Whose Ground Truth? Embracing Ambiguity in Human-Centered AI

查看原文 / View Original 编辑 / Edit

Computer Science > Artificial Intelligence arXiv:2610.10805 (cs) [Submitted on 7 Oct 2026] Title:Whose Ground Truth? Embracing Ambiguity in Human-Centered AI Authors:Jingyao Wu, Mohammad Tariqul Islam, Per Rådberg Nagbøl, Julie Gerlings, Giovanni Leoni, Georgiana Fryza, Lars Pilegaard Thomsen, Jakob Mainz View a PDF of the paper titled Whose Ground Truth? Embracing Ambiguity in Human-Centered AI, by Jingyao Wu and 7 other authors View PDF

计算机科学 > 人工智能 arXiv:2610.10805 (cs) [提交于 2026 年 10 月 7 日] 标题：谁的“基本事实”？在以人为本的 AI 中拥抱模糊性作者：Jingyao Wu, Mohammad Tariqul Islam, Per Rådberg Nagbøl, Julie Gerlings, Giovanni Leoni, Georgiana Fryza, Lars Pilegaard Thomsen, Jakob Mainz 查看标题为《谁的“基本事实”？在以人为本的 AI 中拥抱模糊性》的论文 PDF，作者为 Jingyao Wu 及其他 7 位作者，查看 PDF。

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.

摘要：随着人工智能系统越来越多地与人类互动并针对人类做出决策，理解人类的解读已成为发展以人为本的人工智能的重要组成部分。传统的机器学习和人工智能系统大多是在存在单一明确“基本事实”的假设下开发的，人类标注中的差异通常通过聚合来解决，或被视为噪声处理。

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.

然而，对于许多以人为本的任务而言，人类的解读本质上是模糊的，对同一输入的多种解读可能同时是合理且有效的。将这种模糊性简化为单一目标，可能会忽略有关人类感知、判断和经验多样性的重要信息。

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.

在这篇立场论文中，我们呼吁转向对合理人类判断的解读空间进行建模，同时将有意义的模糊性与标注噪声区分开来。我们认为，这一视角应指导人工智能系统的表示、学习、评估、部署和治理方式，从而支持更以人为本的人工智能，更好地反映人类解读的多样性。