Thomas Kuhn, in
his 1962 book, The Structure of Scientific Revolutions, noted that science is
in part based on consensus. The prevailing scientific views of a time period
are agreed upon by consensus. That consensus becomes the paradigm. Facts are
first discovered and teased out through experimentation. Then they must be
agreed upon by the best scientists in the field in a consensus. Consensus may
introduce some subjectivity into objective science if there is enough
uncertainty in the field.
Bias and censorship exist in
science as well. People have opinions. These opinions mostly don’t affect
science since it is based on facts, but when science influences policy or is
translated into policy, those opinions arise, and factuality becomes less
influential. Less plausible ideas are “censored” simply due to being less
plausible, as they should be. Scientists are most often biased toward the most
plausible consensus, the prevailing paradigm, which is generally, but not
always, a good thing.
Scientific Censorship
A 2023 PNAS paper explores
scientific censorship, defined as:
“…actions aimed at obstructing particular scientific ideas
from reaching an audience for reasons other than low scientific quality.”
The researchers found that
prosocial concerns such as self-protection, benevolence toward peer scholars,
and concern for the well-being of human social groups were motivating factors
for scientific censorship. They note that there is a clear need to improve
transparency and accountability in scientific decision-making. They also note
that the costs and benefits of these types of censorship need to be analyzed
and weighed, since some censorship is sometimes warranted. They note that
scientific censorship is difficult to detect and measure, so it is rarely
studied. In describing and defining censorship, they note:
“Censorship is distinct from discrimination, if not
always clearly so. Censorship targets particular ideas (regardless of their
quality), whereas discrimination targets particular people (regardless of their
merit).”
The following table
distinguishes types of censorship, who the censors typically are, the
motivations of censors, and the outcomes of scientific censorship.
They distinguish two types of
censorship: soft censorship and hard censorship. Hard censorship refers to
censorship from institutional authorities like governments and religious
authorities, and usually involves preventing dissemination or retraction. Soft
censorship can have different motivations, even benevolent ones like protecting
the researcher. It involves “social punishments or threats of them (e.g.,
ostracism, public shaming, double standards in hirings, firings, publishing,
retractions, and funding) to prevent dissemination of research.” It may be
mild, like simply discouraging certain research projects that might negatively
affect careers.
Censors include governments,
typically in authoritarian regimes, educational institutions such as
universities, journals, and professional societies, and more informal threats
of ostracism and reputational damage to both researchers and institutions. An
example of the first type is when a university in Hungary relocated to Austria
due to being censored by the Hungarian government. Educational donors can
threaten to withhold funding if they think the research is not what they want
to see. These kinds of deterrents also influence scientists to self-censor and
avoid controversial research. Most scientists report some kind of
self-censoring.
They also note that soft
censorship can be hard to distinguish from simple scientific rejection. They
also note that scientific rejection can be subjective and influenced by the
pressures noted above.
“…many criteria that influence scientific
decision-making, including novelty, interest, “fit”, and even quality are often
ambiguous and subjective, which enables scholars to exaggerate flaws or make
unreasonable demands to justify rejection of unpalatable findings.”
Thus, they note, bias and
censorship can be mistaken for genuine science-based rejection. Scientists
whose work was legitimately rejected may claim that they are being censored.
Peer review is another
process that can be biased:
“…peer reviewers evaluate research more favorably when
findings support their prior beliefs, theoretical orientations, and political
views.”
They also note that science
is designed to root out bias, which it generally does over time.
In exploring the psychology
of censorship, they note:
“Censorship research typically explores dark
psychological underpinnings such as intolerance, authoritarianism, dogmatism,
rigidity, and extremism. Authoritarianism, on the political right and left, is
associated with censoriousness, and censorship is often attributed to desires
for power and authority.”
They also give some
interesting information about modern levels of scientific censorship:
“Hundreds of scholars have been sanctioned for
expressing controversial ideas, and the rate of sanctions has increased
substantially over the past 10 y. Retractions of scientific articles have
increased since at least 2000, many for good reasons such as statistical
errors, but some were at least partly motivated by harm concerns.”
Some data is given below.
The graph below shows that
scientific censorship can lead to erroneous conclusions or at least less likely
conclusions.
They also note that when
science journals like Nature and Scientific American endorse political
candidates, they are really being censorious and that such actions can erode
trust in science.
“Scientific censorship appears to be increasing.
Potential explanations include expanding definitions of harm, increasing
concerns about equity and inclusion in higher education, cohort effects, the
growing proportion of women in science, increasing ideological homogeneity, and
direct and frequent interaction between scientists and the public on social
media.”
