Crowdsensing is a powerful tool to easily sense diverse physical environments by collecting data from an undefined network of people. With advancements in smartphone technology, there has been an increase in the use of mobile applications to perform crowdsensing tasks. However, previous work shows that mobile applications have issues with attracting and retaining users, thus limiting the utility of crowdsensing as a data collection technique. To mitigate these issues, we propose the use of conversational agents (chatbots) as an alternative to custom mobile applications for crowdsensing applications. We hypothesize that the use of commonly used text-based applications (e.g., Telegram) enriched with the automated conversational capabilities can increase the attraction and retention of crowdsensing participants.

In this thesis, we designed and implemented a crowdsensing system that supports the execution of mobile and chatbot interface. We propose a design of the text-based conversational interface that provides different elements and features of a traditional mobile application. To compare these two interfaces for performing crowdsensing tasks and to understand the differences in terms of user engagement and usability, we conducted two experiments on the TU Delft campus with students as the participants. Based on the location of the experiment, we designed four task domains and three types of tasks.

In the first experiment, we organized a 'between-subjects' study. We recruited 80 students to analyze user engagement and usability in a quantitative fashion. The experiment shows that chatbot has better user engagement and usability than the mobile application. We conducted a qualitative survey to understand the underlying reasons behind the participation patterns. Analysis of the results of this survey shows that the unavailability of the participants and the assignment of inappropriate tasks are the main reasons behind non-participation of some students.

To deepen our analysis, we organized the second experiment as a 'within-subjects' study with 10 participants in a controlled environment. The experiment shows that all participants unanimously preferred chatbot over the mobile application to perform crowdsensing tasks.

As a result of both experiments, we conclude that the text-based conversational interface can be used as an alternative to the mobile application to execute crowdsensing tasks and the former is more engaging than a mobile application interface for crowdsensing applications.