Contributions published at Information Management (Gerhard Schwabe)

The work presented in this Bachelorís Thesis is part of a program, the Community-Based chronic disease Care Lesotho (ComBaCal), that aims to tackle Non-Communicable Diseases such as diabetes and high blood pressure in Lesotho, a landlocked country in southern Africa. One of the program's goals is to support local health workers with an eHealth application (eH-app), running on a mobile device, with clinical algorithms in providing culturally tailored, language-appropriate, and individualized care for a large number of people in rural villages in Lesotho. This Bachelorís Thesis is part of a larger project and placed in the early stages of requirements finding and development of the overall eH-app. The main goal of this thesis is to gain knowledge based on developing an exploratory prototype of the logic and functionality of a eHealth digital agent (eH-DA) to diagnose and treat diabetes. The knowledge gained from this prototype and the improved understanding of the underlying problem will lay the foundation for further work within the eH-app. The work involved the initial definition of the medical flows diagnosing and treating diabetes and the development of neighboring building blocks, in addition to the eH-DA prototype. So was also a Web App Server functionality integrated that allows adding new medical flows, and a graphical user interface for the eH-DA was added using a chatbot representation. The eH-DA prototype was built using a state machine to process medical flows defined in graphical diagrams that are translated into state diagrams represented by JSON files. The developed prototype was demonstrated to a focus group of 10 people representing five different organizations in Switzerland and Lesotho sponsoring the overall activities and evaluated to identify improvement areas and develop recommendations for the next phase of the overall development efforts.

Customer relationships are an essential part of many companies. Therefore companies are constantly looking for possibilities to build and improve these relationships. As a result of digitalization and increasing trade on the Internet, customers are increasingly expressing their opinions about products and services on online platforms in the form of reviews. Handling reviews is often too time-consuming for companies. As a result, many rely on service providers to analyze and respond to reviews for them. The response process is still done manually and no software support exists to support this process. In the context of this project, the review response process of a service provider in this area was analyzed and revised. Criteria and characteristics define the essential contents of a review response. These characteristics and criteria were weighted and combined in an overall evaluation model to obtain a final assessment of the quality of a response. This model was integrated into a tool that independently evaluates the quality of a response. This tool can be integrated into the response process to assist review authors with the evaluation of the quality. By evaluating the tool, it was shown that the quality assessment model can distinguish responses from experienced and inexperienced review authors. More than 95% of the responses were correctly classified. In addition, around 50% of beginner review authors felt supported by the tool, when it was evaluated in a real working environment.

A project by the Information Management Research Group at the University of Zurich is currently developing a digital assistant to support bank advisory services. Due to the novelty of this project, there is little known about the acceptance of such assistants in the context of bank advisory services. The aim of this thesis is, therefore, to define acceptance criteria for digital assistants in banking services. Based on a literature review and analysis of expert interviews, a selection of factors was made to be examined in a survey. These factors were namely the assistant's response time, error rate, transaction recognition, and manner. The results showed some influence of the selected criteria on the acceptance of digital assistants. However, based on statistical analysis, the results could not be attributed solely to the factors examined. The results suggest nevertheless, that these factors should be taken into account during development and must be the subject of further research.

Recent technological advancements in artificial intelligence led to the emergence of research on human-autonomy teaming. Research thereby identifies training as a critical success factor for human-autonomy teams. This thesis addresses the training process of such teams in the domain of advisory. The goal was to develop a specific training framework for a human-autonomy teaming project in the financial advisory sector, namely the project "Heinzelm‰nnli". To do so, the thesis uses design science research to develop two artifacts. The first artifact constitutes of an abstract training methodology and is based on current literature on human-autonomy teaming, end-user training, and service encounters, as well as on the insights of experts in these domains. Thereby, the abstract methodology provides an integrative perspective of the components that might affect training and its outcome in the domain of advisory and human-autonomy teaming. As a second artifact, the thesis presents a specific training framework and its application for the proof-of-concept phase of the project ìHeinzelm‰nnliî. The evaluation showed, based on qualitative data, that the specific training framework is valid and effective to empower advisors for the conversational agent-based assistant ìHeinzelm‰nnliî. In sum, this thesis presents a specific training framework for project "Heinzelm‰nnli" that is based on the development of an abstract training methodology to empower advisors for human-autonomy teaming.

