Understanding metro passenger’s travel plan choice behavior under unplanned service disruptions is vital for transit agencies. It allows capturing the changes in the demands of the passengers, adopting measures aimed at minimizing the impact on the transit system, and ensuring the safety of the interrupted passengers. Different from planned metro service disruptions, unplanned service disruptions cannot be known in advance and have high uncertainty. However, little is known regarding the role of the uncertain duration in the decision-making process and the taste heterogeneity in the behavior under unplanned metro service disruptions. To fill this gap, we first established a time interval in each scenario of a stated preference questionnaire to indicate the uncertain duration and conducted a web-based survey. Based on the survey data collected at Guangzhou, China, we developed an error component latent class model for travel plan choice behavior considering uncertainty and heterogeneity. The model result showed that the population can be classified into two classes, uncertainty pessimists and uncertainty optimists, who have a strong and weak perception of the uncertainty of the disruption duration, respectively, based on socio-demographic and travel characteristic attributes. Concurrently, the perception of uncertainty in both the classes is enhanced during peak hours and when a passenger enters a station. The findings of this study provide more insights into passenger’s travel behavior under unplanned service disruptions with uncertainty. Moreover, they can assist transit agencies in adopting effective management strategies, which, in turn, will aid in improving the service quality for its passengers.