Organized Session

SHOT Joint Session

Another Vast Machine I: Data, Models, and Simulations in the Human Sciences

Organizer

Tabea Cornel

New College of Florida

Chair

Emily Klancher Merchant

UC Davis

Metadata

Session Abstract

Ten years ago, Paul Edwards argued that the "vast machine" of meteorology had neared completion. Global systems of weather observation, data manipulation, and data interpretation have not tamed the climate, but they have established (almost) undeniable facts about it, including the existence of global warming. More recently, data, models, and simulations have become essential to the human sciences. Recidivism algorithms, genetic diagnostics, and multivariate pattern analysis are only three examples of the numerous technologies on which scientists rely to establish facts about humanity and predict the future of individual humans. How did these technologies originate? What consequences are there, if any, of applying technologies to humans that had originally been developed for non-human systems? How stable have such technologies and related concepts been across time, space, and cultures? To what extent have these technologies reinforced or eroded power dynamics within and between scientific communities? In what ways have these technologies helped or hindered establishing (almost) undeniable facts about humanity?

Presenter 1

From Mass to Micro Persuasion in an Age of Big Data

Rebecca Lemov

Harvard University

Abstract

During World War II a small team of behavioral scientists undertook the Mass Persuasion study to trace the mechanisms by which Americans exposed to propaganda messages were compelled to change their normal behavior. Published in 1946, Robert K. Merton and his team laid out aimed to discover strategies by which "technicians of sentiment" influenced a mass audience by triggering patriotic responses. Researchers combined new techniques of data gathering and processing in order to target the domain of human subjective response. This talk traces how such persuasive techniques work today, when they are driven by machine learning, real-time information gathering, and AI in the form of microtargeting. Cambridge Analytica and more recent examples of behavioral engineering are examined.

Metadata

Presenter 2

Hadooping the Genome: The Impact of Text Search Tools in Biomedicine

Hallam Stevens

Nanyang Technological University

Abstract

This talk examines the consequences of the so-called 'big data' technologies in biomedicine. Analyzing algorithms and data structures used by biologists can provide insight into how biologists perceive and understand their objects of study. Here, I draw on published scientific literature and publicly available descriptions of big data algorithms to track the migration of algorithmic tools across domains. Algorithms used for sequence comparison or sequence mapping in genomics are largely derived from the powerful tools for text searching and indexing that have been developed since the 1950s and now play an important role in online search. In biology, sequence comparison algorithms have been used to assemble genomes, process next-generation sequence data, and, most recently, for 'precision medicine.' I argue that the predominance of a specific set of text-matching and pattern-finding tools has influenced problem choice in genomics. It allowed genomics to continue to think of genomes as textual objects and to increasingly lock genomics into 'big data'-driven text-searching methods. In genomics, tools developed for searching human-written texts are now widely deployed for searching for meaning and pattern in biological data. However, genomes and other' omic data are not human-written and are unlikely to be meaningful in the same way.

Metadata

Presenter 3

Fluorescent Sensors in Bayesian Brains: Transgenic Mice as Statistical Technologies in Neurogenomics

Tabea Cornel

New College of Florida

Abstract

At the Allen Institute for Brain Science, researchers use transgenic mice as statistical technologies to gain insight into human cognition. One of the researchers at the Allen Institute stated that his team relied on transgenic mice as "a Bayesian prior." The researchers use two-photon calcium imaging to detect fluorescent protein-based Ca2+ sensors in the brains of transgenic mice, and they use these data to create a Bayesian prior probability distribution to help model visual perception in humans. This high-tech mathematization of genetically altered rodents grew out of two converging developments: first, the advance of model organisms in the life sciences, and, second, the recent rise of the "Bayesian Brain." The latter concept is rooted in the conviction that Bayesian statistics are (more or less) accurate models of human cognitive processing. The "Bayesian Brain" has been touted as an epistemological revolution that will impact our understanding of the human mind and the development of artificial intelligence. Its conceptual roots reach back to the 18th century, when Thomas Bayes pondered the probabilities of random events (e.g., where a ball comes to rest after being thrown upon a table). In this paper, I analyze the epistemic function of mice at the Allen Institute, and I ask in what ways the reliance on triple transgenic mouse lines with built-in fluorescent chemical sensors might prompt us to expand our understanding of the roles of model organisms in neuroscientific research. I also inquire into the extent to which falling back on centuries-old statistical paradigms might be less of an epistemological revolution than evidence of a neuroscientific uncertainty regarding how to interpret big data in the absence of big theories. More broadly, I ask how neuroscientific practice and concepts have been reshaped at the intersection of Bayesianism and neuro(trans)genomics.

Metadata

Presenter 4

Anatomy of a Biomedical Controversy: Cancer Screening, Statistics, and Simulation Modeling

Christopher J. Phillips

Carnegie Mellon University

Abstract

The US Preventive Services Task Force (USPSTF) is a collection of experts who deliberate about evidence taken from empirical studies, biomedical data, and computer simulation models to make recommendations about the frequency and type of screening regimes. Their recommendations are widely used by insurance companies, government regulators, and primary care physicians to determine how frequently mammograms or colonoscopies should be administered, for example. Not surprisingly, the USPSTF consequently produces conclusions that are highly consequential in both financial and public health terms and simultaneously highly contested. For historians of science and technology, the Task Force's deliberations and recommendations provide a rich example of how emerging statistical and simulation technologies play an increasingly central (if controversial) role in the practice of modern medicine. This talk will use the history of the USPSTF and its increasing reliance on computer-based micro-simulation models to explore how novel statistical technologies shape our understanding of human futures, and in particular the possible futures of human health that various screening recommendations would usher in. The talk will use breast and colorectal cancer screening recommendations-two of the most controversial of the recent past-to trace how computerized simulation models emerged in the 1980s and were refined and incorporated more formally in 2000 into the National Cancer Institute-sponsored Cancer Intervention and Surveillance Network. This talk not only reveals how novel statistical technologies have fundamentally shaped modern medicine, but also how the controversial reliance on computer simulation models over traditional clinical trial data raises profound questions about the changing role of empirical evidence, statistical judgment, and data-driven science.

Metadata