Crowdsourcing for social goals (e.g., supporting public libraries or people with disabilities) is a promising area. However, little is known about how to develop active worker communities for such goals. First, we need reliable metrics for the workers' motivation. Second, the characteristics of senior crowd workers have rarely been studied, even though they often play a primary role in social-purpose work. This work introduces a four-quadrant worker motivation model for social-purpose crowdsourcing and describes a system based on that model. Then we investigate the outcomes from the system's operations for six months, which involved both young and senior workers, seeking better ways to build an active community of crowd workers. We analyzed the workers' activities based on the system logs, conducted a survey, assessed the correlations between the subjective values and actual behaviors, and then discuss the implications.