An antidote to the orthodoxy

Western science largely takes a colour-blind, neutral, and apolitical viewpoint. Power asymmetries and social hierarchies are viewed as inconsequential to scientific endeavours. The Western, white, cis-gendered, able-bodied and male perspective – hailed as ‘God’s eye view’ – is the standard against which all science is measured. Data science is no different in inheriting this worldview. Dichotomies such as emotion vs. reason are commonplace; context is seen not as a vital background but as a contaminant; and the data scientist (with her background, interests, motivations, and political inclinations) remains invisible. Data Feminism is an antidote to this orthodoxy, wherein the authors begin by acknowledging their privilege, status, and positionality, and continue to engage reflexively throughout the book. 

The book puts forward a feminist data science by building on the work of scholars and activists from intersectional feminism to sociology, archival studies and more. In doing so, the authors tackle persistent misconceptions: that data are neutral, that ‘numbers speak for themselves’, and that data exist out in the world awaiting collection. Far from being neutral, objective, and apolitical, data cannot be divorced from the inherently oppressive and racist structures in which they are embedded. Our lives are affected by datasets and algorithms which recapitulate historical and existing power asymmetries. Data practices that are unaware of hierarchies of privilege, context and oppressive structures sustain the status quo. Power is wielded where it already is concentrated.  

The book brings humans back to the centre of data science and remains close to them throughout, asking questions such as: Whose views predominate in data science? Who is marginalised and made invisible? Who is forced into visibility and excessive surveillance to the detriment of their safety? Who benefits? Who disproportionately suffers harm? The book shines a light on such issues and concedes that data science should acknowledge and tackle them directly.

In centring humans, the authors attempt to ground data practices in messy and contextual political ecologies and concrete lives. Since most data in data science are about people, it is only realistic that practices such as data visualization reflect human experiences. The authors oppose dichotomies such as emotion vs. reason, in contrast to the orthodoxy which forcefully divorces emotions from visualisation. Instead, they advocate for data visceralization – a type of data visualization framed around emotions where the whole body, not just the eye, can partake in understanding data – for representing the conditions of living bodies. 

As algorithmic decision making increasingly becomes integral to daily life, there is also a growing body of work surrounding ethics in data science and AI. Yet, a substantial amount of the work can be classified as shallow insofar as asymmetrical power distributions and structural inequalities get little, if any, attention. The focus of these ethical analyses is on narrow problems such as improving fairness metrics, often without consideration for the harm data tools can inflict on marginalised and vulnerable communities. While not entirely opposing such shallow efforts, Data Feminism provides principles for understanding and practicing data science. At its core, the book contends that data and data science is not an abstract and apolitical practice that emerges in a vacuum but is inherently historical, cultural and contextual. Consequently, the book is geared towards understanding the broader issues of justice over the narrow concept of ethics; tackling structural oppression over bias; and prioritising equity over fairness. 

Data practices – dominated by white Western men unaware of their oppressive structures – as long as they remain unexamined harm marginalised communities. Those from dominant groups are unable to detect these harms, what the authors call ‘privilege hazard’. Projects that are led by and emerge from minoritized communities hold the key to sustainable solutions. The book is full of examples of ongoing projects from marginalised communities, including data initiatives by black women, transgender and non-binary people, and indigenous people.

If you are interested in how data science can impact our lives directly or indirectly, this book is a good read. It is written in a playful manner and provides accessible examples throughout. Some of the neologisms such as ‘Big Dick Data’ – masculinist, totalising fantasies of world domination as enacted through data – make the book a gem.

For those coming to data justice, ethics, and fairness, from statistics, data science, or computational sciences generally, this book will feel ground-breaking as it radically reimagines these concepts. For those with a background of, for example, Black feminist studies, the background work that the book leans on is not novel. Nonetheless, synthesising theories and ideas from various disciplines is challenging – interdisciplinary work is an uphill climb – and Data Feminism manages this eloquently. 

-       Reviewed by Abeba Birhane, PhD Candidate, University College Dublin & Lero – The Irish Software Research Centre