Optimal Use of Preferences in Artificial Intelligence Algorithms

TL;DR

This paper establishes decision problem-agnostic conditions under which separating preference-free training from preference-based post-processing in AI systems is optimal, providing theoretical foundations for modular and flexible AI pipeline design.

Contribution

It introduces a diminishing-value-of-information condition that explains when preference embedding is beneficial, extending prior work by removing the need to specify downstream objectives.

Findings

Preference embedding reduces the value of information at the margin.

Preference-free training weakly dominates for any expected utility decision.

Automating threshold computation can be advantageous under cognitive constraints.

Abstract

Machine learning systems embed preferences either in training losses or through post-processing of calibrated predictions. Applying information design methods from Strack and Yang (2024), this paper provides decision problem agnostic conditions under which separation training preference free and applying preferences ex post is optimal. Unlike prior work that requires specifying downstream objectives, the welfare results here apply uniformly across decision problems. The key primitive is a diminishing-value-of-information condition: relative to a fixed (normalised) preference-free loss, preference embedding makes informativeness less valuable at the margin, inducing a mean-preserving contraction of learned posteriors. Because the value of information is convex in beliefs, preference-free training weakly dominates for any expected utility decision problem. This provides theoretical foundations for modular AI pipelines that learn calibrated probabilities and implement asymmetric costs through downstream decision rules. However, separation requires users to implement optimal decision rules. When cognitive constraints bind, as documented in human AI decision-making, preference embedding can dominate by automating threshold computation. These results provide design guidance: preserve optionality through post-processing when objectives may shift; embed preferences when decision-stage frictions dominate.