PaperlayerView exact revision on arXiv

Source-linked AI summary

Data Science as Political Action: Grounding Data Science in a Politics of Justice

Ben Green

arXiv:1811.03435v4cs.CYcs.LG

TL;DR

Data science ethics initiatives are insufficient to address the social-justice issues implicated by data-driven systems. The paper frames data science as political action and proposes deliberation about impacts and goals. It argues that new methods and norms can support more rigorous contributions to social justice.

  • Problem

    Ethics initiatives are insufficient to address the broad social-justice issues implicated by data science and its scrutinized algorithms.

    Source: §1 · pp. 1–2

  • Method

    The paper frames data science as political action, urging clear political commitments, rigorous consequence evaluations, and deliberation about desirable impacts with affected communities.

    Source: §4 · p. 13

  • Results

    The paper argues that deliberating about political goals and developing new methods and norms can support more rigorous contributions to social justice.

    Source: §4 · p. 14

  • Takeaways & Limitations

    Deliberation should include scholars from other fields, policymakers, and affected communities, and remain open to non-technological alternatives.

    Source: §4 · p. 13

    Data scientists are only one set of actors in social and political progress and should partner with broader coalitions and social movements.

    Source: §3.4.3 · p. 13

Abstract

from arXiv · show

In response to public scrutiny of data-driven algorithms, the field of data science has adopted ethics training and principles. Although ethics can help data scientists reflect on certain normative aspects of their work, such efforts are ill-equipped to generate a data science that avoids social harms and promotes social justice. In this article, I argue that data science must embrace a political orientation. Data scientists must recognize themselves as political actors engaged in normative constructions of society and evaluate their work according to its downstream impacts on people's lives. I first articulate why data scientists must recognize themselves as political actors. In this section, I respond to three arguments that data scientists commonly invoke when challenged to take political positions regarding their work. In confronting these arguments, I describe why attempting to remain apolitical is itself a political stance--a fundamentally conservative one--and why data science's attempts to promote "social good" dangerously rely on unarticulated and incrementalist political assumptions. I then propose a framework for how data science can evolve toward a deliberative and rigorous politics of social justice. I conceptualize the process of developing a politically engaged data science as a sequence of four stages. Pursuing these new approaches will empower data scientists with new methods for thoughtfully and rigorously contributing to social justice.

1 Introduction

Scrutiny of data-driven systems has prompted ethics initiatives, but ethics alone is insufficient to address the social-justice issues implicated by data science. The article argues that data science is political action and that practitioners must assess its downstream effects on people's lives.

1 Introduction

  • Algorithms used in hiring, healthcare, criminal sentencing, welfare, and social media have been associated with bias, opacity, manipulation, misinformation, and amplified political extremism.The datasets supporting some systems can also expose sensitive and intimate information about people.Source: §1 · p. 1
  • Ethics frameworks can support reflection, but their abstract principles, focus on individuals and design, and incorporation into corporate practices leave them insufficient for addressing social justice.The article describes some corporate uses of ethics language as “ethics-washing” that resists structural reforms.Source: §1 · p. 2
  • The article calls for data scientists to recognize their role in constructing society and evaluate their work by its downstream impacts on people's lives.It frames the argument around why practitioners must recognize themselves as political actors and how they can ground practice in social justice.Source: §1 · p. 2
  • Data scientists act politically by shaping how institutions distribute rights, status, and goods through tools that inform consequential decisions.The article uses politics broadly, beyond partisan or electoral activity, to include actions affecting who gets what, when, and how.Source: §1 · p. 2
  • The proposed shift aims to help data science contribute to equity and social justice without requiring practitioners to abandon technical expertise.The author calls for expanding which problems practitioners take on and how they engage with society.Source: §1 · p. 2

2 Why Must Data Scientists Recognize Themselves as Political Actors?

This section argues that data science is not apolitical: technological choices and claims of neutrality or social good can preserve the status quo. It calls for explicit political principles and evaluation of reform choices against alternatives to guide structural change toward social justice.

2.1 Argument 1: “I am just an engineer”

  • Engineering choices shape which interests technology supports, so data scientists cannot treat their work as politically neutral or responsibility-free.Traffic-signal timing, for example, helped define streets as spaces for cars rather than pedestrians; technological systems contribute to a particular social contract.Source: §2.1 · p. 4
  • Recognizing technological agency does not mean placing all responsibility on data scientists or allowing unaccountable experts to make public decisions.Technology is one tool among many for complex social problems, and democratic governance requires negotiating the limits of expert power.Source: §2.1 · p. 4

2.2 Argument 2: “Our job is not to take political stances”

  • Neutrality is unattainable, and striving for it is a conservative political stance that can reinforce the status quo and obscure data science’s embedded judgments.Data scientists are not responsible for creating every social condition, but they are responsible for how they engage with those conditions.Source: §2.2 · pp. 4–6
  • Data scientists need not share one political vision, but should surface normative disagreements and act according to articulated principles despite uncertainty.Principles may be complex, conflicting, and subject to change.Source: §2.2 · p. 6

