Improving AI Development
As a user experience researcher at Facebook/Meta, my primary focus was on understanding the AI development process to improve the models for better efficiency, harm detection, and product recommendations. In this case study, I will share my experience of conducting user research to enhance AI models' performance.
(Note: this case study is intentionally vague given the nature of the work. Feel free to contact me to discuss further.)
Context
Facebook/Meta is a social media giant that uses AI to personalize content for its users. The AI models have to work efficiently while maintaining privacy and user safety. Hence, my team was tasked with understanding the development process to make the models better.
Research Goal
Our primary research goal was to identify the challenges in the AI development process and to come up with solutions to improve it. We aimed to create a seamless development process that would result in better models for users.
Research Process
Our research process consisted of several stages that helped us to gain insights into the AI development process. The following are the stages we undertook:
Desk Research: The first stage of our research process involved desk research, where we reviewed existing literature on AI development. We studied the best practices in the industry, the latest advancements, and the challenges faced by other companies.
Interviews: Next, we conducted interviews with the AI development team to understand the process's nuances. We aimed to understand the development process from start to finish, identify the bottlenecks and challenges, and identify areas for improvement.
Surveys: To gather data from a wider audience, we conducted surveys with the users of Facebook/Meta. We aimed to understand their expectations and experiences with the current AI models. We also asked for their suggestions to improve the models.
Data Analysis: After gathering the data, we analyzed it to identify patterns, trends, and insights. We grouped the data into themes and identified the key pain points in the development process. We also identified the areas that needed improvement based on the user's feedback.
Recommendations
Based on our analysis, we made recommendations to improve the AI development process. We suggested changes to the process to make it more efficient, less resource-intensive, and better at detecting harm or recommending products to customers. We also suggested improvements in the models' user experience to enhance the user's experience and protect their privacy.
Results
Our research resulted in several key recommendations that were implemented in the AI development process. The development team adopted a more collaborative approach to the development process, which improved the efficiency and effectiveness of the models. We also recommended changes in the user interface to make it more user-friendly and protect their privacy.
Conclusion
Our research helped to enhance the AI development process at Facebook/Meta. We identified the pain points in the development process and suggested improvements to make the models better. Our research resulted in the implementation of several changes, which improved the models' efficiency, safety, and user experience. Overall, our research helped to improve the AI models and provided a better experience for users.