Measuring Human Value Expression in Social Media Texts: Calibrated LLM Annotation and Encoder Transfer
arXiv:2606.11018v1 Announce Type: new Abstract: Measuring subjective constructs in naturally occurring social media text requires annotation procedures that are theoretically grounded, empirically validated, and transferable to an encoder model for scalable prediction. Using non-English social media posts annotated according to Schwartz's theory of basic human values, we investigate how different LLMs, prompts, and instruction languages operationalize the expression of values in text. We argue that although texts may permit multiple plausible interpretations, theory-based value definitions can
The proliferation of social media data and advancements in large language models make it timely to develop sophisticated methods for extracting nuanced human value expressions, moving beyond superficial sentiment analysis.
Accurate measurement of human values from social media allows for better understanding of societal sentiment, cultural nuances, and potential drivers of behavior, crucial for political, economic, and social strategy.
The ability to calibrate LLMs for complex, subjective annotations opens new avenues for scalable and theoretically grounded analysis of public opinion and value systems expressed in vast unstructured text data.
- · Social scientists
- · Market researchers
- · LLM developers
- · Governments/NGOs
- · Traditional survey methods (for some applications)
- · Oversimplified sentiment analysis tools
LLMs can be effectively trained and validated to annotate subjective constructs like human values with higher accuracy and scale.
Improved understanding of societal values and their evolution will lead to more targeted messaging, policy development, and marketing strategies.
The development of 'value-aware' AI agents and systems that can better interpret and interact with human motivations and cultural contexts.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL