Towards Goal-Oriented Volunteering - LLMs to the Rescue?

The UN 2030 Agenda recognizes volunteers as key actors in achieving the Sustainable Development Goals such as quality education, gender equality, environmental conservation and community well-being. At the same time, however, sustainability of the voluntary sector itself is massively endangered by profound changes in demography, social structure, and volunteer motives. This is not least since craving for meaningful volunteering opportunities simultaneously allowing to achieve personal goals is increasingly in the foreground. Thus, adhering to the principle »do good for others and for yourself« we strive to synergistically align volunteers' personal goals with volunteering opportunities proposing a recommender system based on customized cross-encoder models. In this respect, our contribution is threefold: First, we reveal in how far LLMs are able to provide labeled ground truth data for personal goals and volunteering opportunities by comparing existing approaches and propose their adoption to the peculiarities of our domain. Second, we put forward a learning approach for fine-tuned models using transfer learning based on cross-encoder models. Finally, we resolve the feasibility of the different labeling approaches and the resulting models based on appropriate metrics and statistical tests, reflected upon through complementing lessons learned.