Mechanized Significance and Machine Learning: Why It Became Thinkable and Preferable to Teach Machines to Judge the World

Abstract

The slow and uneven forging of a novel constellation of practices, concerns, and values that became machine learning occurred in 1950s and 1960s pattern recognition research through attempts to mechanize contextual significance that involved building “learning machines” that imitated human judgment by learning from examples. By the 1960s two crises emerged: the first was an inability to evaluate, compare, and judge different pattern recognition systems; the second was an inability to articulate what made pattern recognition constitute a distinct discipline. The resolution of both crises through the problem-framing strategies of supervised and unsupervised learning and the incorporation of statistical decision theory changed what it meant to provide an adequate description of the world even as it caused researchers to reimagine their own scientific self-identities.

3Es for AI: Economics, Explanation, Epistemology

“…The prior hierarchies and vulnerabilities that already signify our episteme, risk becoming further datum for the systematically variable outcomes that will be installed by the logics of the machine. In other words, “the insight and intelligence with which we address machine learning systems today will be the linchpin of future bad laws that we must later protest” (Mendon-Plasek, 2021 , p. 57–58). The assessment of technologies is a complex matter, and as Sclove ( 1995 , p. 4–5) points out, this is not simply reducible to the questions that are typically asked by newspapers, public-interest groups, corporate leaders, and governmental bodies—i.e., Is it workable?…”

“…To begin with, there is the historical perspective on the intimate links between the development of economics and of machine learning from the middle of the twentieth century onwards. I bring together the two separate accounts, by Mirowski ( 2002 ) and Mendon-Plasek ( 2021 ), which highlight these rare histories for neoclassical economics and machine learning, respectively. In Cold War/post-war U.S, the research agendas of mechanistic conceptualization and realization of human behaviors, motivations, and interactions was furthered in both these disciplines from similar institutions, funding bodies (for instance RAND) and even individuals in mathematics, game theory, engineering, and decision sciences who moved between the disciplines.…”

“…The footprint of a mechanistic and computational view of describing and predicting human behavior became clearer with the development of “neoclassical price theory, game theory, rational expectations theory, theories of institutions mechanism design, the nascent program of ‘bounded rationality,’ computational economics, ‘artificial economies,’ ‘autonomous agents,’ and experimental economics” (Mirowski, 2002 , p. 9). Meanwhile, there was the impulse toward pattern recognition in order to mechanically identify significance and reproduce human judgment based upon the beliefs that “there was little difference between looking for patterns in the natural world and looking for patterns in human society” (Mendon-Plasek, 2021 , p. 33).…”

“…“The most secure part of our present knowledge is knowing what it is that we cannot know or do” [Louis Fein, Impotence Principles for Machine Intelligence , 1968 ,p. 443, in Mendon-Plasek ( 2021 ), p. 53]…”

Impact of Machine/Deep Learning on Additive Manufacturing: Publication Trends, Bibliometric Analysis, and Literature Review (2013-2022).

“…Historically, the idea behind machine learning (ML) stretches back to the early 1940s and 1950s when computer scientists proposed creating machines that could self-learn [1]. This era was marked by the creation of the "Perceptron algorithm" by Frank Rosenblatt in 1958 [2], which is considered by many scientists as one of the earliest and most influential contributions to the field [3,4].…”