# An Analytics of Culture: Modeling Subjectivity, Scalability, Contextuality, and Temporality

**Authors:** Nanne van Noord, Melvin Wevers, Tobias Blanke, Julia Noordegraaf, Marcel Worring  
**arXiv:** [2211.07460](https://arxiv.org/abs/2211.07460)  
**Date:** 2022-11-14  
**Summary page:** https://www.summarizepaper.com/en/paper/2211.07460v1/

## Key points

- The relationship between culture and AI is bidirectional
- AI models are used to analyze culture while shaping our understanding of it
- Four concepts must be incorporated into AI research to do justice to the complexity of human culture:
- Subjectivity (individual nature of cultural experiences and interpretations)
- Scalability (considering different scales at which cultural phenomena occur)
- Contextuality (situating cultural artifacts within their historical and social contexts)
- Temporality (recognizing the crucial role time plays in shaping cultural meaning formation)
- Working with archival data requires expertise in determining meaningful information beyond metadata
- Computational models must contextualize data while distinguishing between incidental and meaningful information
- Longitudinal data is required to capture how meaning evolves over time by encoding how meanings are retained or forgotten over time
- Collaboration between humanistic scholars and AI researchers can develop more accurate models that better capture the intricacies of culture while avoiding issues such as bias.

## Summary

The relationship between culture and AI is bidirectional, with AI models being increasingly used to analyze culture while shaping our understanding of it. To overcome the tension that arises from these models being trained on collections of cultural artifacts that may not always correctly encode expressions of culture, four concepts must be incorporated into AI research to do justice to the complexity of human culture: subjectivity, scalability, contextuality, and temporality. Subjectivity refers to the individual nature of cultural experiences and interpretations; scalability involves considering different scales at which cultural phenomena occur; contextuality emphasizes the importance of situating cultural artifacts within their historical and social contexts; and temporality recognizes the crucial role time plays in shaping cultural meaning formation. Working with archival data requires expertise in determining what is meaningful beyond metadata such as where, when, and by whom an artifact originated. Computational models must be developed that can contextualize data while distinguishing between incidental and meaningful information. Additionally, capturing how meaning evolves over time through longitudinal data requires encoding how meanings are retained or forgotten over time. In conclusion, this paper invites productive engagement and collaboration between humanistic scholars and AI researchers to develop more accurate models that better capture the intricacies of culture while avoiding issues such as bias.