This paper discusses the challenges of creating user models for computers to treat users as individuals with distinct personalities and goals. It proposes using stereotypes as a mechanism to build user models based on limited information. The paper describes a system called Grundy that uses stereotypes to create user models and recommends books based on them. The system must resolve conflicts in uncertain knowledge and adapt stereotypes based on user experience. The paper analyzes Grundy's performance, showing that its user models effectively guide its behavior. It also discusses the importance of user models in improving system usability and performance. The paper outlines the three major dimensions of user models: whether they model a canonical user or individual users, whether they are explicitly constructed or inferred from behavior, and whether they contain short-term or long-term information. The paper focuses on individual user models that are implicitly constructed and long-term. It describes how stereotypes are used to build user models, how they are activated based on triggers, and how they are combined to form the User Synopsis (USS). The paper also discusses how the USS is used to guide the system's task of recommending books. The paper concludes that user models can serve various functions in a system, even a simple one like Grundy. It also discusses the limitations of the system in exploiting the USS and the importance of learning and adapting stereotypes based on user experience. The paper emphasizes the need for a hierarchical memory system to organize knowledge and the importance of adapting stereotypes to improve system performance.This paper discusses the challenges of creating user models for computers to treat users as individuals with distinct personalities and goals. It proposes using stereotypes as a mechanism to build user models based on limited information. The paper describes a system called Grundy that uses stereotypes to create user models and recommends books based on them. The system must resolve conflicts in uncertain knowledge and adapt stereotypes based on user experience. The paper analyzes Grundy's performance, showing that its user models effectively guide its behavior. It also discusses the importance of user models in improving system usability and performance. The paper outlines the three major dimensions of user models: whether they model a canonical user or individual users, whether they are explicitly constructed or inferred from behavior, and whether they contain short-term or long-term information. The paper focuses on individual user models that are implicitly constructed and long-term. It describes how stereotypes are used to build user models, how they are activated based on triggers, and how they are combined to form the User Synopsis (USS). The paper also discusses how the USS is used to guide the system's task of recommending books. The paper concludes that user models can serve various functions in a system, even a simple one like Grundy. It also discusses the limitations of the system in exploiting the USS and the importance of learning and adapting stereotypes based on user experience. The paper emphasizes the need for a hierarchical memory system to organize knowledge and the importance of adapting stereotypes to improve system performance.