Conservative AI and social inequality: conceptualizing alternatives to bias through social theory

Abstract

In response to calls for greater interdisciplinary involvement from the social sciences and humanities in the development, governance, and study of artificial intelligence systems, this paper presents one sociologist's view on the problem of algorithmic bias and the reproduction of societal bias. Discussions of bias in AI cover much of the same conceptual terrain that sociologists studying inequality have long understood using more specific terms and theories. Concerns over reproducing societal bias should be informed by an understanding of the ways that inequality is continually reproduced in society -processes that AI systems are either complicit in, or can be designed to disrupt and counter. The contrast presented here is between conservative and radical approaches to AI, with conservatism referring to dominant tendencies that reproduce and strengthen the status quo, while radical approaches work to disrupt systemic forms of inequality. The limitations of conservative approaches to class, gender, and racial bias are discussed as specific examples, along with the social structures and processes that biases in these areas are linked to. Societal issues can no longer be out of scope for AI and machine learning, given the impact of these systems on human lives. This requires engagement with a growing body of critical AI scholarship that goes beyond biased data to analyze structured ways of perpetuating inequality, opening up the possibility for radical alternatives.

Artificial intelligence, algorithms, and social inequality: Sociological contributions to contemporary debates

“…While the conventional approach in AI and data science treats bias as ‘error’ – inaccurate modeling or prediction – a broader definition of bias that is also used (but less often articulated) equates it with whatever might be undesirable in society (Mitchell et al., 2020; Zajko, 2021). In short, while an algorithmic model that accurately reproduces society's ‘unequal ground truth’ might be considered unbiased in the first narrow sense of bias as accuracy, it may be biased according to this second view of bias as an undesirable tendency.…”

“…Whether we are looking at algorithmic search results, hiring decisions, or recidivism prediction, we will continue to confront examples of harmful discrimination that cannot be reduced to issues of accuracy; any attempt at a solution will be inherently political, and avoiding the issue or treating it as being ‘out of scope’ for data science amounts to a conservative orientation (Green, 2018; Zajko, 2021). As a consequence, a growing movement among AI researchers has been taking a more radical path to altering relations of power (Kalluri, 2020), but interdisciplinary contributions from scholars of power and inequality remain badly needed.…”

“…None of this has been surprising to scholars in fields such as science and technology studies (STS), critical data studies, or critical algorithm studies, who have long examined the relationships between technology and social structures (Felt et al., 2017; Gillespie & Seaver, 2016; Iliadis & Russo, 2016). Sociologists in general have been less engaged in these issues thus far, which is unfortunate given that social theory can fill the conceptual chasm currently identified as ‘bias’ in data science (Joyce et al., 2021; Zajko, 2021).…”

Speeding up to keep up: exploring the use of AI in the research process

“…Interviewees foresaw issues with the ways in which AI might reflect and further existing inequalities and bias. A large amount of research and regulation has targeted the minimisation of this bias (Caliskan et al 2017;Röösli et al 2021;Zajko 2021), but the hidden consequences of AI adoption with respect to research have yet to be fully addressed. If AI challenges institutional and academic identities and helps shape the future role of the academy, it may be pertinent to ask how technology can support, rather than impede, the process.…”

Ennalta määrätyn episteeminen kulttuuri ja politiikan liikkumatila

“…Keinoäly tarjoaa useita hyviä esimerkkejä ennalta määrätyn episteemisestä kulttuurista. Teknologia ei ole moraalisessa mielessä neutraalia (Cooley 1995) ja viime aikoina keinoäly on herättänyt pohdintaa teknologian eettisistä ja poliittisista seurauksista (Zajko 2021). Algoritminen päätöksenteko omaksuu ja toistaa olemassa olevia yhteiskunnallisia vinoumia: "objektiivisen arvostelukyvyn lupauksen varaan luodut järjestelmät päätyvät uusintamaan vinoumia ja eriarvoisuuksia, joita niiden 'kouluttamiseen' käytetyt yhteiskunnat sisältävät" (emt.).…”