# The Problem with Metrics is a Fundamental Problem for AI

arXiv: https://arxiv.org/abs/2002.08512 · HTML: https://mlsanity.com/paper/2002.08512/

## Key facts

- Cited by 37 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 20 Feb 2020.

## Built on by (most cited first, 30 of 37)

- [Defining and Characterizing Reward Hacking](https://mlsanity.com/paper/2209.13085/) (2022)
- [Large language models can accurately predict searcher preferences](https://mlsanity.com/paper/2309.10621/) (2023)
- [Evaluation Gaps in Machine Learning Practice](https://mlsanity.com/paper/2205.05256/) (2022)
- [Eliciting and Learning with Soft Labels from Every Annotator](https://mlsanity.com/paper/2207.00810/) (2022)
- [What are you optimizing for? Aligning Recommender Systems with Human Values](https://mlsanity.com/paper/2107.10939/) (2021)
- [Reality Check: A New Evaluation Ecosystem Is Necessary to Understand AI's Real World Effects](https://mlsanity.com/paper/2505.18893/) (2025)
- [Aligning Offline Metrics and Human Judgments of Value for Code Generation Models](https://mlsanity.com/paper/2210.16494/) (2022)
- [Promises and pitfalls of artificial intelligence for legal applications](https://mlsanity.com/paper/2402.01656/) (2024)
- [Scaling Laws Do Not Scale](https://mlsanity.com/paper/2307.03201/) (2023)
- [Benchmark Data Repositories for Better Benchmarking](https://mlsanity.com/paper/2410.24100/) (2024)
- [GPT for Games: An Updated Scoping Review (2020-2024)](https://mlsanity.com/paper/2411.00308/) (2024)
- [Predictable Artificial Intelligence](https://mlsanity.com/paper/2310.06167/) (2023)
- [A Review of the Evidence for Existential Risk from AI via Misaligned Power-Seeking](https://mlsanity.com/paper/2310.18244/) (2023)
- [The Inadequacy of Offline LLM Evaluations: A Need to Account for Personalization in Model Behavior](https://mlsanity.com/paper/2509.19364/) (2025)
- ["This is not a data problem": Algorithms and Power in Public Higher Education in Canada](https://mlsanity.com/paper/2403.13969/) (2024)
- [Identifying the Context Shift between Test Benchmarks and Production Data](https://mlsanity.com/paper/2207.01059/) (2022)
- [Quantitative AI Risk Assessments: Opportunities and Challenges](https://mlsanity.com/paper/2209.06317/) (2022)
- [Beware of "Explanations" of AI](https://mlsanity.com/paper/2504.06791/) (2025)
- [MassSpecGym in the Wild: Uncovering and Correcting Evaluation Pitfalls in AI-Driven Molecule Discovery](https://mlsanity.com/paper/2606.19624/) (2026)
- [Mapping the Challenges of HCI: An Application and Evaluation of ChatGPT for Mining Insights at Scale](https://mlsanity.com/paper/2306.05036/) (2023)
- [Beyond Accuracy: EcoL2 Metric for Sustainable Neural PDE Solvers](https://mlsanity.com/paper/2505.12556/) (2025)
- [The Strong, Weak and Benign Goodhart's law. An independence-free and paradigm-agnostic formalisation](https://mlsanity.com/paper/2505.23445/) (2025)
- [The Different Faces of AI Ethics Across the World: A Principle-Implementation Gap Analysis](https://mlsanity.com/paper/2206.03225/) (2022)
- [Challenges in Explanation Quality Evaluation](https://mlsanity.com/paper/2210.07126/) (2022)
- [Positive AI: Key Challenges in Designing Artificial Intelligence for Wellbeing](https://mlsanity.com/paper/2304.12241/) (2023)
- [Benchmarking and Evaluation of AI Models in Biology: Outcomes and Recommendations from the CZI Virtual Cells Workshop](https://mlsanity.com/paper/2507.10502/) (2025)
- [Against Proxy Optimization](https://mlsanity.com/paper/2606.23597/) (2026)
- [Manipulation-Proof Oblivious Audits against Deceptive Model Providers](https://mlsanity.com/paper/2608.04365/) (2026)
- [Branching Out: Broadening AI Measurement and Evaluation with Measurement Trees](https://mlsanity.com/paper/2509.26632/) (2025)
- [White-Box Sensitivity Auditing with Steering Vectors](https://mlsanity.com/paper/2601.16398/) (2026)

## Abstract (arXiv)

Optimizing a given metric is a central aspect of most current AI approaches, yet overemphasizing metrics leads to manipulation, gaming, a myopic focus on short-term goals, and other unexpected negative consequences. This poses a fundamental contradiction for AI development. Through a series of real-world case studies, we look at various aspects of where metrics go wrong in practice and aspects of how our online environment and current business practices are exacerbating these failures. Finally, we propose a framework towards mitigating the harms caused by overemphasis of metrics within AI by: (1) using a slate of metrics to get a fuller and more nuanced picture, (2) combining metrics with qualitative accounts, and (3) involving a range of stakeholders, including those who will be most impacted.