Position: Every Ground Truth is a Human Construction, not an Objective Truth
(2026) 43rd International Conference on Machine Learning, ICML 306. p.1-13- 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 data sets 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... (More)
- 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 data sets 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. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/833b9fcf-5d41-41b0-ad8f-61bee0d79973
- author
- Högberg, Charlotte
LU
; Johnson, Ericka
and Wagstaff, Kiri L.
- organization
- publishing date
- 2026-04-30
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- in press
- subject
- keywords
- Ground truth, Machine learning, benchmark dataset, data
- host publication
- Proceedings of the 43 rd International Conference on Machine Learning, Seoul, South Korea
- volume
- 306
- pages
- 13 pages
- publisher
- PMLR
- conference name
- 43rd International Conference on Machine Learning, ICML
- conference location
- Seoul, Korea, Republic of
- conference dates
- 2026-07-06 - 2026-07-11
- DOI
- 10.48550/arXiv.2607.09668
- project
- When N equals 1: Ethics and epistemology of personalisation and group belongings in data-driven prediction medicine
- language
- English
- LU publication?
- yes
- id
- 833b9fcf-5d41-41b0-ad8f-61bee0d79973
- date added to LUP
- 2026-05-03 20:09:36
- date last changed
- 2026-08-10 10:09:52
@inproceedings{833b9fcf-5d41-41b0-ad8f-61bee0d79973,
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 data sets 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<br/>ground truths can support transparency, accountability, and interdisciplinary work.}},
author = {{Högberg, Charlotte and Johnson, Ericka and Wagstaff, Kiri L.}},
booktitle = {{Proceedings of the 43 rd International Conference on Machine Learning, Seoul, South Korea}},
keywords = {{Ground truth; Machine learning; benchmark dataset; data}},
language = {{eng}},
month = {{04}},
pages = {{1--13}},
publisher = {{PMLR}},
title = {{Position: Every Ground Truth is a Human Construction, not an Objective Truth}},
url = {{https://lup.lub.lu.se/search/files/253015983/Hogberg_Johnson_Wagstaff_Position_Every_Ground_Truth_is_a_Human_Construction_not_an_Objective_Truth.pdf}},
doi = {{10.48550/arXiv.2607.09668}},
volume = {{306}},
year = {{2026}},
}