The authors write that peer
review was designed to be anonymous and confidential, but that this could
increase bias and censorship, rather than reduce it, as preferred. They suggest
opening up the process more. The goal is to eliminate bias and censorship
regarding the acceptance or rejection of papers. They also think that
scientific journals and institutions should be audited for procedural
unfairness. Such audits and evaluations could make academic journals more
competitive. They also call for better documentation and better data
availability for retractions.
Expertise and Policy
This section is a review of
the ideas of Daniel Sarewitz and the late Steve Rayner on the subject. The
article, published in 2021, was mostly written before COVID hit. They speak of
a “post-truth condition,” due to science illiteracy, populist politics,
and the proliferation of unverifiable information via the Internet and social
media. Expertise in the form of skills like those of a surgeon or a pilot is
not being contested, but only certain types of expertise:
“Clearly, what is contested is not all science, all
knowledge, and all expertise, but particular kinds of science and claims to
expertise, applied to particular types of problems.”
They note that science is
limited in the kinds of questions it can answer and the types of problems it
can solve. The physicist Alvin Weinberg argued in an influential 1972 article
that certain questions transcend the ability of science to answer them. These
questions typically involve complex and socially divisive topics. They also
note that “risk” has become a more prominent concern in modern society.
Concerns about public health and environmental health rise in developed and
wealthy societies. Risk acceptance or rejection is often determined by
scientists and policymakers.
“It is thus no coincidence that the 1980s and 90s saw
“risk” emerge as the explicit field of competing claims of rationality.”
“…starting in the 1970s, there has been a rapid expansion
in health and environmental disputes, not-in-my backyard protests, and concerns
about environmental justice, invariably accompanied by dueling experts, usually
backed by competing scientific assessments of potential or actual damage to
individuals and communities. These types of disputes constitute an important
dimension of today’s divisive national politics.”
Scientists interpret nature,
and speak for it, they note, in order to advise policymakers. Political
divisiveness has made it so that there are experts who speak for each political
viewpoint. Thus, even views of nature can vary due to political persuasions. As
an example, he cites Johan Rockstrom and colleagues' “planetary boundaries” way
of looking at environmental issues vs. challenges to those ideas by more
pragmatic writers like Ted Nordhaus, who argued that the thresholds they chose
are not “non-negotiable,” but arbitrary. I have argued similarly. This is an
important example because it shows that catastrophism in the form of ideas like
crossing planetary boundaries, tipping points, ecosystem collapse, etc., has
gained traction in recent years. Such pessimistic views, especially if not
warranted, can distort science.
The authors also talk about
the power of numbers that back up scientific claims, but also can distort them
if what is being measured is more abstract than concrete. They give two main
examples. The first is the prediction of the percolation flux of groundwater
that may encounter nuclear waste stored underground. The second is the idea of
climate sensitivity, which has remained more or less in the same predicted
range of 1.5-4.5 deg C.
“The legacy of research on climate sensitivity is thus
remarkably similar to that of percolation flux: decades of research and
increasingly sophisticated science dedicated to better characterizing a
numerical abstraction that does not actually describe observable phenomena,
with little or no change in uncertainty.”
They note that scientific
expertise is often called on to basically predict the future. This is often
done with numerical modeling. Here, they note that weather prediction has
gotten very good due in part to the fact that it is a closed system, but also
due to having many local predictions and model learning with massive amounts of
data. We can predict the weather accurately for about a week in advance.
Predicting climate is another matter and deals far more in the abstract.
“The contrast between weather and climate forecasting
could not be clearer. Weather forecasts are both reliable and useful because
they predict outcomes in relatively closed systems for short periods with
immediate feedback that can be rapidly incorporated to improve future
forecasts, even as users (picnickers, ship captains) have innumerable
opportunities to gain direct experience with the strengths and limits of the
forecasts.”
“Using mathematical models to predict the future global
climate over the course of a century of rapid sociotechnical change is quite
another matter. While the effects of different development pathways on future
atmospheric greenhouse gas concentrations can be modeled using scenarios, there
is no basis beyond conjecture for assigning relative probabilities to these
alternative futures. There are also no mechanisms for improving conjectured
probabilities because the time frames are too long to provide necessary
feedback for learning. What’s being forecast exists only in an artificial
world, constituted by numbers that correspond not to direct observations and
measurements of phenomena in nature, but to an assumption-laden numerical
representation of that artificial world.”
They argue that while
modeling is numerical, it is predictive and has some level of unresolvable
uncertainty. Here, they invoke Alfred North Whitehead’s 1929 “Fallacy of
Misplaced Concreteness,” in which abstractions are taken as concrete facts.
