Masterarbeit
Enhancing a Security Assessment Tool with Semantic Knowledge Graphs and Interactive Conversational Interfaces
Completion
2026/03
Research Area
Students
Shariful Islam
Advisers
Verena Traubinger M.Sc.
Dr.-Ing. Sebastian Heil
Description
Chatbots are increasingly used by public services as a communication channel for citizens. Especially with the higher use of large language models, it is necessary to evaluate how governmental chatbots can make their reasoning transparent and provide information about possible resources. Trust in governmental chatbots is important as they are often used to gain access to official data or even legislature. This thesis aims to systematically investigate and categorize transparency options in governmental chatbots, with a particular focus on how reasoning and sources can be presented to end users.
This includes: (1) analyzing the state of the art in chatbot design and transparency in digital government, (2) describing a list of transparency mechanisms in governmental use cases, (3) developing prototype chatbots that implement these mechanisms or a subset of them, and (4) conducting a structured user evaluation on citizens’ preferences. From the analysis and evaluation results, guidelines or best strategies should be formulated.
The objective of this thesis is the creation of a solution or the combination of existing approaches to solve the problem described above of exploring transparency strategies in governmental chatbots. This comprises the following parts. An analysis of the state of the art on governmental web chatbots, trust in chatbots (including LLMs), current chatbot technologies, and other relevant work. The prototypical implementation should include the design and coding of different transparency options for the user study. A quantitative user study should be conducted where the prototypes are compared with each other. The user study and the prototype have to be based and evaluated on their compliance with elicited requirements based on the literature research.
