Matti Nelimarkka
@matnel.bsky.social
Conceptual alignment? We find that neither more traditional supervised models nor advanced LLM-based models seem to fully capture the intended conceptual richness for social science content classifications.
Conceptual alignment? Using Explainable AI models to examine how AI classifies social science content
Summary of our poster at the International Conference on Computational Social Science
open.substack.com
August 5, 2025 at 6:57 AM
Conceptual alignment? We find that neither more traditional supervised models nor advanced LLM-based models seem to fully capture the intended conceptual richness for social science content classifications.
Can unsupervised image clustering imitate social scientists' gazes?
Short answer: no; supervised image clustering cannot – at least yet – replace social scientists, not even with visual language models. Here is our reasoning for this conclusion:
open.substack.com/pub/socially...
Short answer: no; supervised image clustering cannot – at least yet – replace social scientists, not even with visual language models. Here is our reasoning for this conclusion:
open.substack.com/pub/socially...
Can unsupervised image clustering imitate social scientists' gazes?
Summary of our poster at the International Conference on Computational Social Science
open.substack.com
July 29, 2025 at 11:20 AM
Can unsupervised image clustering imitate social scientists' gazes?
Short answer: no; supervised image clustering cannot – at least yet – replace social scientists, not even with visual language models. Here is our reasoning for this conclusion:
open.substack.com/pub/socially...
Short answer: no; supervised image clustering cannot – at least yet – replace social scientists, not even with visual language models. Here is our reasoning for this conclusion:
open.substack.com/pub/socially...