What is Culture Analytics ?


(illustration by Katherine Kinnaird)

We are a group of researchers defining a new field of study.  We came together in the spring of 2016 at IPAM‘s Culture Analytics Institute. This site summarizes and publicizes the outcomes of the program, but also, more importantly, it keeps the conversation going.  If you are a culture analytics researcher, please register, log in, and join the conversation!  (To register, please send a request to ajofre -AT- faculty -dot- OCADU -dot- ca)


“Culture Analytics seeks to increase human comprehension of digitized culture at scales that preclude immediate deployment of traditional methods. Enabled by breakthroughs in algorithmic sophistication, computational power, and digitization rate, CA is intimately bound with advances in data science. Yet successful CA projects ideally support analysis, interaction and insight on a spectrum between macroanalysis and close reading. The field thus holds implications for both STEM and Arts & Humanities fields, and currently profits from productive critiques emerging from both.” 
-Peter Leonard


“Cultural Analytics is the data driven study, by way of testing hypotheses and discovering new ideas and patterns, of human activities and interactions.Today, we are at a stage where an incredible quantity of human activity and interactions is captured digitally or is in the process of being digitized and archived. Activities such as the dissemination of new ideas (ranging from political ideologies and propaganda to scientific discoveries and artistic artifacts) are almost exclusively taking place on the internet. Similarly, an enormous amount of human social interactions are taking place digitally over social media websites, e-mail, text-messages, etc. All these activities and interactions capture the culture of our modern times. In addition, numerous institutions have been digitizing human activities (such as books, photographs, audio recordings, etc) that took place in the past for preservation.

Using all these digital records, both from the past and present, provides us a unique opportunity to perform analytics on these large datasets. This allows us to test existing hypotheses on popular narratives (potentially disrupting many assumptions) and discover new ideas that were not known before. Since this is a Big Data problem, we need computational methods from machine learning and data science to assist in the discovery process. As in many other Big Data endeavours, such as computational biology, this process involves an expertise of both computational methods and the domain knowledge of the social sciences and humanities.
-Mehrdad Yazdani


“Culture analytics is a triangulation of qualitative and quantitative data driven approaches to understand macro, meso and micro level human living environments where people share norms, rules, beliefs, languages, and organizations. Human living environments consist of physical emotional and spiritual boundaries.”
– Sunmoo Yoon, RN


“The rapid accumulation of digitized cultural data, including historical, off-line, and online, has made development of culture analytics increasingly important. Following the general practice in analyzing big data, it is easy to see that the emerging field of culture analytics will adopt network analysis, statistics, data mining, and machine learning as the major analysis tools. These methods are clearly important for analyzing big data as well as cultural data. In particular, machine learning, and its newest phase, deep learning, has been and will continue to be pivotal for the ever increased automation the technology has been offering. However, machine learning utilizes black-box approaches and thus lacks fundamental understanding. A bigger problem with these approaches is the difficulty in translating the results by these approaches to specific problems, as they do not truly target the actual problems domain experts are interested in. To help overcome the limitations of these “mainstream” approaches, we advocate to employ complexity science to analyze cultural data. It is most important to note that complexity science not only has a number of models with rich dynamical behavior, including sudden changes in system dynamics (e.g., bifurcations), but also has many treasured concepts such as nonlinearity, multiscale, sensitive dependence on small disturbances, fractal self-similarity, long memory, extreme variations, and nonstationarity. These concepts all possess certain universality, and thus have wide applicability. More importantly, reasoning with concepts can lead to better formulated scientific problems. This is of critical importance; without the guidance of good problems, one would feel clueless or even become paralyzed when facing massive cultural as well as other types of big data. In those situations, big cultural data becomes burden rather than wealth. Complexity science will be especially important for developing cultural complexity measures, identifying phase transitions from cultural data, as well as developing a calculus of culture analytics. Indeed, our recent experience of finding scaling laws governing the long-range correlations in sentiment time series extracted from novels, paintings, K_pop video streams, as well as the evolution of world-wide political instability and conflicts, has greatly enhanced our belief.”
-JianBo Gao


More here by Lev Manovich