In an environment that is plagued by distrust and intransparency, such as the used car market, blockchain-based certified vehicle data offers a compelling solution to these issues. One particular manifestation of such blockchain-based car data is the cardossier project. While previous research has examined the possibilities and effects of cardossier data, the role of commercial used car dealers has been largely unexplored. Thus, this work is dedicated to the role of used car dealers in a market with cardossier data. Through a design science approach, the role of dealers was incorporated into an existing experimental online market game. Subsequently, the developed software artifact was used for a market simulation experiment. The results thereof are a conceptual abstraction of used car dealers into the domain of an experimental market game, the implementation of said abstraction as well as insights into the effects of blockchain-based certified car data onto a used car market environment, that contains not only sellers and buyers but also dealers.

Banks have started to integrate Digital Assistants (DAs) more and more into their advisory sessions. To realize their full potential, several design aspects have to be considered, one of which is a DA’s degree of anthropomorphism. Since the banking sector is especially characterized by the importance of trust, this thesis aims to better understand the influence anthropomorphism on this metric. For this, we employ a mixed-methods approach. We let people fill out a survey and conduct interviews in which they evaluate three DAs exhibiting different degrees of anthropomorphism. The results show that anthropomorphism only has an overall minimal positive effect on trust, validating the prevailing perspective in current research. However, the qualitative data suggest that for at least a share of people, the degree of anthropomorphism plays a more decisive role when it comes to perceptions of trust. In the future, banks should consider utilizing a DA only in advisory sessions with customers that fit a specific set of characteristics and even flexibly employ various assistants with different degrees of anthropomorphism.

Conversational Agents have aroused great interest in recent years. Software with improving capabilities driven by advances in machine learning and natural language processing that interacts with users through natural language provides new opportunities for people to interact and participate ubiquitously. Although conversational agents have the ability to automate and innovate a range of activities in different areas, such as in customer service or marketing and sales, there is still limited shared knowledge about guidance for the design for conversational agents in the financial industry. This study aims to determine design features that a conversational agent should have in order to be utilized in a bank investment advisory environment and to be accepted by its potential users such as the investment advisors and their customers. To address this gap we present our design science research study on principles for a design of conversational agents. Based on a systematic literature review to identify an initial set of design features of conversational agents from existing research, twelve user interviews, and five workshops we created a list of functional and non-functional requirements which we integrated into eight design principles. While the design principles were being established we evaluated them in an iterative manner based on five evaluation dimensions (Importance, Usefulness, Memorable, Practicability, and Unambiguousness) with an expert-based evaluation. In order to also validate the design principles, storyboards were created which enabled us to analyze the design principles in more detail. Our findings supplement prior research’s design characteristics and play a significant role in the development of conversational agents in the financial industry. The requirements we developed can be useful to engineers, software architects, and developers.

null

The BCCG was developed for the cantonal administration of Zürich to assess the eligibility of blockchain for use cases in the Swiss public sector in order to stimulate the adoption of blockchain technology. Therefore, this thesis aimed to evaluate the BCCG, assess how well it supports the public sector, and identify potential improvements. By assessing 11 potential blockchain use cases with employees of the Swiss public sector, it was found that the BCCG is a practical and time-efficient method, for it provides a structured assessment. However, it was identified that improvements can be made concerning the provision of greater clarity and support for the future implementation of blockchain projects.

