Topics for doctoral, master’s and bachelor’s theses

The scope, methodological requirements and depth of the subject matter vary depending on the type of degree (PhD, Master’s or Bachelor’s)!

Do not send me KI-generated proposals!

Innovation economics

1. Various topics relating to patent analysis (data is available), for example:
   1. Network analyses linking patents
   2. Trend analyses on innovation in fields such as PV, wind energy, etc.
   3. Text analyses using natural language processing (software and a brief introduction are provided)
   4. … = feel free to suggest your own topics

Socio-ecological systems

1. Various topics relating to socio-ecological systems, for example:
   1. Collection of case studies and extraction of the success factors mentioned in the articles
   2. Meta-analysis of case studies
   3. Mapping success factors in articles to an existing framework
   4. Use of LLM to systematically extract information from articles
   5. … = your own topics are also welcome

Cooperation

1. Various topics relating to cooperation, for example:
   1. How can altruists be distinguished from egoists, and how can these distinctions be maintained over time?
   2. How can altruists be reliably identified on the basis of known characteristics (e.g. blood donations)?
   3. How can egoists be reliably identified on the basis of known characteristics?
   4. What are the benefits for communities consisting solely of altruists?
   5. Meta-analysis of public goods games, dictator games, ultimatum games, or other well-known behavioural economics experiments
   6. Compilation of influencing factors examined in public goods games
   7. Street experiments on altruism and egoism
   8. … = feel free to suggest your own topics

Bionics

1. Various topics on bionics and innovation, for example:
   1. Compilation of creativity techniques and their evaluation
   2. Classification of bionic inventions
   3. Literature review: compilation of classifications relating to evolutionary adaptations, bionic inventions, or new taxonomic developments
   4. Design of a categorisation system linking technology and biology
   5. Classification of evolutionary adaptations and creation of a taxonomy
   6. Finding Biological Solutions to Technical Problems
      Nature has already developed solutions to many technical problems: geckos adhere to surfaces without glue, and lotus leaves clean themselves. However, finding these solutions in the biological literature is difficult because biology and engineering speak different languages. You will build an evaluation dataset of technical problems and corresponding biological solutions. Using training data from bio-inspired patents and synthetically generated queries, you will fine-tune an embedding model specifically for this cross-domain translation task. You will then evaluate whether the adapted model identifies better and more surprising biological analogies than standard approaches. Methods: Python, embedding model fine-tuning, annotation, information retrieval evaluation.
   7. … = own topics are also welcome

Democracy, Coordination, Participation

1. Democratic Provisioning Systems
   Cooperative supermarkets, energy communities, and housing projects organize the provision of basic human needs democratically, collectively, and sustainably. You investigate what enables successful cooperation in such projects. This includes, for example, how teams stay motivated, work together effectively, design robust structures, and build community. Methods: case studies, interviews, participatory action research.
2. Organizing: How Can Large Numbers of People Act in a Coordinated Way?
   Social movements and grassroots initiatives need to mobilize large numbers of people to act together. One approach that explains how this can be achieved is community organizing. It emphasizes relationships, narratives, strategy, and distributed leadership. You select case studies of successful or unsuccessful organizing projects and investigate the factors that determine their success. Methods: case studies or system modeling to make complex interrelationships visible.
3. Living Lab in the Former Leiner Building in Graz
   A space for living labs focused on socio-ecological transformations is being established in the former Leiner building in Graz. Researchers and citizens will jointly develop and test new everyday solutions. You implement an idea of your own, such as a repair café or a swap market. Through interviews or surveys, you investigate how participants’ cooperation and practices change. The outcome is an understanding of the conditions under which such initiatives succeed.
4. Democratic Economic Systems
   Various proposals exist for how an economy organized democratically could work. Examples include Participatory Economics (Parecon), negotiated coordination, and multicriteria decision-making procedures. You select one of these proposals and analyze it from a systems science perspective. In particular, you explore what kinds of dynamics such a system might develop. Methods: institutional analysis or system modeling.

LLMs

1. Reconstructing Codebooks with LLMs
   In qualitative research, texts are coded using a codebook. However, it often remains unclear whether the coding actually reflects the original codebook. You develop a method in which a large language model iteratively reconstructs a codebook from previously coded data and compare it with the original. This makes it possible to determine whether coders have deviated from the codebook while also investigating how LLMs form and organize categories. Methods: LLM pipelines (Python), experiments with systematically biased datasets, comparison metrics, qualitative content analysis.
2. Decision Models for Qualitative Coding
   New open-source decision models, such as Cloudflare’s Clef, do not generate text. Instead, they answer structured questions by assigning probabilities to possible answers. This makes them ideally suited for classifying texts according to a codebook. How well do these models code qualitative data? Is fine-tuning with a small number of examples worthwhile, and how reliable are their confidence scores? The goal is to determine which passages the model can code autonomously and which require human review. All models run locally, so sensitive data never leave your own server. Methods: Python, LoRA fine-tuning, comparative experiments with generative LLMs and BERT, calibration analyses.