arXiv cs.AI· Jingyao Wu, Mohammad Tariqul Islam, Per R{\aa}dberg Nagb{\o}l, Julie Gerlings, Giovanni Leoni, Georgiana Fryza, Lars Pilegaard Thomsen, Jakob Mainz·· 9 天前AI 评分24
谁的标准答案？以人为中心的 AI 应拥抱模糊性
Whose Ground Truth? Embracing Ambiguity in Human-Centered AI
AI 导读
一篇被 NeurIPS 2026 Trustworthy AI for Good Workshop 接收的立场论文呼吁，以人为中心的 AI 应从追求单一标准答案转向建模合理解释空间。作者指出，传统机器学习常把人类标注差异通过聚合消除或当作噪声，但许多任务中同一输入的多种解释可同时合理，压缩为单一目标会丢失人类感知与判断的多样性信息。该视角应贯穿 AI 系统的表示、学习、评估、部署与治理。
正文 · 原文
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.
| Comments: | Accepted to NeurIPS 2026 Trustworthy AI for Good Workshop |
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.10805 [cs.AI] |
| (or arXiv:2610.10805v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.10805 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jingyao Wu [view email]
[v1]
Wed, 7 Oct 2026 19:05:31 UTC (1,554 KB)
来源：arXiv cs.AI · arxiv.org