To make the WISE survey of the Center for Dental Medicine of the University of Zurich more attractive for patients a chatbot was developed. It assists patients during the participation in the survey and guides them through it. The chatbot was evaluated for its hedonic quality and the findings showed that the WISE survey with chatbot was more attractive for participants and made them feel more motivaed. The design and implementation of the chatbot as well as the implementation and results of the survey are discussed in this thesis.

Banks need a lot of information from their customers in order to offer them the best possible services and solutions. The Hypothekarbank Lenzburg and their employees, especially customer consultants, experienced difficulties in this information collecting process in the past. Therefore, in this thesis, three new digital prototypes were created in Figma and designed to facilitate the information collection from bank customers. The prototypes focus on usability, usefulness and further on disturbing the consultant-customer meeting as little as possible. In evaluation, customer consultants were impressed by the features the prototypes have to offer and by the simplicity in which they are designed. Additionally, users made some improvement suggestions for further development.

The used car market is characterized by information asymmetries and mistrust. Blockchain technology promises to resolve these problems using a system which stores data over the life cycle of a vehicle. However, while blockchain technology is strong in preserving the stored information, sense-making of this information is still essential to bring value to end consumers of the system. In this paper, we take an exploratory approach and create a prototype, which is then evaluated in realistic car sale conversations between buyers and sellers. We demonstrate and discuss how the interplay of different design elements of an application, built on top of a blockchain-based platform, helps to reduce information asymmetries and enhance trust. Our findings suggest that though providing more information about a used product (a car) leads to fewer information asymmetries in general, a reputation mechanism and data analysis are both beneficial in improving the situation further. As for trust, such a system enhances trust between buyers and sellers and, in general, makes the overall purchase process more trustworthy. However, to achieve these positive effects, the quality of the stored information should be guaranteed and properly communicated to the end-user.

The used-car market struggles with missing trust, lack of transparency and information asymmetries. To address these shortcomings, the cardossier project was started which records events along the life cycle of a car. Until now, the role of used-car dealers was hardly examined in this context. To learn more about the effects of cardossier on used-car dealers, this thesis integrates them into a market simulation platform and simulates the used-car market. The obtained data confirms the increase of transparency on the market. For the business of used-car dealers, it turned out that their role is still important on the market and that although more information is available, dealer warranties are still important for customers.

This thesis deals with the conception and technical implementation of the WISE-survey with an integrated chatbot. The previous WISE-survey, which is used by the Center of Dental Medicine for the individual assessment of orofacial pain and temporomandibular disorders, served as a basis. Patients of the Center of Dental Medicine rate the previous WISE-survey as too long and detailed. In addition, it offers too few options to address their individual complaints. By integrating the chatbot, which interacts with the users to answer mostly open questions and accompanies them through the survey, an increase in the intention to use and the user experience should be achieved. To counteract the problems of the previous survey, the chatbot interprets the patients' statements and asks more precise follow-up questions if necessary. After the implementation of the WISE-survey with chatbot, an evaluation with twelve participants was conducted. The participants had to complete the WISE-survey with and without chatbot and answer questions about the perceived user experience and the intention to use. The results of the evaluation show that the WISE-survey with chatbot led to an increase in the intention to use and the surveyed user experience scales (perspicuity, dependability, novelty, output quality) compared to the previous WISE survey without chatbot.

Digital assistants (DA) have so far almost not been used in bank service encounters. Therefore, it has not been investigated whether known theories on the anthropomorphizing of a DA can also be applied to DA in bank service encounters. This paper investigates quantitatively (N=76) and qualitatively (N=19) what influence the anthropomorphizing of a DA can have on satisfaction and service quality in a bank service encounter. It was shown that if the anthropomorphizing of a DA does not please the client, that this has a negative influence on the variables. On the other hand, no positive influence of anthropomorphizing could be quantitatively demonstrated. However, it was found that trust to the DA has a great influence on satisfaction and service quality.

