Healthy disagreement
among scientists drives the creation of new knowledge and is a necessary
precursor to consensus upon which technologies, policies, and new
knowledge can be built. Yet, in spite of its prominence
in popular and theoretical models of scientific progress, disagreement
has received little empirical attention, with progress stymied by a lack
of appropriate data and widely-accepted quantitative indicators. In
this talk, we outline progress in overcoming
these challenges, illustrating how increasingly-available full-text
data and new approaches to measuring disagreement are paving the way for
a more comprehensives, empirical, and quantitative understanding of the
salience and features of disagreement in science
at multiple levels of analysis. Using a rigorously-validated cue-word
based approach, instances of disagreement are identified from the
citation sentences of millions of publications, and incorporated into a
singular indicator of disagreement. Using this indicator,
we simultaneously reveal the structure of disagreement between
macro-level fields and the enormous heterogeneity across meso-level
subfields. At the micro-level, we complement these data with published
comments—the most unambiguous instance of criticism in
science—in order to better understand the sociological drivers of
disagreement, including author gender, seniority, prestige, and more.
This project establishes a firm methodological and empirical foundation
for a science of scientific disagreement, which
will prove essential for validating theories of scientific progress,
building tools for scholarly search and discovery, designing
consensus-aware science policy, and for effectively communicating
epistemic uncertainty and consensus to the public.
Dakota Murray is a
Postdoctoral research associate at the Center for Complex Network
Research at Northeastern University, and has demonstrated experience
studying sources of bias in scientific careers, the
mobility of scientists, and disagreement in science. He holds a
doctorate in Informatics from Indiana University Bloomington. His
current research focuses on leveraging full-text data and computational
tools for developing an empirical and quantitative approach
for studying scientific debates and consensus.