Master of Sc. Christoph Hein

UX Research & Design | AI & LLM Interaction

About Me & My Research

Welcome to my interactive portfolio. I am looking for opportunities to design and influence future interactions with AI and LLMs. I achieved a master degree in Human Computer Interaction in 2025. For the past few years, I've been absolutely captivated by the world of AI and LLM interactions. How make them as useful, accessible, and visible as possible for users! What novel areas to explore, such as deploying them on live streaming sites to let viewers interact with them, facilitating new opportunities for creative engagements. Therefore I created this site to showcase and update and link recent projects and ideas. Ask the LLM currently supported with RAG about my projects and scientifc work or check my CV and other information.

Research Assistant

Suggested Questions:

Professional Experience

08/2024 - 04/2025

Interactive Technologies Institute (Portugal)

Research Assistant & Master Thesis

  • Accomplished writing a scientific article about Prompting Dog-Centered Perspectives: Investigating Anthropocentrism in LLM. This was done by comparing data from different interviews with output data from prompt engineered LLMs with methods such as knowledge graphs. During this internship I had one supervisor for feedback.
  • Master thesis written in cooperation with ITI under the title LLMs as Narrative Facilitators: Successes and Limitations for Multi-user Prompt Fusion for Creating More-than-Human Narratives. For this purpose I developed an LLM interface with Twitch integration and measured its performances in a user study with metrics such as game experience questionnaire, interviews and a quantative analysis of the output and chain of thought. In the meantime I also layed the foundation for a LLM selection method: "Hein's Quick Model Choice" using the LLM as judges paradigm paired with human evaluation to define the most suitable LLMs for my work.
10/2023 - 01/2024

Edag Engineering Group AG

Working Student - Human Factors

  • Design & concept for FlexiSteer: Accomplishing the design of relevant functionalities in Figma in several versions, clickable prototype to present the concept and expanding functionalities during consultation with the project manager.
  • Other works involved redesigning a customer application in close exchange with the developers, creating icons for different projects and using LimeSurvey for questionnaires in usability tests.
2022 - 08/2023

BREDEX GmbH

Working Student - UX/UI Department

  • I supported the conception and design of a cutomer project: Accomplished creating screens as low fidelity prototypes as well as in Figma with a few details, in close exchange with the project manager and UX designer.
  • For the quarterly workshops in the design apartment, I was responsible for 50 percent of the workshop content, which included creating slides and tasks for participants, as well as giving a successful presentation.
2021 - 12/2021

BREDEX GmbH

Mandatory Internship & Bachelor Thesis - UX/UI Department

  • During my Bachelor thesis in cooperation with BREDEX: Design and Prototypical Implemantation of an ITR (Interactive Text Response) chatbot, a conversational interface was created and evaluated in a user study.
  • Accomplished creating a questionnaire with evaluation matrix based on customer effort scores for BREDEX, in consultation with the CEO, for customer centricity measurements used in future projects. This involved holding multiple workshops with different company departments on Miro and evaluating data in order to create a set of questions and guidelines for how and when to use the survey.
2020

Complero GmbH

12-week Internship

  • Accomplished concept to market launch of a dashboard for customers
  • I worked out requirements, developed early prototypes and designed the interface & supported the implementation

Education

10/2022 - 06/2025

Bauhaus Universität Weimar

M.Sc. Human Computer Interaction (taught in english)

09/2016 - 12/2021

Hochschule Kaiserslautern University of Applied Sciences

B.Sc. Digital Media Marketing

Skills

Design Tools

  • Adobe Creative Cloud
  • Figma
  • Balsamiq
  • Miro
  • Adobe XD
  • DaVinci Resolve

Software & Development

  • MS Office 365
  • Sharepoint
  • Java Script
  • HTML
  • Python
  • Flask
  • GitHub
  • LimeSurvey
  • Cloudflare & Render
  • Twitch & OBS

User Experience

  • Interviews
  • Usability Studies
  • Focus Groups
  • Requiements Engineering
  • Field Research
  • Surveys
  • Personas
  • Journey Maps
  • Prototyping
  • Human Centered Design
  • Heuristic Evaluation
  • A/B Testing
  • Design Thinking

Selected Scientific Works

LLMs as Narrative Facilitators: Successes and Limitations for Multi-user Prompt Fusion for Creating More-than-Human Narratives

Hein, C. (2025) | Master Thesis

This study investigates the integration of LLMs into Audience Participation Games to mediate collaborative storytelling on Twitch, addressing the lack of frameworks for equitable multi-user prompting. A mixed-methods approach is used to evaluate the LLM as a game agent and record the user experience...

