The inevitable weakness of metrics

2026-06-19 · Source: MIT Technology Review · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Intermediate, long

Summary

The article explores the inherent weaknesses of metrics, drawing on the author's decade-long experience with self-quantification and philosophical concepts like "value capture" by C. Thi Nguyen. It argues that while metrics can serve vital functions, they often obscure or corrupt nuanced goals, leading to a phenomenon where the measure itself replaces the original intent. The author's journey with devices like Fitbit and social media analytics (e.g., Chartbeat) revealed that quantifying life's minutiae did not lead to self-knowledge but rather redefined what was important, shifting focus from qualitative goals to hitting numerical targets like 20,000 daily steps. The piece highlights that institutional quantification simplifies complex realities for legibility and aggregation, stripping nuance and causing individuals and organizations to outsource their values to external metrics such as Yelp ratings or GPAs. This numeric worldview, where "knowing has become numeric," risks devaluing human experience and purpose, especially in the age of AI.

Key takeaway

For Directors of AI/ML or AI Ethicists designing performance metrics, recognize that quantification inherently simplifies and can lead to "value capture." Your systems risk redefining user or organizational goals if metrics become the sole target. Critically question the belief that numbers fully capture human needs and desires. Prioritize qualitative understanding alongside quantitative data to avoid inadvertently outsourcing core values to easily measurable but potentially misleading scores.

Key insights

Metrics, while useful, inherently simplify complex goals, leading to "value capture" where the measure replaces the original intent.

Principles

• Measurement inevitably begets more measurement.
• Metrics redefine your core sense of what's important.
• Institutional quantification simplifies for legibility.

In practice

• Question the belief that numbers fully reflect human needs.
• Re-evaluate personal and institutional reliance on metrics.
• Prioritize qualitative understanding over simple scores.

Topics

• Value Capture
• Self-Quantification
• Metric Design
• Goodhart's Law
• Data Philosophy
• AI Ethics

Best for: Consultant, Director of AI/ML, AI Ethicist

Related on AIssential

• Overconfidence at Scale — AI scales decisions exponentially, but human responsibility does not, creating systemic risks from scaled, flawed assumptions.
• AI Values Institute Launches New White Paper Exploring the AI Value Alignment Gap — The AI Value Alignment Gap highlights the critical disconnect between AI's generated value and an organization's stated human-centered values.
• Gig workers are endlessly exploited. AI could make more of us share their fate — AI adoption is driving a broad "gigification" of the workforce, eroding traditional employment benefits for cost savings.
• A Framework for Measuring Appropriate Reliance on Set-Valued AI Advice — Appropriate reliance on AI advice requires distinct metrics for set-valued predictions, separating quantity from quality of reliance.
• UC Berkeley bans AI use for law students — AI's rapid advancement creates both transformative potential and urgent ethical, economic, and infrastructural challenges.
• The “End of Coding” is a Myth. The Reality is Much Harder — AI shifts the bottleneck from code generation to ensuring quality, trust, and business alignment in complex systems.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by MIT Technology Review.