Using Large Language Models to Classify Existential Themes from Social Media Data
Method Stream: Text Mining
ECs: Both 14 and 23 EC thesis possible
Description:
In this assignment, you will bridge existential psychology with modern Natural Language Processing by training a model to recognise positive and negative core themes (e.g., meaning, freedom, shame, guilt, demoralization, coherence, hope, etc.) in everyday language. Drawing on classic existential thinkers such as Tillich, Yalom and Antonovsky, you will examine how these psychological processes surface in social-media, with the goal of fine-tuning a transformer model that reliably detects these themes across diverse corpora, including social media, electronic health records, diaries and literary works.
Who are we looking for?
We are looking for students who enjoy turning language into insight. You will gather publicly available texts—such as Reddit threads—clean and organise them into a research corpus and manually annotate a sample for existential themes. The project suits anyone fascinated by meaning, freedom, shame and related motifs and who is comfortable working with large bodies of text. Prior programming experience is optional; a willingness to learn basic Python along the way is enough. Close attention to annotation quality, working with codebooks, respect for data ethics, and genuine intellectual curiosity are highly valued.
What do we offer?
This is an ambitious project in scope. You should expect to work collaborative with other students and you can expect active engagement from your supervisors.