Culture Analytics

Materials for the DH 2019 Workshop

Culture Analytics

James Abello, Edoardo Airoldi, Cecilia Aragon, Katy Börner, Russel E. Caflisch, Maria-Rita D’Orsogna, Tina Eliassi-Rad, Jacob G. Foster, Jianbo Gao, Blake Hunter, Lev Manovich, Isabel Meirelles, Filippo Menczer, Vwani Roychowdhury, Maximilian Schich, Timothy R. Tangherlini. (For affiliations and corresponding author see below.)

Culture analytics holds enormous potential, allowing cultural understanding to be leveraged for intelligent policy making in the face of diverse global challenges.

Culture analytics (CA) investigates cultures and their temporal evolution by combining data-driven analysis and mathematical modeling with qualitative theories of culture. The rapid development of CA over the past decade has been powered by enormous gains in computational power, the emergence of sophisticated algorithms to address complex data, and the large-scale digitization of culture in archives, museums and libraries, and in the digital expressions of billions of people. Insights drawn from studies of these complex cultural datasets have real world applicability, including the protection of endangered cultural sites, the documentation of the world’s cultures and their histories, and the discovery of newly emerging cultural phenomena. Understanding culture’s role in areas as diverse as education, tourism, healthcare, business, urban development, food security or law provides a solid foundation for sound policy making. For example, art historians, mathematicians, and computer vision experts studying the paintings of early masters have helped identify art counterfeiters (1) and preserve legitimate artifacts, such as a fifteenth century altarpiece in Ghent (2). Researchers studying the personal anecdotes of tens of thousands of parents posting on social media have revealed how vaccine hesitancy has taken hold in some communities (3), helping public health officials to devise culturally attuned messaging that may avert future health crises. And as food scarcity becomes an increasing global challenge (4), collaborative work among computer scientists, chemists, linguists, and historians on recipes from around the world has identified the flavor networks at the root of diverse cuisines, a finding which can be harnessed to develop sustainable foods that are culturally recognized and “taste right” (5) [Figure 1].

Figure 1

Figure 1: The complex interplay of the geography of food (a), its representation in art (b), flavor networks (c), and food security (d). The tomato migrated from South America to Central America, where it became a mainstay of Mayan culture. It was transported to Europe in the sixteenth century by Spanish conquistadors who were intrigued by its ubiquity in Aztec civilization. On its arrival in Europe, the tomato became not only a central ingredient of Mediterranean cuisine but also an object of study in pictorial art. Initially abhorred by the British, it intersected with various flavor networks and gained popularity in colonial era America, ultimately resulting in hybrid foods such as ketchup, the ubiquitous hotdog condiment, which was originally a South East Asian fermented fish sauce. (6) Recent work on sustainable farming in drought stricken areas such as California has developed the tomato, given its broad appeal in many cuisines and its health benefits, as a plant suitable for shifting climates, resulting in a dramatic increase in small tomato farms throughout sub-Saharan Africa. Integrating cultural understanding of diet, health, and agricultural practice is now seen as a key aspect of ensuring food security in the future. (4)

Despite these impressive gains, such early successes highlight the complex, systems- science challenges facing the field.

Can we formalize the basic terms and measures for the study of culture? Even though digital repositories allow scholars to explore aspects of culture such as literature, music, and art in great detail, the lack of consistent formal terms and measures inhibits work beyond a single collection. There are currently no common formalizations of terms or measures that adequately capture the very broad range of cultural practices–from pictorial art, to statuary, to literature, to religious practice, to music, to architecture, to cuisine, to family and community organization– that contribute to cultural complexity. For instance, while there are excellent digital resources to study the pictorial art and statuary housed in the medieval Spanish cathedral of San Juan de Compostela (7), there are few avenues for aligning this collection with analogous collections from around the world. This alignment problem becomes more pronounced as the scale of the collections and the diversity of the artifacts increases. Researchers studying early Asian religions would face considerable hurdles if they wanted to coordinate their investigations of the temple complex in Angkor (8) with studies of potentially related sites such as the great Buddhist temples of Bhamala Stupa near Haripur, Pakistan (9). Yet such coordination is necessary if we are to compare cultures consistently across space and time, and devise strategies for preserving cultural heritage.

Can we find scalable algorithms to detect the culturally meaningful structures and transitions in heterogeneous data, and their influence on one another? Cultural practices are embedded in a thick context of interdependence involving many players, but there is no consensus as to how such relationships can be captured. Developing consistent approaches to modeling interdependence will offer insight into how changes in one aspect of a culture create ripple effects through other aspects of that and other cultures. For example, by working with court records from England, historians and data scientists have shown how changing class perceptions exerted gradual influence on the conceptualization of crime and the moderation of punishment up through the nineteenth century (10). Work by literary scholars and statisticians explored how these same changing class perceptions influenced literature and Londoner’s ideas about their city, attaching positive and negative sentiments to different neighborhoods (11). Combined, these studies become far more powerful, offering a model of how interdependent cultural practices influence both urban development and legal practice. In one of the most ambitious projects to date, archeologists, architects, literary scholars, historians, and geographers are creating the “Venice Time Machine”, a vast digital collection based on one thousand years of archival material. This collection can be used to explore the interrelationships between trade, science, art, and politics, and it will benefit greatly from algorithms designed to model and visualize interdependence (12). As these methods are developed and refined, researchers will be able to pinpoint areas where developments in one area of a culture have significant impact on others, thereby providing new historical insight into the complexity of culture, and identifying areas where decision-makers can concentrate efforts to effect positive change.

