Political Philosophy
What is new, and what is old, in fairness and machine learning
This article explores the issue of normative distinctiveness in machine learning decision systems, arguing that while machine learning does not pose uniquely new normative problems, its mere existence expands practical possibilities and places new justificatory demands on both new technical systems and prevailing human-centered decision regimes. It emphasizes that depoliticizing conventional bureaucratic structures leads to a missed opportunity for a broader normative reevaluation of what we owe to each other in a world of expanded practical possibility.
Executive Impact: Reframing AI's Ethical Challenge
Machine learning systems demand a new lens for evaluating ethical frameworks and institutional design, moving beyond simplistic comparisons to unlock transformative potential.
0%
Expanded Possibility
New avenues for institutional design
New avenues for institutional design
0%
Accountability Clarity
Improved tracing of system-wide outcomes
Improved tracing of system-wide outcomes
0%
Democratic Control Potential
Increased capacity for public input
Increased capacity for public input
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
The first thing that one learns about machine learning is that it is only as good as the data on which it relies. The first thing that one learns about fairness in machine learning is that, alas, those data are often not very good. They encode problematic social biases alongside useful predictive patterns. The difficulty of learning the one but not the other is what gives rise to the problem of bias in machine learning. Lacking a solution, the result is a reproduction of all of the patterns in the data: good, bad, and ugly.
0.7-0.8 AUC
Typical ML model accuracy for child abuse prediction, better than a coin toss.
Many concerns raised about machine learning (e.g., inaccuracy, bias) have equal moral significance for the legitimacy of human systems. The mistaken attribution of normative difference often stems from a perceived difference in accountability, but centralized ML systems can actually simplify causal tracing for system-wide outcomes.
Enterprise Process Flow
ML model availability
→
Increased measurability of accuracy
→
Accuracy becomes a planned design choice
→
New moral questions for institutional design
| Feature | Machine Learning Systems | Traditional Human Systems |
|---|---|---|
| Decision Structure |
|
|
| Accountability Tracing |
|
|
| Democratic Control Potential |
|
|
Hayek's error in distinguishing between market outcomes (spontaneous order) and planned economies (deliberate design) for social justice parallels how fairness discourse sometimes naturalizes decentralized human systems, exempting them from scrutiny while scrutinizing ML systems. This leads to a depoliticization of existing institutions.
Expanding Moral Possibility via ML
The advent of machine learning expands our practical possibilities, creating new choice points about morally significant matters. Just as ultrasound changed moral choices around pregnancy, ML forces us to proactively plan for values like accuracy in institutional design. This new capacity means we are morally responsible not just for ML system design, but also for the choice to cling to traditional human-centered systems when alternatives exist.
Quote: "Algorithms force us to be explicit about what we want to achieve with decision-making. And it's far more difficult to paper over our poorly specified or true intentions when we have to state these objectives formally. In this way, machine learning has the potential to help us debate the fairness of different policies and decision-making procedures more effectively."
Source: (21)
Quantify Your AI Advantage
Estimate the potential efficiency gains and cost savings for your organization by integrating advanced AI decision systems.
Estimated Annual Savings
$0
Reclaimed Annual Hours
0
Your AI Transformation Roadmap
A structured approach to integrating AI, ensuring ethical considerations and normative re-evaluations are at the core of every phase.
01 Discovery & Assessment (1-2 Weeks)
Initial consultation to understand your current bureaucratic decision-making processes, identify existing biases and inefficiencies, and assess data readiness for ML integration.
02 ML Feasibility & Normative Impact Analysis (3-4 Weeks)
Evaluate the technical feasibility of ML solutions for specific decision domains. Conduct a deep normative analysis comparing human-centered vs. ML-augmented systems, focusing on accuracy, fairness, and accountability implications.
03 Pilot Design & Ethical Framework Development (4-6 Weeks)
Design a pilot ML system, explicitly defining accuracy thresholds and other normative goals. Develop a robust ethical framework for deployment, incorporating insights on centralization, decentralization, and democratic participation.
04 Implementation & Justification Strategy (6-10 Weeks)
Deploy the ML pilot, with continuous monitoring for performance and fairness. Develop a public justification strategy for the chosen decision-making structure, addressing the expanded moral responsibilities introduced by ML capabilities.
Ready to Re-evaluate Your Institutional Design?
Connect with our experts to discuss how expanded practical possibilities and new justificatory demands can transform your organization's decision-making for a more just and efficient future.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat