Position: Every Ground Truth is a Human Construction, not an Objective Truth

Jul 16, 2026

Upload by Ivy

Ground truth datasets play a foundational role as reference values for training and evaluating machine learning models. This position paper argues that ground truths are not neutral objective measurements provided by nature, but are constructed through human decisions and technological arrangements. Highlighting the situated, context-dependent character of reference datasets improves reliability by clarifying where, when, and how datasets and the models shaped by them should be used. The paper calls for “situated reliability,” emphasizing limits and truth claims, and for greater transparency, accountability, and interdisciplinary collaboration.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

Introduction

Ground truth as constructed

Black-box decisions and contingency

Ground truthing and definition

Construction pathways and implications

Position: Every Ground Truth is a Human Construction, not an Objective Truth Charlotte Hgberg 1 Ericka Johnson 2 Kiri L. Wagstaff 3 arXiv :2607 .09668v1 [ cs .LG] 28 May 2026 Abstract Ground truth datasets play a fundamental role as reference values in the training and evaluation of machine learning models. This position paper argues that ground truths are not neutral objective measurements that are naturally given, but instead that they are constructed by arrangements of humans and technologies. We argue that the ML community will benefit from articulating and discussing these often invisible or unreported choices and acknowledging that reference datasets are contingent, not universal. Focusing on the situated and context-dependent nature of ground truths can improve reliability by enabling a better informed perspective on where, when, and how the datasets, and the models they have shaped, can best be used. We argue for increasing ‘situated reliability’ which includes articulating the limits and strengths of models and their truth claims. Finally, paying more attention to the construction of ground truths can support transparency, accountability, and interdisciplinary work. 1. Introduction In machine learning (ML) research and development, the term “ground truth” commonly refers to datasets viewed as containing the true values of a given concept (Kang, 2023) that are used to train and evaluate ML models. An often overlooked aspect is that these ground truths are not neutral objective measurements naturally given. All of our “truths” are constructed. We use ground truth as a countable noun in this paper to emphasize that a ground truth for an ML project is only one of many potential ground truths. Constructing aground truth encompasses work beyond just aligning with domain knowledge, yet these additional choices and their consequences are often not discussed or reported. 1Lund University, Lund, Sweden 2Linkping University, Linkping, Sweden 3 Oregon State University, Corvallis, OR, USA. Correspondence to: Charlotte Hgberg <char[lotte.hogberg@lth.lu.se](lotte.hogberg@lth.lu.se) >. Proceedings of the 43 rd International Conference on Machine Learning, Seoul, South Korea. PMLR 306, 2026 . Copyright 2026 by the author(s) . Ground truth construction pathways Where / when / whom to observe How (what features) to observe Who defines the label ontology Who labels the observations [And so on] GT1 GT2 GT3 Figure 1. Every ground truth is constructed as the result of multiple decisions, not a single objective truth. Different pathways yield different ground truths (GT1, GT2, GT3, ...) for the same task. Our position is that ground truths are constructed, and that ML researchers should articulate the choices involved in their construction and discuss the impact those choices have on their results. The concept of ground truth has been defined in various ways. The term has been used in remote sensing since the 1960s and been adopted across fields such as computer science and engineering, psychology, neuroscience, and economics (Woodhouse, 2021) . One view regards ground truth as the fundamental, real, or underlying facts collected at the source. In remote sensing, it has been defined as “information obtained by direct measurement at ground level, rather than by interpretation of remotely obtained data”(Woodhouse, 2021) . A more inclusive definition is “information obtained by direct observation of a real system, as opposed to a model or simulation; a set of data that is considered tobe accurate and reliable, and is used to calibrate a model, algorithm, procedure”(Woodhouse, 2021) . Kang (2023) argues that the concept basically refers to information assumed to be true in the development of ML models. We follow Jaton’s (2017) definition of ground truths as the result of ground truthing, which is the practice of defining the problem to be solved and what the input values and desired output targets should

FAQs

What does the paper argue about ground truth datasets in machine learning?

Ground truths are constructed rather than neutral objective measurements. They result from multiple human and technological decisions that shape reference values used in ML.

How does viewing ground truth as contextual help with reliability?

Focusing on situated and context-dependent properties improves reliability by clarifying the limits and best-use conditions of datasets and the models trained or evaluated with them.

What is “situated reliability” and why is it important?

“Situated reliability” involves articulating the strengths and limits of models and their truth claims. It supports more transparent, accountable, and interdisciplinary use of ML evidence.

Jul 16, 2026

Upload by Ivy