Assessing environmental, economic and social impacts is an established approach in industry and research to analyse the sustainability of products. However, conducting such assessments is time-consuming. In particular, collecting life cycle inventory data and linking it to databases requires considerable time and expertise. Prospective assessments covering numerous future scenarios are also computationally intensive. Artificial intelligence methods such as large language models and machine learning promise significant relief here. It remains open, however, how reliable and transparent the resulting outcomes are.
The aim of this thesis is to develop and test an AI-supported approach to sustainability assessment, using energy-intensive products as an example. Possible directions are:
- Automated collection of life cycle inventory data from scientific literature and environmental product declarations using large language models
- AI-supported matching of inventory flows to datasets of established databases such as ecoinvent
- Machine learning-based surrogate models for the rapid assessment of prospective scenarios
- ...
The specific direction will be agreed upon before the start of the thesis based on a short exposé.
Note: Programming experience in Python is recommended.
If you are interested, please contact Alexander Barke.