Puy Arnald, Bacon Ethan, Carmona Alba, Flinders Samuel, Gefen David, Khanjani Mohammad, Larsen Kai R, Lachi Alessio, Linga Seth N, Lo Piano Samuele, Melsen Lieke A, Murray Emily, Sheikholeslami Razi, Sobhani Ariana, Wei Nanxin, Saltelli Andrea
School of Geography, Earth and Environmental Sciences, University of Birmingham, Birmingham B15 2TT, UK.
Department of Modern Languages, College of Arts and Law, University of Birmingham, Birmingham B15 2TT, UK.
iScience. 2025 Mar 13;28(4):112184. doi: 10.1016/j.isci.2025.112184. eCollection 2025 Apr 18.
Several modern scientific fields rely on computationally intensive mathematical models to study uncertain, complex socio-environmental phenomena such as the spread of a virus, climate change, or the water cycle. However, the degree of epistemic commitment of these fields is unclear. By using machine learning to extract the knowledge claims of around 755,000 abstracts from 14 scientific fields spanning the human and physical sciences, we show that epidemic, integrated assessment, and water modeling display a degree of linguistic assertiveness akin to physics. Water modeling surpasses even the most accurate physical sciences in substantiating knowledge claims with numbers, which are largely produced without accompanying uncertainty and sensitivity analysis. By exploring the balance between doubt and certainty in academic writing, our study reflects on whether the strong conviction and quantification of fields modeling socio-environmental processes, especially water modeling, are epistemically justified.
几个现代科学领域依靠计算密集型数学模型来研究不确定的复杂社会环境现象，如病毒传播、气候变化或水循环。然而，这些领域的认知承诺程度尚不清楚。通过使用机器学习从涵盖人文科学和自然科学的14个科学领域中提取约755,000篇摘要中的知识主张，我们发现流行病、综合评估和水模型显示出与物理学相似的语言确定性程度。水模型在通过数字证实知识主张方面甚至超过了最精确的自然科学，而这些数字大多在没有伴随的不确定性和敏感性分析的情况下产生。通过探索学术写作中怀疑与确定性之间的平衡，我们的研究反思了对社会环境过程进行建模的领域，尤其是水模型，其强烈的信念和量化在认知上是否合理。