This study investigates the integration of LLMs into Audience Participation Games to mediate collaborative storytelling on Twitch, addressing the lack of frameworks for equitable multi-user prompting. A mixed-methods approach is used to evaluate the LLM as a game agent and record the user experience. This approach combines qualitative interviews, game expereince questionnaire surveys, and exact match analysis. This process is supposed to provide in sights that contribute to a better understanding of the user experience when LLMs/GPTs are integrated into gaming contexts. The findings indicate that, while the used LLM (Qwq) effectively merged numerous inputs into a cohesive narrative, it exhibited an "early-input bias," preferring initial contributions and thereby compromising inclusivity in later turns. This resulted in a decrease of more than 50 percent in the EM score by the fifth turn. Participants reported moderate engagement and low perceived competence, suggesting deficiencies in user experience design concerning visibility and feedback. Positively there was next to no tension and frustrations recorded. Mutual influence while prompting the LLM to create the narrative received moderate ratings, indi cating a moderate impact. Additionally, the LLM’s adherence to more than human perspectives was noted as inconsistent with anthropocentric framing appearing in a few occasions. However, humor increased enjoyment and was partially successful combined with critiques of human ecological impact. Over all, the study is able to contribute to human-AI collaboration in multi-user contexts, but the LLM instructed as a game agent is not able to fulfill the task sufficiently all the time. Due to limitations such as small sample sizes and reliance on prompts to facilitate the agent, the thesis explores more the feasibility of creating narrations with users remotely providing multiple inputs in a prompt.

Keywords: LLM Interaction Prompt Fusion Live Streaming Audience Participation Games
View Full Thesis

Prompting Dog-Centered Perspectives: Investigating Anthropocentrism in LLMs

Interactive Technologies Institute (2024) | unpublished

This study investigates the anthropocentric bias in large language models when prompted to describe canine behavior when reacting or perceiving certain stimuli such as "food" or "friend". In particular, we compare three different prompt strategies: dog-centred perspective, an anthropocentric-debiased perspective, and a control without any bias...

This study investigates the anthropocentric bias in large language models when prompted to describe canine behavior when reacting or perceiving certain stimuli such as "food" or "friend". In particular, we compare three different prompt strategies: dog-centred perspective, an anthropocentric-debiased perspective, and a control without any bias. The outputs were evaluated against three baseline data sources: Video analyses of dog behavior, interviews with dog owners, and interviews with canine behavior experts. Findings suggest that the default prompt produced the most anthropocentric responses, while the anthropocentric-debiased prompt aligned more closely with expert perspectives. The dog-centric prompt, though aiming for a canine viewpoint, often anthropomorphized dogs through human-like language.

Keywords: Anthropocentric Bias More-Than-Human Canine Behavior Prompt Engineering
View Full Paper

Understanding Public Transportation Abroad: A Qualitative Inquiry Using Semi-Structured Interviews

Bauhaus University Weimar (2024) | unpublished

In addressing urban mobility challenges, mitigating environmental impacts, and participating in daily life activities, public transport (PT) systems play a critical role. This qualitative study examines the key factors influencing PT experiences abroad and explores potential implications for improving PT in Germany...

In addressing urban mobility challenges, mitigating environmental impacts, and participating in daily life activities, public transport (PT) systems play a critical role. This qualitative study examines the key factors influencing PT experiences abroad and explores potential implications for improving PT in Germany. Through semi-structured interviews with PT users who have lived or spent time abroad, the study examines various aspects of PT use, including accessibility, reliability, ticketing, control, and their commuting experiences. Findings reveal common themes such as the importance of flexible fare structures and reliable information dissemination to enable seamless travel. Participants emphasize the need for user-centered approaches and innovative solutions to improve the usability and effectiveness of public transport. Recommendations include the adoption of contactless payment methods, the introduction of express services, and the expansion of PT coverage in underserved areas with alternative PT modes. Further research is recommended to explore additional or more specific modal shift strategies and user experiences with alternative PT to evaluate their impact on PT use and environmental sustainability.