2.3 Argument 3: “We should not let the perfect be the enemy of the good”

  • Appeals to incremental “good” lack robust definitions of “good” and “perfect” and fail to explain how to evaluate their relationship.The argument often assumes technology-centered incremental reform advances social progress, although some efforts may conflict with substantive equality and anti-oppression.Source: §2.3 · p. 6

2.3.1 Data science lacks a thorough definition of “social good”

  • Because “social good” lacks a clear definition, projects with conflicting political aims can share the label, while dominant groups’ judgments pass as neutral.The section contrasts projects seeking non-punitive alternatives to incarceration with work enhancing police operations, and argues that vague claims evade debate about impacts.Source: §2.3.1 · p. 7
  • Incremental reforms should be evaluated against debated visions of a more just society, including their short- and long-term effects and the paths they promote.A narrow reform that improves immediate conditions may hinder broader structural change; the section distinguishes reformist from non-reformist reforms.Source: §2.3.1 · p. 8
  • Data science’s emphasis on accuracy and efficiency favors reformist changes within existing systems, which can entrench the status quo and limit more egalitarian structural alternatives.Like a greedy algorithm, this strategy seeks local improvements that may obstruct better solutions to complex political problems.Source: §2.3.1 · p. 8
  • Pretrial risk assessments improve selection for detention, while abolitionists seek to end detention; these reforms embody conflicting visions of how criminal justice should work.Abolitionists favor reforms that reduce or replace carceral responses, rather than merely making criminal law administration more humane.Source: §2.3.1 · p. 9
  • Data science interventions should be assessed against alternative reforms, not only the status quo, without presuming machine learning is appropriate for every problem.Pretrial risk assessments can legitimize policies that drive racial injustice and mass incarceration, whereas ending cash bail and pretrial detention is a possible non-technological reform.Source: §2.3.1 · p. 9
  • Avoiding politics overlooks technology’s social impacts, privileges the status quo, and narrows possible reforms; data science cannot meaningfully advance social justice without accepting its political role.Articulating political visions is necessary to identify tensions among reforms and debate which to pursue.Source: §2.3.1 · p. 9

3 How Can Data Scientists Ground Their Practice in Politics?

The article maps a possible four-stage path from interest and reflection through new applications to practice-level change. It argues that justice-oriented data science requires participation, rigorous impact methods, and broader structural reform.

3.1 Stage 1: Interest

  • Programs should expand pathways for social-impact work and prioritize social and political reforms over deploying technology.The author notes that opportunities to pursue an articulated vision of social benefit are scarce; programs can offer methods, support, and pathways while treating algorithms as one imperfect intervention.Source: §3.1 · p. 10

3.2 Stage 2: Reflection

  • Reflection must extend beyond biased data to the political histories shaping crime definitions, institutions, and interventions.The author also calls for interdisciplinary and community expertise, including STS in data science training.Source: §3.2 · p. 10

3.3 Stage 3: Applications

  • Stage 3 uses critical, non-reformist frameworks to redirect applications toward injustice and power shifts, including community partners and institution-focused policing analyses.Examples include working with community groups rather than law enforcement and analyzing police officers or police behavior.Source: §3.3 · pp. 10–11
  • Stage 3 remains limited when politics stay unspoken, existing methods cannot rigorously represent social contexts, or applications proceed without deliberation and accountability.The author says covert approaches may advance only minor reforms; larger shifts require explicit defense and community grounding.Source: §3.3 · p. 11

3.4 Stage 4: Practice

  • Stage 4 calls for new epistemologies, methods, and cultures that orient data science around social justice.The proposed shift remains somewhat speculative, though the author identifies broad directions.Source: §3.4 · p. 11, §3.4.2 · p. 12

3.4.1 Participatory data science

  • Participatory practice means replacing detached objectivity with situated perspectives, designing with affected communities, and subjecting decisions to democratic oversight.Participation should address whether to develop an algorithm and how it should be used, not only its design.Source: §3.4.1 · pp. 11–12

3.4.2 New methods and cultures

  • Data science needs methods and institutions that evaluate real-world social impacts rigorously and foster technical and nontechnical contributions.The author calls for interdisciplinary collaboration and for workshops, conferences, and journals that support blended methods.Source: §3.4.2 · pp. 12–13

3.4.3 Engaging with the broader political context

  • Justice-oriented data science also depends on structural reform and on data scientists acting as partners in coalitions and social movements.The author describes worker organizing and support for communities and activists as examples, while noting data scientists cannot promote progress alone.Source: §3.4.3 · p. 13

4 Conclusion

Data science must abandon neutrality and orient its values and practices toward social justice, grounding work in clear political commitments and rigorous evaluation of consequences. This requires deliberating over desirable impacts, including affected communities, and remaining open to non-technological alternatives.

  • Data science must abandon neutrality, orient its values and practices toward social justice, and ground its work in clear political commitments and rigorous evaluation of consequences.Source: §4 · p. 13
  • Political data science requires debating desirable impacts, rigorously evaluating issues, and remaining open to non-technological alternatives.Source: §4 · p. 13
  • Deliberation should include data scientists, scholars, policymakers, and affected communities; new frameworks and norms can support more rigorous contributions to social justice.Source: §4 · pp. 13–14
Loading 1811.03435v4…