Is Environmental Science Less Objective?

An analysis of the use of subjective language in scientific papers.

An interesting paper, "Science versus Subjectivity in Environmental Research," by Dominic Parker and Ngoc Ha Do at the University of Madison. Here is the abstract:

Objective science aims to uncover truths about environmental conditions, including those related to emission trends, biodiversity loss, climate change, and their economic costs. When and why do scientists use subjective language to describe these conditions with bias, emotion, or advocacy? To shed light, we study determinants and trends in subjective language within the abstracts, introductions, and conclusions of over 170,000 peer-reviewed articles published during 1990-2023 in top economics and science journals. Using machine learning to label and score text by subjectivity, we find that subjective language is more prevalent in articles focused on environmental topics relative to non-environmental topics: by 12% in economics and by 42% in science. After accounting for factors such as article age, author experience, number of authors, and journal quality, higher subjectivity scores associate positively with future academic citations in both science and economics, especially if the article is focused on an environmental topic. These findings, which imply subjective language has been rewarded in the academy, are consistent with informal theory suggesting subjective language will be used more frequently by scientists when expected rewards are high and risks are low.

The obvious question is whether this is just a finding about the use of language, or whether it indicates something about the underlying work.

Comments

[Stephen Lathrop]

The file with the entire paper will not load for me.

Two questions. and a point.

The first question: I ask it because I doubt an assertion in the abstract relating most to economics, but applicable across the entire range of academic research. It is about reliance on correlations. Do the authors class as unobjective suggested correlation-based explanations which lack proof of causation? How about if those explanations result from insight into correlations among data ranked objectively to be of reliable quality? If so, then I insist that tears a hole in their reasoning about objectivity.

Data useful in science can be—and often are—accurately judged to have been objectively collected. That can turn into a distraction to confuse their users about conclusions. Objective conclusions about cause and effect can require—and too often do require— standards for relevance which too often prove impossible to meet objectively.

The second question: Some sciences—ecology, meteorology, physical oceanography, climatology come to mind, there are many others—take for subject matter processes implicating so many hard-to-predict interactions among incompletely understood sets of variables that scientific objectivity demands cautious phrasing about suggested results. That caution counts more in favor of scientific reliability than against it.

The point: So what?

Natural sciences of perfectly objective pedigree—geology and zoology, for instance—were for centuries almost entirely observational in character. That did not prevent Agassiz from using both to arrive at an admirably objective conclusion that startled the world: that the face of much of the earth bore evidence of recent ice ages. Rarely if ever has, "objective," science delivered insight to compete with that one for influence on human affairs, extending as it did even to questions of religious belief.

The science most urgently in need of present advancement is one destined to remain mostly observational for at least decades, and maybe forever. It depends for its data on observations so numerous, and for its reasoning on interactions so ramified and permeated with happenstance, that it all but makes a mockery of the notion of scientific objectivity.

It is the science of ecology. Yet, to give ecology the influence in human affairs which the fate of humanity requires it to have, has become a stumbling block in the practice of that least objective of human activities, politics.

Adler's pointer to the abstract above will not advance much objective insight if it turns predictably into yet another self-regarding STEM-insistent argument for yet more cultural dominance by, "objective," scientists and engineers. When it comes to objectivity about human affairs, those too often do not know what they are talking about.

[Drewski]

And as we all know, LLMs are almost as good at stastical analysis as they are at law or counting. This paper has as much meaning as a dart thrown by a blindfolded drunk.