Concern 249 · Working draft · editorial review in progress
Over-optimization
Over-optimization occurs when pushing a performance measure harder makes it a worse guide to the outcome people actually want.
Understanding the concern
A score may track quality reasonably well within familiar conditions. Strong optimization can drive the system toward unusual cases where that relationship breaks down. A reward model can then report improvement even as an independent measure of quality falls.
This is related to Goodhart's law, a warning about turning measures into targets. It differs from deliberate tampering because the scoring system can remain unchanged. Independent evaluation, limits on optimization, and attention to real outcomes help reveal when the score has stopped being informative.
An example
A school improves a narrow test score by dropping valuable learning that the test does not measure.
Illustrative unless identified as a documented case.
Words in plain English
- Model
- The learned part of an AI system. It uses patterns found during training to produce an answer or prediction.
- Reward
- A signal used in some training methods to favor certain behavior. A high reward score does not necessarily mean a good result for people.
Evidence and its limits
The study examines reward-model optimization and independent quality estimates under tested training conditions; the school example is an analogy.
Gao, Schulman, and Hilton · ICML 2023 · Research study
Scaling Laws for Reward Model Overoptimization ↗ (opens in a new tab)Examines optimization of an imperfect proxy reward with an experimental reference reward; not a complete model of human values.
Broader frameworks and background
These resources provide context. They do not establish every claim in this concern.
A question to discuss
At what point does further improvement on this metric stop improving the underlying outcome?
Keep track of different views, what needs checking, and any next step in your own discussion notes.
Open discussion notes →Principles for the conversation
Before you finish · reflect together
Virtues and principles to consider
Virtues are qualities we practice in how we treat people and make choices. They are not scores for judging someone. Use these ideas to explore what a thoughtful response could look like.
Prudence
Prudence means using careful judgment before and during action. It balances worthwhile opportunities with the need to avoid harm that could have been anticipated.
Start with a limited trial, set stopping points, and prefer choices that can be undone when uncertainty is high. Review what happens rather than assuming the original plan remains sound.
Ask together: What is a careful next step, and when should we stop or change course?
Humility
Humility means recognizing the limits of our knowledge and being willing to learn from others. Expertise can be valuable without answering every question.
State where evidence is weak, invite criticism, and listen to people with different experience. Leave room to revise a decision when new information appears.
Ask together: What might we be missing, and who could help us see it?
Stewardship
Stewardship means caring for resources and systems that others depend on, including people who will use them later. It looks beyond the benefit of a quick launch.
Plan for maintenance, energy use, data protection, and eventual shutdown. Ask who will look after the tool when the original team moves on and how future users can avoid inheriting hidden costs.
Ask together: What are we responsible for caring for over time?
Editorial ratings
These are editorial judgments from the register, not measured probabilities or predictions about your situation.
About the register and scoring →More ways to explore
Search for more writing
For more specific or recent material, open a search below or ask for a web summary. AI-generated summaries can contain mistakes; open the original articles to check them.