and relevant answers for most of the user queries, as well as links to more detailed resources if the user wants to learn more.

    3. The results suggest that the bot is a robust and reliable system that can handle a variety of user queries related to responsible business practices. However, the results are not fully generalizable across use cases that were not part of the evaluation, as there may be some limitations and challenges that the system may face in real-world scenarios, such as:

      • The system may encounter user queries that are ambiguous, incomplete, or out of scope, which may require more clarification or feedback from the user or the system.

      • The system may not have the most up-to-date or comprehensive information for some user queries, as the online documentation that the bot uses may not cover all aspects of every one of Microsoft’s commitments or may not reflect the latest changes or updates.

      • The system may face some technical issues or errors, such as network latency, server downtime, or system bugs, which may affect the system’s performance or availability.

  • What are the limitations of the Responsible Business Practices bot? How can users minimize the impact of the bot’s limitations when using the system?

    1. The bot is not designed or evaluated for user queries that are not related to responsible business practices, such as queries about general topics, personal matters, or other Microsoft products, services, and policies that are not covered by the system’s knowledge base. The system may not be able to provide accurate or relevant answers for such queries or may provide generic or default responses instead.

    2. Like any AI system, it is not perfect and may sometimes produce incorrect answers. This is called hallucination, which means that the system generates something that is not based on reality or evidence, but rather on its own internal logic or errors.

    3. To minimize the impact of the bot’s limitations when using the system, users can take the following steps:

      • Users can use simple and clear language when asking queries related to responsible business practices, and avoid using jargon, slang, or abbreviations that the system may not recognize or understand.

      • Users can provide enough information and context when asking queries related to responsible business practices and avoid leaving out important details or assumptions that the system may not infer or guess.

      • Users can verify the accuracy and relevance of the answers provided by the system and check the links to more detailed resources if they want to learn more or confirm the information.

      • Users can report any errors, issues, or feedback that they encounter when using the system, and help the system improve its performance and functionality over time.

  • What operational factors and settings allow for effective and responsible use of the bot?

    1. The system is expected to perform reliably and safely within the domain of responsible business practices, which means that the user queries should be related to Microsoft’s responsible sourcing and human rights commitments, and the system should provide information that is accurate, relevant, and helpful for users.

    2. The system is also expected to perform reliably and safely within the scope of natural language processing, which means that the user queries should be expressed in plain English, and the system should generate natural language answers that are clear, coherent, and engaging for users.