Utilizing Prompt Caching to Improve User Experience with AI Chatbots; Understanding Urban Impacts and Adoption Factors of Autonomous Vehicles

As artificial intelligence continues to integrate more into our daily workflows, the usability/accessibility of AI systems significantly influences their effectiveness and adoption. My capstone project aimed to enhance the usability of a business intelligence chatbot at MicroStrategy by implementing an autocomplete system. This project was driven by user reported issues surrounding user experience, response latency, and consistency with the chatbot. My STS research examined adoption factors and urban impacts of autonomous vehicles (AVs). With AVs being an emerging technology, it is important to understand adoption factors and urban impacts to maximize its benefits when implementing the technology. Both projects underscore that usability is a fundamental aspect of AI systems that significantly enhance effectiveness and adoption of the technology. Whether it’s a chatbot interface or AV infrastructure, it’s important to emphasize how we interface with the underlying AI system to maximize its benefits. MicroStrategy, a business intelligence software company, provides an AI chatbot that allows companies to analyze their data dashboards. However, users reported pain points concerning having to manually type repetitive questions, long response times, and inconsistent responses when interacting with the chatbot. To address these issues, I developed an autocomplete system using vector similarity search and prompt caching techniques. The autocomplete system made it so that as users type a query into the chatbot, historically semantically similar questions to what the user has typed would be suggested and could be autocomplete, leveraging past responses to improve response latency and consistency. The autocomplete system provided suggestions with roughly 300 millisecond latency and improved response times of historically similar questions by a factor of approximately 20. The system provided relevant suggestions and faster, more consistent responses. This project was the first time that data from a telemetry database that stored interactions with the product was used, showing how leveraging previous data could be a potential future direction in improving analytical capabilities and user experience with the product. My STS research seeks to better understand how autonomous vehicles impact urban dynamics, and what factors primarily contribute to accelerated adoption in certain cities like San Francisco. Because AVs are a relatively new and emerging technology, it is important to understand its impacts on urban dynamics to ensure that its implementation addresses existing challenges in urban environments rather than exacerbating them. It is also important to understand adoption factors because they influence how AVs are used, guide the development of the technology, and influence policymaking surrounding AVs. Through a literature review, analyzing case studies, and using Actor-Network Theory (ANT) as an analytical framework, my research provides an overview of adoption factors and impacts surrounding the current state of AV adoption. My research highlights that because AVs are still relatively new technology, its impacts on urban environments are unclear. While AVs have shown promise in things like reducing congestion and improving environmental sustainability, they also present potential challenges related to equity and urban sprawl. Using ANT as an analytical framework also suggests that adoption factors depend on the alignment of human and non-human actors within the network specific to a certain urban environment, so it is difficult to conclude that a select few factors in isolation primarily influence adoption outcomes more than others. Going forward, it is important to view AVs as sociotechnical systems rather than strictly technical systems so that cities can implement them in a way that maximizes benefits while also mitigating potential harms.