CONCEPT Cited by 1 source

Utility function optimization

Definition

Utility function optimization in system design refers to selecting actions by maximizing a composite function that explicitly balances multiple competing objectives — typically positive engagement signals against negative cost signals.

In notification/messaging systems, this takes the form:

U(member, action) = Σ wₖ · Rewardₖ(member, action) − Cost(action)

Where: - Positive signals: likelihood of valuable engagement (views, clicks, satisfaction) - Negative signals: fatigue propensity, opt-out risk - Universal message cost: an empirically-tuned additive cost per send that keeps the function concave and prevents degenerate "always send" policies

The universal cost is necessary because sparse negative feedback alone under-penalizes over-messaging. Tuning uses a combination of online A/B experiments and offline evaluation metrics.

Seen in

• sources/2026-06-19-netflix-thinking-fast-slow-for-a-personalized-notification-system — Netflix's Slow Policy maximizes a utility function trading engagement signals against cost+fatigue signals, with ~O(100) discrete actions representing push+email frequency combinations.