Keywords: Public transport Interviews Improvements
View Full Paper

User-Centered Design Approaches of a Color-Based Communication System for Emotional Support of Employees as a Ubiquitous Computing Solution

Bauhaus University Weimar (2023) | unpublished

This paper focuses on a possible ubiquitous computing solution that provides for: transporting emotions among employees through the transmission of colors in order to motivate and support each other. The efforts explained in the paper are based on already researched and user tested proposals to extend the communication of individuals with colors...

This paper focuses on a possible ubiquitous computing solution that provides for: transporting emotions among employees through the transmission of colors in order to motivate and support each other. The efforts explained in the paper are based on already researched and user tested proposals to extend the communication of individuals with colors. With this knowledge, user-centered design is explained and used in a theoretical approach to identify the strengths and weaknesses of the application. The overall goal of this paper is to come to a conclusion on the desirability of implementation by exploring UCD methods and a theoretical application in the described home office scenario.

Keywords: Ubiquitous Computing User-Centered-Design Communication with Color
View Full Paper

Projects

Coming Soon: More LLM studies on Twitch

AI Project | N/A

In the future I will conduct more studies on Twitch with LLMs, to explore how they can be used to facilitate more-than-human narratives and other creative engagements.

Stay tuned for updates on upcoming studies and projects involving LLMs on Twitch. These studies will aim to explore the potential of LLMs in enhancing user engagement and creating innovative interactive experiences. Further I want to explore how viewers can argue with a LLM to convince it pressing a symbolic button. In an other future project I will target developing a gaming experience in which a GEN is used to create live new contents such as bosses in a video game based on the viewers input. This will allow for a dynamic and engaging gaming experience where the LLM adapts to the players' actions and decisions in real-time.

Stay Tuned for more information and updates on these exciting projects!

Exploring the Realism of AI-Generated Onomatopoeia: An Interactive Demo

AI Project | 2024

Onomatopoeia guessing Interface

The project investigates the ability of GEN, specifically text-to-speech (TTS) systems, to produce natural-sounding onomatopoeia, and evaluates the human ability to distinguish these from real human-made sounds.

Onomatopoeia guessing Interface

This paper explores the comparison of human and AI-generated onomatopoeia within an interactive demo exhibited at 'summaery' 2024 at Bauhaus university Weimar. The project investigates the ability of deep learning models, specifically text-to-speech (TTS) systems, to produce natural-sounding onomatopoeia, and evaluates the human ability to distinguish these from real human-made sounds. Using open-source datasets and models such as AudioLDM2 and ElevenLabs' monolingual model, the demo involved participants listening to and voting on various sounds categorised as farm animals, instruments and domestic noises. The results show a significant challenge for participants in accurately identifying AI-generated sounds, with a near 50/50 split between correct and incorrect votes, highlighting both the advances in AI audio generation and the limitations of human recognition capabilities.

Key Words:
  • Interactive Demo
  • Measurements of Performance
  • Generation of Onomatopoeia
  • Voice Cloning

Social Engineering - The Adventure!

AI/Social Engineering Project | 2023/24

Interface of the game showing a scene in an office with multiple choices

This semester project was accomplished in a diverse team with backgrounds in different disciplines. My contributions are mostly about level and interaction design, creative writing and storytelling, as well in the written parts. The report is a summary of the project "Social Engineering - The Adventure" that was driven by seven Bachelor and Master students as well as three supervisors. Over the course of a semester, the team was dedicated to exploring the techniques of social engineering while developing the prototype of a game in the end.

