This study aims to examine the effect of dynamically aligning the conversation style to the user’s preference in a conversational agent environment. We suggest a technique to identify the user’s preferred conversation style, by scoring statements based on their writing style. We designed a within-subject experiment intended to measure the user engagement and satisfaction (with style alignment × without style alignment). We found an increase in self reported satisfaction ratings and reflect on how familiarity with the entity behind the chatbot may affect the preference of the conversational style. We finish this study with concrete suggestions to designers and developers of future chatbots and how should the alignment of conversation style be assigned and adapted

With the increase of anthropomorphic conversational agents, the research on the effects of anthropomorphism in conversational agents is on the rise. However the studies cannot seem to agree on many of the raised questions and the research on the proper design of conversational agents is still lacking. This research paper attempted to contribute to filling this gap by analysing whether anthropomorphic visual cues used in conversational agents have an effect on the trust and satisfaction of the users. In particular, a between-subject experiment was performed, where the use of emojis and a profile image of four different levels of anthropomorphism were manipulated in a conversational agent based on Telegram. A total of 120 participants, split between eight experimental groups, had a conversation with the agent and filled in a survey about their experience. The results concluded that individual visual cues as well as combination of them did not have any significant effects on the trust and satisfaction of the users.

The rise in the use of crowd computing platforms led to the birth of Dandelion, a conversational crowd computing platform developed at TU Delft with the main goals being to connect students with researchers and to allow students to report on their well-being by using a friendly interface. Dandelion was tested manually up to the time of drafting this paper; thus, the primary motivation behind this paper is to ensure the robustness and measure the responsiveness of Dandelion.

Robustness was exercised by utilizing a simulated user behaving unexpectedly. The testing framework then classifies the behaviour of Dandelion according to the C.R.A.S.H. scale. The testing framework is validated by altering Dandelion's behaviour and ensuring that the test results reflect the change. Furthermore, a lower bound to the run time of a task will be estimated using a Multi-Agent System (M.A.S.) simulation on Dandelion.

Upon verifying the correctness of the robustness test, a faulty assumption was uncovered on which the user's input validation was based. Furthermore, the M.A.S. simulation run estimated a lower bound of $\approx 5.788$ seconds, while revealing a lack of user's input validation before posting them in the database.

Conversational agents are being used more often. Many conversational agents have been designed to make them more humanlike to increase the engagement of the user. This research is looking into a couple of aspects of anthropomorphism to determine whether these aspects improve the engagement of the users. These aspects are typos with time delay and empathy. 3 conversational agents were created for this research, one control group and one for both aspects. For each conversational agent a group of volunteers tested the bots and filled in a survey after the conversation. While the conversational agent with typos and time delay didn’t show a significant change, the conversational agent with empathy did. When testing the control conversational agent and the conversational agent with typos and time delay multiple participants told that they felt ignored when the conversational agent doesn’t show empathy, this implies that not having empathy has negative effects on the conversation. Therefore giving the conversational agent empathy results a better engagement of the user.

With the growing popularity of conversational interfaces, new use-cases in which conversation may be an effective medium for human-computer interaction are explored. Recent works have shown how conversational interfaces can improve worker engagement in micro-task crowdsourcing and that task performance is comparable to web-based interfaces. This research explores how conversation workflows can be designed with micro-breaks to effectively support task execution in conversational crowdsourcing. A database model is proposed that supports conversation workflows with micro-breaks.
Furthermore, two experiments were held with 10 participants each to evaluate how micro-breaks affect worker engagement, preferences and task performance. The results show that break proposals by the conversational system slightly increase worker engagement, task quality, and task duration. Moreover, it was found that workers take less breaks than they prefer to and that break preferences vary among workers.

Crowd-mapping is a relatively new field of research, which involves the collection of geographic data by a crowd of workers. The collection of said data is of great importance for organizations like municipalities, where it is used for applications such as maintaining streets and greenery. The benefit of crowd-mapping over traditional mapping methods, where workers physically observe the area, is that it has the potential to be far more cost-effective and time-efficient. As this should not come at the cost of losing accuracy, research needs to be done on how to effectively map objects in a city.

Although previous work has focused on mapping urban objects using street-level imagery, they are all specifically aimed at a single type of object. Furthermore, they do not offer a general method for geo-location estimation and do not estimate the height of the objects. All of the systems designed in previous work only support task execution using a web platform. As crowd-mapping is nothing without the crowd, it is important to keep the workers engaged. No research had been done on how the task execution platform and type of task could affect the worker engagement and satisfaction.

In this thesis we will design a system for crowd-mapping urban objects using street-level imagery. We will propose novel methods for geo-location and object height estimation. Experimentation showed that the proposed geo-location method was able to deliver an accuracy with up to 83% of the estimations being within 2.5 meters of the ground truth with a mean distance of 1.85 meters. The height estimation showed up to 85% of the estimations being within 30 centimeters from the ground truth with a mean difference of 15 centimeters. Furthermore, the system supports task execution on three platforms; web, mobile and mobile virtual reality. We demonstrated the feasibility of executing mapping-, data-enrichment- and verification tasks on each of these platforms. Experimentation with the different platforms showed that the type of task and execution platform affects user engagement, cognitive load, satisfaction and execution time.

Knowledge Base Construction (KBC) is a challenging and complex task involving several substeps and many experts of a knowledge domain. Crowdsourcing approach has been used to support KBC with promising scalability and output quality, but to enable even more people to participate in KBC, there is a need to broaden the pool of workers beyond the ones who are already familiar with existing crowdsourcing platforms. Meanwhile, the number of people who use messaging platforms has been increasing. There is also a renowned popularity of text-based conversational agents -- chatbots -- existing on these messaging platforms. By leveraging the fact that there is a large number of users who are familiar with conversational interfaces, we see an opportunity to broaden the participants of crowdsourced KBC by using chatbots to execute KBC tasks. In this thesis, we investigate the use of chatbots to enable crowdsourced construction of knowledge bases. We design a conversational crowdsourcing platform to support the execution of KBC tasks. An experiment involving 43 students using our system and interviews with 7 participants were conducted to evaluate the system within the context of constructing a knowledge base for the TU Delft campus. From the results, we show that the platform is suitable to be used for crowdsourced KBC with justifiable execution time, accuracy, and completeness. We also laid out the potential future work to improve and extend the functionalities of the chatbot system.