Source-linked AI summary
Roles for Computing in Social Change
Rediet Abebe, Solon Barocas, Jon Kleinberg, Karen Levy, Manish Raghavan, David G. Robinson
TL;DR
The paper addresses concerns that technical fairness work may leave structural injustice untouched. It analyzes four modest roles for computing—diagnostic, formalizer, rebuttal, and synecdoche—and concludes that computing can support broader social change without solving social problems on its own.
Problem
Recent technical work on fairness, bias, and accountability may treat problematic features of the status quo as fixed while failing to address deeper injustice and inequality.
Method
The paper analyzes four potential roles for computing research in supporting broader social change while examining their opportunities and risks.
Results
Computing can diagnose social problems, shape their formal definitions, establish limits on technical possibilities, and make longstanding problems newly salient.
Takeaways & Limitations
Computational work can support fundamental social and political change when used as one contribution among many rather than as a substitute for broader reform.
Takeaways & Limitations
Computing is not sufficient on its own and carries hazards, including metric distortion and compatibility with particular sociopolitical arrangements.
Abstract
from arXiv · showhide
A recent normative turn in computer science has brought concerns about fairness, bias, and accountability to the core of the field. Yet recent scholarship has warned that much of this technical work treats problematic features of the status quo as fixed, and fails to address deeper patterns of injustice and inequality. While acknowledging these critiques, we posit that computational research has valuable roles to play in addressing social problems -- roles whose value can be recognized even from a perspective that aspires toward fundamental social change. In this paper, we articulate four such roles, through an analysis that considers the opportunities as well as the significant risks inherent in such work. Computing research can serve as a diagnostic, helping us to understand and measure social problems with precision and clarity. As a formalizer, computing shapes how social problems are explicitly defined --- changing how those problems, and possible responses to them, are understood. Computing serves as rebuttal when it illuminates the boundaries of what is possible through technical means. And computing acts as synecdoche when it makes long-standing social problems newly salient in the public eye. We offer these paths forward as modalities that leverage the particular strengths of computational work in the service of social change, without overclaiming computing's capacity to solve social problems on its own.
1 INTRODUCTION
The paper responds to concerns that technical fairness efforts may leave structural injustice unaddressed by identifying modest ways computing can support broader social change. It frames computing as one contribution among many, not a unilateral solution.
- Motivation: Fairness, bias, and accountability have become central concerns in computer science and technology practice, especially for high-stakes decisions.Algorithmic systems may improve prediction and resource allocation while also introducing, perpetuating, or worsening inequality.
- Problem: Critics argue that interventions focused on automated decisions and resource allocation can obscure the background conditions sustaining injustice.Examples include improving admission parity instead of low-income high-school instruction and changing employment selection while leaving hostile workplace cultures untouched.
- Research question: The paper asks whether computing can contribute to wholesale social and political reform as well as incremental intervention within existing systems.It rejects technological solutionism and seeks roles through which computing supports rather than supplants other ways of addressing social problems.
- Contribution: The authors propose four potential roles for computing research that may align with broader social change.These roles are presented as illustrative, nonexclusive, and intentionally modest, with hazards explored alongside their opportunities.
2 COMPUTING AS DIAGNOSTIC
Computing can diagnose and precisely characterize social problems, particularly their manifestations in technical systems, while remaining part of a broader multidisciplinary inquiry. Diagnostic work can expose discriminatory patterns and influence systems, but metrics, access limits, and diagnosis alone constrain its value.
- Computing as diagnostic: Computational methods can measure social problems and diagnose how they manifest in technical systems.Used with other empirical methods, they provide evidentiary support for examining values in technology without constituting a sufficient remedy.
- Computing as diagnostic: Computational studies have documented discriminatory patterns in advertising, word embeddings, machine translation, and facial analysis systems.These analyses reveal the depth, scope, pervasiveness, and mechanisms of problems affecting racialized and gendered groups.
- Computing as diagnostic: Diagnostic work is especially valuable for auditing black-boxed systems whose decision criteria are obscured by complexity, trade secrecy, or limited transparency.Principled approaches can rigorously document machine-learning practices without presenting themselves as solutions.
- Computing as diagnostic: Diagnostic findings can affect investigated systems without resolving the underlying social problems.Following research on facial analysis, Microsoft and IBM reported improved accuracy across gender and racial lines, alongside continued research, advocacy, and policymaking.
- Caveats: Computational diagnosis should be combined with disciplines such as STS, sociology, economics, and ethnography to analyze sociotechnical systems holistically.These approaches capture social and organizational dimensions that computational methods may address poorly.
- Caveats: Metrics can become targets, causing their usefulness as measures to deteriorate over time.The paper uses the 4/5ths rule as an example of a diagnostic guideline becoming an enforcement target that can displace attention from other forms of discrimination and bias.
- Caveats: Diagnostic approaches may be constrained when technical systems are insulated from scrutiny or when legal restrictions limit researchers’ activities.Studies of proprietary news curation, resume search, and advertising systems must work around institutional constraints.
- Caveats: Precisely identifying a social problem does not itself ensure that relevant parties will remedy it.The paper argues that audit data require narrative tools because data do not necessarily change hearts, minds, or policy.
