This paper presents an end-to-end prototype for a dashboard that increases transparency and control over conversational AI systems. The system aims to make chatbots more transparent by revealing internal representations of users, such as age, gender, education level, and socioeconomic status. The dashboard allows users to see and modify these internal models, which can help them understand and control the chatbot's behavior. The system was tested with a user study, where participants interacted with the chatbot and provided feedback on the dashboard's usability and impact on their experience. The study found that users appreciated the dashboard, which provided insights into chatbot responses, raised awareness of biased behavior, and gave them control over the system. Participants also suggested future directions for both design and machine learning research. The dashboard was implemented as a web application using React and a Flask-based REST API, connecting to the LLaMa2Chat-13B model. The user study showed that the accuracy of the user model improved as conversations progressed, with an average accuracy of 78% across age, gender, and education after six turns of dialogue. Users found the dashboard engaging and useful for understanding chatbot responses, especially inappropriate or incorrect ones. However, some users expressed concerns about privacy and the potential for biased behavior. The dashboard also allowed users to control the chatbot's internal model of themselves, which helped them explore and mitigate biases. Users were able to adjust the chatbot's responses based on their demographic attributes, leading to more personalized and accurate answers. The study also highlighted the importance of user research in interpretability, as participants uncovered subtle types of biases around features such as socioeconomic status that were not anticipated. The paper concludes that the dashboard provides evidence of a design pathway toward more transparent and user-controlled AI systems. Future work includes exploring more detailed user models, addressing privacy concerns, and developing task-oriented dashboards for various AI applications. The user experience of the dashboard itself is also a rich area for further investigation, including how to handle sensitive user attributes and the potential for dashboards in voice-based or video-based systems.This paper presents an end-to-end prototype for a dashboard that increases transparency and control over conversational AI systems. The system aims to make chatbots more transparent by revealing internal representations of users, such as age, gender, education level, and socioeconomic status. The dashboard allows users to see and modify these internal models, which can help them understand and control the chatbot's behavior. The system was tested with a user study, where participants interacted with the chatbot and provided feedback on the dashboard's usability and impact on their experience. The study found that users appreciated the dashboard, which provided insights into chatbot responses, raised awareness of biased behavior, and gave them control over the system. Participants also suggested future directions for both design and machine learning research. The dashboard was implemented as a web application using React and a Flask-based REST API, connecting to the LLaMa2Chat-13B model. The user study showed that the accuracy of the user model improved as conversations progressed, with an average accuracy of 78% across age, gender, and education after six turns of dialogue. Users found the dashboard engaging and useful for understanding chatbot responses, especially inappropriate or incorrect ones. However, some users expressed concerns about privacy and the potential for biased behavior. The dashboard also allowed users to control the chatbot's internal model of themselves, which helped them explore and mitigate biases. Users were able to adjust the chatbot's responses based on their demographic attributes, leading to more personalized and accurate answers. The study also highlighted the importance of user research in interpretability, as participants uncovered subtle types of biases around features such as socioeconomic status that were not anticipated. The paper concludes that the dashboard provides evidence of a design pathway toward more transparent and user-controlled AI systems. Future work includes exploring more detailed user models, addressing privacy concerns, and developing task-oriented dashboards for various AI applications. The user experience of the dashboard itself is also a rich area for further investigation, including how to handle sensitive user attributes and the potential for dashboards in voice-based or video-based systems.