Whose Ground Truth? Embracing Ambiguity in Human-Centered AI

arXiv:2610.10805v1 Announce Type: new Abstract: As AI systems increasingly interact with people and make decisions about them, understanding human interpretations becomes an important part of developing human-centered AI. Conventional machine learning and AI systems are largely developed under the assumption that a single definitive ground truth exists, with variability in human annotations often resolved through aggregation or treated as noise. However, for many human-centered tasks, human interpretation is inherently ambiguous, and multiple interpretations of the same input may be simultaneously reasonable and valid. Reducing such ambiguity to a single target risks overlooking meaningful information about the diversity of human perception, judgment, and experience. In this position paper, we call for a shift towards modeling the interpretation space of plausible human judgments, while distinguishing meaningful ambiguity from annotation noise. We argue that this perspective should guide how AI systems are represented, learned, evaluated, deployed, and governed, supporting more human-centered AI that better reflects the diversity of human interpretation.

Calls for shift in AI development to account for ambiguity in human interpretation.

The paper is a theoretical call to action with no empirical deployment or measured outcome, so its effect on human flourishing is indirect and uncertain. The direction is mildly positive if adopted, since it favors preserving human interpretive diversity over forced consensus.

This paper advocates for a conceptual shift in AI development, proposing that systems should account for the inherent ambiguity in human interpretations rather than assuming a single ground truth. It suggests that this change would lead to more human-centered AI. While the goal is positive, it is a conceptual argument and not a report of an implemented system or its measured impact.

The paper advocates for AI systems that embrace ambiguity in human interpretation. This could lead to more respectful and effective AI interactions. The paper does not discuss risks or negative implications.

SOURCE     arXiv cs.AI (tier 1)
   ↓
DOCUMENT   123ab2bd-65f1-450b-b60b-cb3b11ac80ca
           https://arxiv.org/abs/2610.10805
   ↓
EVIDENCE   2 extracted excerpts
   ↓
MODEL RUN  4 runs, methodology 1.0.1
   ↓
SCORE      -7  (Mixed or uncertain)
   ↓
CONFIDENCE 65%