[Paper Review] Algorithmic Injustices: Towards a Relational Ethics

Abeba Birhane, Fred Cummins|arXiv (Cornell University)|Dec 16, 2019

Ethics and Social Impacts of AISocial Sciences19 references41 citations

TL;DR

The paper argues for relational ethics that centers vulnerable groups in addressing algorithmic injustice, beyond purely technical fixes.

ABSTRACT

It has become trivial to point out how decision-making processes in various social, political and economical sphere are assisted by automated systems. Improved efficiency, the hallmark of these systems, drives the mass scale integration of automated systems into daily life. However, as a robust body of research in the area of algorithmic injustice shows, algorithmic tools embed and perpetuate societal and historical biases and injustice. In particular, a persistent recurring trend within the literature indicates that society's most vulnerable are disproportionally impacted. When algorithmic injustice and bias is brought to the fore, most of the solutions on offer 1) revolve around technical solutions and 2) do not focus centre disproportionally impacted groups. This paper zooms out and draws the bigger picture. It 1) argues that concerns surrounding algorithmic decision making and algorithmic injustice require fundamental rethinking above and beyond technical solutions, and 2) outlines a way forward in a manner that centres vulnerable groups through the lens of relational ethics.

Motivation & Objective

• Acknowledge that data-driven tools encode human culture, values, and morality and thus require more than technical solutions.
• Center the disproportionately affected individuals and groups when designing and evaluating algorithms.
• Critically examine underlying assumptions about bias, justice, and ethics as dynamic and context-dependent.
• Promote understanding and contextual analysis over mere prediction in algorithmic systems.

Proposed method

• Advocate for relocating ethics from purely technical solutions to philosophical and social inquiry.
• Emphasize relational perspectives that consider data as tied to culture, meaning, and social power.
• Propose treating biases and fairness as dynamic, contingent concepts needing ongoing revision.
• Argue that algorithms shape social norms and orders, not just perform classifications.
• Recommend continual reframing of what constitutes fairness and ethics as society evolves.

Experimental results

Research questions

• RQ1 How should algorithmic injustice be addressed beyond technical fixes to center those most affected?
• RQ2 What does relational ethics require when evaluating and deploying algorithmic systems in socially contested domains?
• RQ3 How do concepts like bias, fairness, and justice change over time and across contexts?
• RQ4 In what ways do algorithmic classifications contribute to social order, and how can this be addressed?

Key findings

• Relational ethics centers disproportionately impacted groups as the starting point for any solution.
• Understanding and contextual grounding should precede predictive modeling in sensitive domains (e.g., prisons).
• Algorithms are not just tools but participants in creating and reinforcing social norms and orders.
• Fairness and ethics are moving targets that require open, iterative re-evaluation across contexts and time.
• Technological fixes alone are insufficient to address deep-rooted biases and injustices in society.