3 COMPUTING AS FORMALIZER
Computing acts as a formalizer by requiring explicit choices about how social problems, goals, inputs, and constraints are represented. This can clarify and contest policy judgments, but formalization may also distort values or divert attention from substantive goals.
- Computing as formalizer: Formalization requires explicit choices about a problem’s inputs, objectives, constraints, and assumptions, shaping how the problem is understood.Computational models replace generalized standards with more explicitly specified, rule-based models whose rankings reflect concrete judgments.
- Computing as formalizer: Formalization cannot replace substantive policy debate, and computational approaches may lack dimensions captured by ethnography, sociology, economics, and STS.The paper distinguishes procedural values such as transparency and accountability from decisions about what rules should actually accomplish.
- Computing as formalizer: Formalizing policy goals can force stakeholders to state their objectives more precisely, creating opportunities for contestation and democratic deliberation.The model-building process can provide intermediate targets for advocacy and discussion among officials, stakeholders, and the public.
- Computing as formalizer: Pretrial risk tools illustrate how formalization can conflate legally distinct risks, such as missing an appointment and absconding or causing violence.Such conflation is a substantive decision that can stigmatize accused people and operate to their detriment.
- Computing as formalizer: Different objective functions can produce very different resource allocations even when the broad policy goal remains the same.The paper uses subsidy allocation after income shocks to illustrate how minsum and alternative objectives encode different priorities.
- Computing as formalizer: Formalization is not necessarily transparent, accountable, or inclusive, and available data may favor measures that distort values or omit important aspects.Technical concerns can draw attention away from underlying policy goals, while powerful institutions may focus moral evaluation on individuals.
4 COMPUTING AS REBUTTAL
Computing acts as rebuttal by challenging entrenched assumptions about what policies and technical systems can accomplish. It can expand the policy imagination or establish rigorous limits, while its critiques risk redirecting attention toward technical improvement or away from useful interventions.
- Computing as rebuttal: Technical experts can use critiques of computational capability and neutrality to challenge political uses of technology and support broader change.The paper treats recognizing technical work’s political valence as a way to contest claims that computational systems are neutral or inherently beneficial.
- Computing as rebuttal: Formal results can prove that some desired algorithmic guarantees are impossible when groups have different base rates.For risk scores, calibrated predictions necessarily produce disparities in inaccurate high- and low-risk labels across groups with differing base rates.
- Computing as rebuttal: Critiques of technical systems can expose limits in policy frameworks that assume fairness is achieved by changing assessments at discrete decision points.The paper identifies frameworks borrowed from discrimination law, including protected categories and the 4/5ths rule, as examples.
- Computing as rebuttal: Computing can broaden political debate by revealing achievable outcomes that policy frameworks or allocation practices treat as unavailable.Matching research and related work can challenge narrow assumptions about the constraints governing scarce-resource allocation.
- Computing as rebuttal: Computational research has established infeasibility in secure fully electronic voting and secure encryption backdoors, contributing to resistance to broad surveillance mandates.These results illustrate how technical limits can challenge proposed policies rather than merely improve implementation.
- Computing as rebuttal: Rebuttal can misdirect policy toward improving a tool, or lead policymakers to reject computational approaches that could still have a positive role.The paper warns that criticizing imperfect performance does not by itself establish that all use should stop.
- Computing as rebuttal: Facial-recognition research distinguishes reducing algorithmic error from restricting a use that could amplify oppression even if accuracy were perfect.The paper notes that the appropriate alternative to a bad algorithm may sometimes be no algorithm at all.
5 COMPUTING AS SYNECDOCHE
Computing can make entrenched social problems newly visible by focusing public attention on their technological manifestations. This focus can create a tractable route into broader problems, but it risks narrowing attention to technical aspects and obscuring other causes and responses.
- Automated public-service systems have made poverty, homelessness, and child-welfare scrutiny newly visible through their concrete technical operations.Examples include welfare eligibility determination, housing allocation, and risk modeling for child abuse and neglect.
- Computing acts as a synecdoche when a technological part stands in for a larger social problem in discourse and critique.This tractable focus can renew attention to complex problems that may otherwise receive less public notice.
- Technology-focused framing can attract resources and audiences that broader critiques of poverty policy might not receive.Eubanks’s work received attention from different spheres as inequality through technology rather than inequality in general.
- A computing lens can mask other facets of a problem and divert political capital from non-technical responses.The risk is especially acute when technical attention restricts inquiry to technological aspects alone.
- Computing-centered critique can exploit computing’s public prominence without challenging its broader hegemony and may reinforce computational framings.The paper identifies a tenuous balance between avoiding overemphasis on technology and recognizing how technologies reinforce social systems.
- Technologies may fit particular sociopolitical arrangements, as content-moderation systems closely fit exploitative labor practices based on their scale and speed.The paper uses technology’s compatibility with social systems to show why computing cannot simply be treated as incidental to surrounding power arrangements.
6 CONCLUSION
The paper concludes that technical work may appear to favor incremental social change, yet computing research can also support more fundamental changes. It identifies four roles—diagnosis, formalization, rebuttal, and salience—while treating computing as one component of broader social and political work.
- Computing research can support fundamental social and political change through diagnosis, formalization, rebuttal, and synecdoche.These roles respectively characterize problems, shape their definitions, illuminate technical limits, and make long-standing problems newly salient.