With increasing technological progress, the number of customer feedbacks to be answered grows. To facilitate the response process for the authors, different artificial intelligence can be used as support. This bachelor thesis investigates how the interaction between humans and artificial intelligence can be supported with the help of a graphical user interface. For this purpose, the thesis examines how design rules in combination with human-AI principles can be used to enable a more efficient response procedure. To reduce the information load in the interface, an algorithm that prioritises the displayed information in their order, considering the review to be answered is designed. Furthermore, it will be investigated how authors can be sufficiently informed about the quality of the response during and after the process. Quality assurance is also presented in the graphical user interface, taking design rules into account. From these questions, it is possible to define different goals for the graphical user interface. As a result, an interface is developed and a prioritisation system is designed, which implements the questions described above. Finally, an evaluation is carried out in which the newly designed interface is compared with the existing one. It is possible to show that the authors can work more efficiently when using the new interface and are 43.75% faster in answering reviews.

Current documentation practices of financial service providers are mainly designed to comply with existing regulations and neglect the clients' need for individualized written information. Clients are often given only standardized product information or a concise enumeration of the topics covered. This lack of client specific information compels clients to make important investment decisions based on standard documents and explanations given only verbally. The increasing prevalence of digital assistants creates new opportunities to personalize advisory documentation and automate its creation. This paper explores what types of content and functionality are best for client advisory documentation. In several design iterations, prototypes have been developed and evaluated. The results show that clients want to receive a summary of the most important outcomes from the advisory session in order to reflect on the advice provided. Advisors also see great potential in documentation that is specifically customized to the clients, but are already heavily burdened with administrative tasks due to demanding regulations. Based on the findings of interviews and user tests, design implications for the provision of advisory documentation as well as suggestions for improvement and ideas for further development of the prototype are elaborated. The application of these results will contribute to a high-quality, personalized, and compliant advisory documentation which satisfies both clients and advisors.

Patients in psychotherapy currently do not have an app that supports processes between therapy sessions. This Bachelor’s Thesis aims to develop an early prototype of a digital agent called AkThi that supports intersession activities of patients undergoing psychotherapeutic treatment. For this purpose, a digital agent as part of the AkThiv app was developed together with the Department of Psychology UZH. Twelve subjects participated in the evaluation, which was conducted using semi-structured interviews and a standardised survey. Our analysis suggests that an interaction with AkThi should be based on established and disease-relevant questionnaires and a perceived personality of positive traits. Essential functions of such a digital agent are the embodiment, tutorial, regular mood and symptom monitoring, reflecting on activities and therapy content, and full integration with other functions of the AkThiv app. The evaluation participants identified different approaches for establishing contact with the therapist in crisis situations that could be implemented in future versions.

The purpose of this thesis is the design and development of an app-prototype which supports intersession activities of patients undergoing psychotherapy. Intersession processes can include thoughts and feelings that patients experience between therapy sessions. To accomplish this, important requirements of the app are selected in an iterative process involving existing literature and feedback of experts in psychology. Suitable technologies and architectures for the implementation are selected in accordance with the gathered requirements. As a result, a fully functioning mobile app-prototype is developed and eventually evaluated. The participants provide initial feedback and generally show a high intention to use the app. The findings of the evaluation suggest functionalities, design principles, and potential challenges of selected aspects of the app. Initial insights and foundations for future development result.

The popularity of the Internet and the possibility for consumers to share their opinions and experiences with products and services online has a significant influence on consumer behavior. Increasing numbers of consumers are using online consumer reviews to make their purchase decisions. This consumer trend has attracted the attention of companies that want to make profitable use of this information, which increasingly leads them to new challenges. In this thesis, online customer feedback management in the hospitality industry was examined. The aim of this work is to show the challenges of online customer feedback management from the perspective of restaurant managers and examine the extent to which those challenges can be addressed with artificial intelligence. For this purpose, interviews were conducted with managers from different restaurant businesses. The results show that AI-based IT solutions can provide support for both the analysis of and response to consumer reviews. However, AI support is limited by the lack of available information, the characteristics of the review sites, and the current values and norms of society.