How do we measure the rate of change in culture, and identify what influences this change? Even though the idea that cultures rise and fall is commonplace (13), there are still no standard mathematical methods for modeling these dynamics. Pioneering work has shown that the rise and fall of cultural centers can be traced by mining the birth and death records of thousands of cultural figures, thereby capturing the migrations that led to phenomena as diverse as the emergence of competing cultural centers during the Renaissance, the slow decline of Rome as a hub in the first millennium, and the impact of missionary work in Edo-era Japan (14). Nascent work in cultural evolution and cultural attraction present other potentially fruitful avenues of inquiry (15). Yet measuring complex phenomena, such as changes in people’s use of the pilgrimage route of the Camino de Santiago from the age of the medieval ascetic to the current age of the selfie-snapping tourist remains an open challenge (16). The dynamic nature of culture requires the development of novel algorithms that can capture these complex phenomena, perhaps allowing us to identify the cultural forces that shape business organizations in different countries (17), contributed to the historical collapse of the Maya (18), or may lead to future forms of cultural hybridization. With an ever-increasing number of data resources to support the dynamic multiscale analysis of culture, we are poised to be able to answer these questions in a consistent and reproducible manner across many cultures and time scales. As these methods are refined, the ability to model real-time data may lead to predictive models of culture change.

Despite the scale of these unresolved challenges, CA is producing important, actionable insights. Because it is grounded in established theories of culture, it offers perspectives on where we have been and where we might be going. By combining the collective expertise of data scientists and cultural researchers while making use of high quality, culturally thick data at scales previously unimaginable, CA presents clear paths for the development of culturally informed decision-making to deal with difficult problems while providing great insight into cultural history.

References

  1. S Jafarpour, et al. Stylistic analysis of paintings using wavelets and machine learning. European Signal Processing Conference (EUSIPCO-2009), Glasgow, Scotland (2009).
  2. A Pizurica, et al. Digital image processing of the Ghent Altarpiece: Supporting the painting’s study and conservation treatment. IEEE Signal Processing Magazine 32: 112-122 (2015).
  3. T Tangherlini, et al. “Mommy Blogs” and the Vaccination Exemption Narrative: Results From A Machine-Learning Approach for Story Aggregation on Parenting Social Media Sites. JMIR public health and surveillance 2 (2017).
  4. HC Godfray, et al. Food security: the challenge of feeding 9 billion people. Science 12: 812-8 (2010).
  5. YY Ahn, et al. Flavor networks and the principles of food pairing. Scientific reports 1 (2011).
  6. D Jurafsky. The Language of Food: A Linguist Reads the Menu. New York: WW Norton (2014).
  7. J Dagenais. Romanesque Redivivus: a full-scale 3D computer reconstruction of the Medieval Cathedral and town of Santiago de Compostela (ca. 1211) (2013).
  8. D Evans, et al. “A comprehensive archaeological map of the world’s largest preindustrial settlement complex at Angkor, Cambodia.” PNAS 104, 14277-14282 (2007).
  9. A Lawler. Huge statue suggests early rise for Buddhism. Science 353, 336 (2016).
  10. S Klingenstein, et al. The civilizing process in London’s Old Bailey. PNAS 111, 9419-9424 (2014).
  11. R Heuser, et al. The emotions of London. Stanford Literary Lab Pamphlet 13 (2016).
  12. A Abbot. The “time machine” reconstructing ancient Venice’s social networks. Nature 546, 341-344 (15 June 2017)
  13. J Diamond. How Societies Choose to Fail or Succeed. New York: Viking (2005).
  14. M Schich, et al. A network framework of cultural history. Science 345, 558-62 (2014).
  15. A Mesoudi. Pursuing Darwin’s curious parallel: Prospects fora science of cultural evolution. PNAS Culture Analytics 5114, 7853-7860 (2017).
  16. RNC Lois Gonzalez. The Camino de Santiago and its contemporary renewal: Pilgrims, tourists and territorial identities. Culture and Religion 14, 8-22 (2013).
  17. M Castells. The rise of the network society: The information age: Economy, society, and culture. New York: John Wiley & Sons (2011).
  18. GH Haug, et al. Climate and the collapse of Maya civilization. 299, 1731-1735(2003).

Acknowledgements

The authors thank the participants of the NSF’s Institute for Pure and Applied Mathematics (IPAM) Culture Analytics program. More information, including participants, presentations, and white papers can be found here.

Image Credits

1a. H Hussein. Food Chains. Lapham’s Quarterly 3 (2011). 1b. L Melendez. Still Life (1784). 1c. YY Ahn, et al. Flavor networks and the principles of food pairing. Scientific reports 1 (2011). 1d. L Chari. Small fields, big harvests. http://www.technoserve.org/blog/small-fields-big-harvests

Author Affiliations

Abello: Rutgers University • Airoldi: Harvard University • Aragon: University of Washington • Börner: Indiana University • Caflisch: UCLA, Institute for Pure and Applied Mathematics • D’Orsogna: California State University, Northridge • Eliassi-Rad: Northeastern University • Foster: University of California, Los Angeles • Gao: Guanxi University • Hunter: Claremont McKenna College • Manovich: City University of New York Graduate Center 12 OCAD University • Menczer: Indiana University • Roychowdhury: University of California, Los Angeles • Schich: The University of Texas at Dallas • Tangherlini: University of California, Los Angeles.

*Correspondence to: tango@humnet.ucla.edu