Models used to predict the future are often successful but not always. There
are many examples of models being wrong, especially when the model assumptions
applied were wrong.
Next, they introduce three
interrelated conditions that allow the establishment of causal relationships
that can guide understanding and action.
“First is control: the creation or exploitation of
closed systems, so that important phenomena and variables involved in the
system can be isolated and studied. Second is fast learning: the availability
of tight feedback loops, which allow mistakes to be identified and learning to
occur because causal inferences can be repeatedly tested through observations
and experiments in the controlled or well-specified conditions of a more or
less closed system. Third is clear goals: the shared recognition or stipulation
of sharply defined endpoints toward which scientific progress can be both
defined and assessed, meaning that feedback and learning can occur relative to
progress toward agreed-upon outcomes that confirm the validity of what is being
learned.”
Technology influences the
fulfilment of these three conditions. Technology is what makes science real for
us, they note. It is essentially proof that science works. Unfortunately for
complex social problems, the three conditions are rarely fulfilled. There is
just too much uncertainty. So-called experts who opine about such complex
problems often insist that they can predict the future with modeling. However,
that does not decrease the inherent uncertainty. They label such expertise as “inappropriate
expertise.” It is inappropriate because it assumes they can make the
uncertain certain with their knowledge. This is in contrast to
“expert-practitioners” who can readily show that their knowledge is correct.
They offer an alternative:
“Decision-makers tasked with responding to controversial
problems of risk and society would be better served to pursue solutions through
institutions that can tease out the legitimate conflicts over values and
rationality that are implicated in the problems. They should focus on designing
institutional approaches that make this cognitive pluralism explicit, and they
should support activities to identify political and policy options that have a
chance of attracting a diverse constituency of supporters.”
They give three examples. The
first is an environmental rule in Massachusetts that moved the arguments for
and against the use of toxic substances from one of conflict to one of
collaboration by reframing it as a rule to replace toxics with non-toxics. The
second is the idea of hydrocarbon reserves, which is abstract in the sense that
how many reserves there are depends on factors that often change, such as
technological capabilities and costs to extract. Here, they note that making
the process more pluralistic has led to better and more accurate predictions of
hydrocarbon reserves than the USGS alone predicted. The third example involves
macroeconomic models, which some economists argue have been very wrong over the
years. Discussing the models is only a part of the Fed’s decision-making
process, which involves reaching an overall consensus on interest rate
determinations.
“Truth, it turns out, often comes with big error bars,
and that allows space for managing cognitive pluralism to build institutional
trust.”
What, then, is appropriate
expertise?
“Appropriate expertise emerges from institutions that
ground their legitimacy not on claims of expert privilege and the authority of
an undifferentiated “science,” but on institutional arrangements for managing
the competing values, beliefs, worldviews, and facts arrayed around such
incredibly complex problems as climate change or toxic chemical regulation or
nuclear waste storage. Appropriate expertise is vested and manifested not in
credentialed individuals, but in institutions that earn and maintain the trust
of the polity.”
They also note that it is
easier to trust the concrete expertise of practitioner-experts than to trust
those inappropriate experts who are over-leveraged in modeling.
“People still listen to their dentists and auto
mechanics. But many do not believe the scientists who tell them that nuclear
power is safe, or that vaccines work, or that climate change is real.”
At the end of the essay, they
reiterate the importance of pluralism and consideration of the impacts on and
views of other stakeholders in decision-making and policymaking. Here, they
invoke the importance of democracy in policymaking.
“Successfully navigating the divisive politics that
arise at the intersections of technology, environment, health, and economy
depends not on more and better science, nor louder exhortations to trust
science, nor stronger condemnations of “science denial.” Instead, the focus
must be on the design of institutional arrangements that bring the strengths
and limits of our always uncertain knowledge of the world’s complexities into
better alignment with the cognitive and political pluralism that is the foundation
for democratic governance — and the life’s blood of any democratic society.”
This essay is certainly food for thought on the subject of scientific expertise in the complex modern world.
References:
Policy
Making in the Post-Truth World: On the Limits of Science and the Rise of Inappropriate
Expertise: Steve Rayner and Daniel Sarewitz. The Breakthrough Journal. Winter
Issue 13, 2021. Journal
Winter Issue 13_2021_PRINT_rev4.indd
Prosocial
motives underlie scientific censorship by scientists: A perspective and
research agenda. Cory J. Clark, Lee Jussim, Komi Frey, +35 , and William von
Hippel. PNAS. Vol. 120 | No. 48. November 20, 2023. Prosocial motives
underlie scientific censorship by scientists: A perspective and research agenda
| PNAS
No comments:
Post a Comment