Cognition · Data artifact

Utility Function Specification

Data artifactCognitionCognition & Memoryarc:UtilityFunctionSpecification

A declarative specification mapping outcomes to utility values: the objectives, their per-objective transforms, trade-off weights and risk-attitude shape.

Responsibility. Encodes a principal's preferences as a utility function used to rank outcomes.

Also known as: Utility function, Multi-objective utility function, Objective weights, Preference model, Cost function, Preference weights

Direct neighbourhood (hover for relationship types)

Relationships

configures structural

Utility-Based Decision Maker Ch5.10 Ch5.12 +1

is configured by structural

Risk Attitude Policy abstract Ch5.10

is read by dependency

is written by dependency

Preference Elicitor abstract Ch5.10

receives data from dynamic

is evaluated by assurance

Evaluation Harness Ch5.10

Design guidance

SHOULD identify objectives, define quantitative metrics, set trade-off weights and validate resulting behaviour before deployment.
SHOULD treat objective weights as context-dependent parameters explored via sensitivity analysis rather than fixed constants.
MUST be validated against historical decisions and simulated scenarios before production deployment.
SHOULD NOT default to linear (risk-neutral) utility; the shape must reflect stakeholder risk attitudes.

Quantitative guidance

As stated by the sources; verify before use.

Example weights: AV 0.5 safety/0.3 speed/0.2 comfort; ride-sharing variant 0.6 safety/0.25 speed; recommendation 0.7 relevance/0.3 diversity; procurement 0.5 cost/0.3 reliability/0.2 speed; grid 0.6 cost/0.4 reliability (Ch5.10).
Raising procurement cost weight to 0.7 flips selection from supplier B (EU 0.765) to supplier A (EU 0.78) (Ch5.10).
Worked example: U(route) = 0.5 time + 0.3 safety + 0.15 comfort + 0.05 cost (Ch5.13).

Classification

Patterns: Weighted linear combination u = sum w_i f_i(o_i)von Neumann-Morgenstern axioms (completeness, transitivity, continuity, independence)Reward engineering / utility design
Quality attributes: Maintainability (ISO/IEC 25010)Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)
Risks mitigated: Misspecified utility causing unintended behaviourUnintended negative externalities on uncaptured objectivesNon-monotonic utility preference reversals

Sources

Ch5.10: T. Nguyen, "Utility-Based Decision Making Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.10. ISBN: 9798244538229.
Ch5.12: T. Nguyen, "Learning-Based Decision Making Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.12. ISBN: 9798244538229.
Ch5.13: T. Nguyen, "Hybrid Decision Systems Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.13. ISBN: 9798244538229.