Interface of the game showing a scene in an office with multiple choices

This report is a summary of the project "Social Engineering - The Adventure" that was driven by seven Bachelor and Master students as well as three supervisors. Over the course of a semester, the team was dedicated to exploring the techniques of social engineering while developing the prototype of a game in the end. The goal of the game is to educate the average internet user on the dangers of social engineering and the techniques used by attackers to access someone’s personal data. The game is designed such that the player takes on the role of the attacker, slowly uncovering a political scandal. The player is asked to gather information on parties involved such as companies and politicians. The story progresses until the scandal is revealed. By providing a choice for different actions, players learn more about different social engineering techniques, making them more aware of its dangers and possibly spread awareness by sharing this knowledge with their social environment. This can be seen as an additional goal, extending and supporting employee education in the public and private sector on privacy and security measures, possibly outperforming other mediums such as educational videos. For the prototype, the story, graphics and programming mini-games were implemented to make the game both educational and fun. At the current state, the story is developed until level 4. The game engine used for development is Ren’Py, and the code can be found on GitHub. The game is mainly story based with the player being able to choose actions to take next. Some of those options will definitely lead to the access of information, some won’t and could perhaps lead to the player getting caught. This means that the game creates a tree of choices for the player to make with some branches leading to a positive and some to a negative outcome.

Key Words:
  • Game Development
  • Social Engineering
  • Group Project
  • Level Design

Air-Balanced Bites

Research Project | 2023

This project examines the need for the physicalization of abstract data using inflatables in order to make it more accessible and palpable. Our project aims to allow users to build their usual meal and see how it meets the proposed standards.

In our data-driven society, data representation is a crucial problem that affects decision-making in many different sectors. This project examines the need for the physicalization of abstract data using inflatables in order to make it more accessible and palpable. Our project aims to allow users to build their usual meal and see how it meets the proposed standards. Also, the user is given the opportunity to change their choice in the process by removing or adding certain products, which helps them think about possible changes in their diet and makes their further choice more conscious. In our case, inflatables act as a pneumatic system that pushes out the table legs at indicators above or below the norm, which makes it possible to clearly see the result of their choices. The table's four inflated components, which correlate to important nutritional indicators and dynamically display meal balance, provide a useful and memorable way to comprehend dietary choices. The study describes the theoretical underpinnings, driving forces, and design methodology of "Air-Balanced Bites," an interactive dining table that physically depicts the nutritional balance of meals in real time. In this TEAM-project my tasks and contributions were mostly about ideation, prototyping and building of the final version, while my team mates were more concerned with programming and the design

Key Words:
  • Interactive Data Visualization
  • Inflatables
  • Adulino
  • Balanced Meal Preparation

Comparison of two signal forms for a take-over request (TOR) in the context of autonomous driving

Research Project | 2019

Onomatopoeia guessing Interface

The vibration signal and the acoustic signal for the “Take over Request” are examined by evaluating reaction times and the deviation from the route traveled. In addition, a preference of the test person is determined with a subsequent questionnaire.

Onomatopoeia guessing Interface

Autonomous driving and the vision of a safe, efficient and convenient mobility concept still have a few hurdles to overcome, but are steadily making their way into our everyday lives (Sokolov, 2015). As reported by Der Spiegel, the behavior and functionality of autonomous vehicles in city traffic is already being tested on a test track in Hamburg (cfr, 2019). Semi-autonomous systems (Advanced Driver Assistance Systems) such as lane departure warning systems in cars and trucks are just one example of the advanced technology in the mobility sector. Over the last 40 years, systems such as anti-lock braking system (ABS), electronic stability program (ESP) and adaptive cruise control (ACC) have become established, which initially support the driver in the longitudinal and increasingly also in the lateral guidance of the vehicle (Beiker, 2015, p. 200; Johanning, Mildner, 2015, p. 61). The underlying scenario of autonomous level four is that the driver remains capable of driving at all times and takes control of the vehicle in the event of an unexpected situation. For this reason, the following question is investigated in this thesis in a study using a driving simulator: Which signal enables a safe and reliable transfer of control of the vehicle, even if the driver is no longer in the “loop” due to another activity? This means that the driver is no longer actively participating in the driving process. The vibration signal and the acoustic signal for the “take over request” are examined on the basis of an evaluation of reaction times and the deviation from the route traveled. In addition, a preference of the test person is determined with a subsequent questionnaire. Finally, the results are presented and a recommendation for future studies is made.

Key Words:
  • Autonomous Driving
  • User Study
  • Take over Request

Get In Touch

Email Inquiry

For research collaborations or professional opportunities

Send Email