UC Santa Barbara

Along with the increasing digitization of society, our personal data has been explicitly or implicitly collected and shared through a plethora of digital devices and online social media services. This personal data has become of vital importance for researchers, designers, and artists to represent an impression of our datafied society and to depict the images of data subjects through various forms of data representations. Meanwhile, the explosive increase of personal data has also been accelerated by commercial or governmental entities behind services or technologies to monetize the customer data or surveil citizens. These entities use the data to categorize and predict our behaviors, preferences, and identities through machines that are designed for services or applications, such as content recommendation and facial recognition.

As our society is increasingly datafied, we see ourselves through our data, which is collected and processed by the machines, for self-representation and self-understanding. Moreover, responses from the machines also affect how we behave and understand ourselves. Within this data-centered human-machine interrelationship, the Human-Data-Machine Loop, the machines see us through available, measurable data obtained from us and through stochastic, algorithmic processes to generalize individuals. Here, uncertainty exists because personal data is not an objective representation of oneself, and the machines are not perfect; they can be erroneous and biased upon the data or humans. These issues of uncertainty are difficult to estimate and represent, and they are problems to be solved, especially in scientific domains. But, this uncertainty perspective can be a creative force or theme in data art. This dissertation proposes an artistic approach to reflecting the uncertainty in the Human-Data-Machine Loop through data art. To this end, this dissertation defines the Human-Data-Machine (HDM) Loop as the main conceptual research framework for viewing our datafied society, along with possible types/sources of uncertainty in the Loop. Second, I propose three types of data art practice based on the HDM framework: Artist-Centered Practice, Artist-Machine Collaborative Practice, and Machine-Centered Generative Practice. Last, this dissertation explains and evaluates the author’s data-driven audiovisual art projects as an empirical case study of each data art practice.

This dissertation aims to contribute to expanding data art practice with the perspective of uncertainty in data practice and to raising the audience's awareness of uncertainty in data practice through artistic approaches.