Beyond Evidence-Based
Belief Formation:
How Normative Ideas Have
Constrained Conceptual Change Research
Stellan Ohlssona
aUniversity of Illinois at Chicago, Chicago,
United States
Article
received 4 September
2013 / revised 13
December 2013 /
accepted 13 December 2013/ available online 20 December 2013
Abstract
The cognitive sciences,
including
psychology and education, have their roots in antiquity. In
the historically
early disciplines like logic and philosophy, the purpose of
inquiry was
normative. Logic sought to formalize valid inferences, and the
various branches
of philosophy sought to identify true and certain knowledge.
Normative principles
are irrelevant for descriptive, empirical sciences like
psychology. Normative
concepts have nevertheless strongly influenced cognitive
research in general
and conceptual change research in particular. Studies of
conceptual change
often ask why students do not abandon their misconceptions
when presented with
falsifying evidence. But there is little reason to believe
that people evolved
to conform to normative principles of belief management and
conceptual change.
When we put the normative traditions aside, we can consider a
broader range of
hypotheses about conceptual change. As an illustration, the
pragmatist focus on
action and habits is articulated into a psychological theory
that claims that
cognitive utility, not the probability of truth, is the key
variable that
determines belief revision and conceptual change.
Keywords: Belief formation; Belief
revision; Cognitive utility;
Conceptual change; Descriptive vs. normative inquiry; Pragmatism
http://dx.doi.org/10.14786/flr.v1i2.58
Cognitive scientists pride
themselves on
their interdisciplinary approach, drawing upon anthropology,
artificial
intelligence, evolutionary biology, linguistics, logic,
neuroscience,
philosophy, psychology, and yet
But the
cognitive sciences
exhibit principled differences that might get in the way of
interdisciplinary
efforts. Inquiry into cognition was originally rooted in the
desire for human
betterment. The first cognitive disciplines, including logic,
epistemology, and
linguistics, were normative disciplines. Logicians wanted to
systematize valid
inferences, as opposed to whatever
inferences, including fallacious ones, that people make;
philosophers sought to
identify criteria for certain
knowledge, as opposed to describe all knowledge[1];
and early linguists were more concerned with codifying correct grammar than with cataloguing grammatical
errors.
Historically, these and related disciplines mixed normative and
descriptive
elements in a way that is quite foreign to the contemporary
conception of a
natural or social science. In this respect, they resembled
aesthetics, ethics,
and legal scholarship more than biology, chemistry, and physics
as practiced
since the Scientific Revolution (Butterfield, 1957; Osler,
2000).
Because the
normative
disciplines were historically prior, the concepts, practices,
and tools of
normative inquiry became part of the intellectual infrastructure
of the
self-consciously descriptive sciences like neuroscience and
experimental
psychology that became established in the latter half of the 19th
century. Concepts like abstraction,
association, and imagery are obvious examples of such imports. This
intellectual
inheritance helped the new sciences get started, in part by
suggesting
questions and problems (How
and when are
associations formed?). But normative and descriptive
disciplines are
different enough in their goals and methods so that it is
reasonable to ask
whether that inheritance has had a negative impact as well.
In this
article, I argue that
certain normative ideas have led the study of cognition in
general and
cognitive change in particular down an unproductive path. The
types of
cognitive changes I have in mind are those that psychologists
call conceptual change,
belief revision, and theory
change (Carey, 2009; Duit & Treagust, 2003;
Nersessian, 2001; Thagard,
1992; Vosniadou, Baltas & Vamvakoussi, 2007) For purposes of this
article, I use these
terms as near-synonyms. When a collective label is needed, I
call them non-monotonic
change processes (Ohlsson,
2011).
The argument proceeds
through
seven steps. In Section 1, I elaborate on the distinction
between normative and
descriptive inquiry. Section 2 highlights the role of normative
concepts in
what I call the Ideal-Deviation paradigm, a particular style of
research that
is common in cognitive psychology. In Section 3 I show that this
research
paradigm is also present, albeit implicitly, in conceptual
change research.
From a normative perspective, people ought to base their
concepts and beliefs
on evidence and revise them when they are contradicted by new
evidence. The assumption
-- sometimes implicit, sometimes explicit -- that people’s
cognitive systems
are designed to operate in this way has focused researchers’
attention on the
deviations of human behaviour from normatively correct belief
management (using
the latter term as a convenient shorthand for “belief formation
and belief
revision”). But the normative perspective is irrelevant to the
scientific study
of cognitive change; hence, so are the deviations between the
norms and actual
cognitive processing. But if the deviations are irrelevant, so
are our
explanations for them. To go beyond the current state of
conceptual change
research, we need to make explicit the influence of the
normative perspective,
identify the constraints it has imposed on theory development,
and relax those
constraints. When we cultivate a resolutely descriptive stance,
the space of
possible theories of conceptual change expands. As a first step
towards a new
theory, Section 4 argues that the notion that people form
concepts and beliefs
on the basis of evidence might be fundamentally incorrect. The
problem of why
people do not revise their misconceptions when confronted with
contradictory
evidence then dissolves, and other questions move to the
foreground. In Section
5, I outline an approach to conceptual change that is inspired
by the
pragmatist notion that concepts and beliefs are tools for
successful action.
According to this perspective, the key variable that drives
conceptual change
is not the strength of the relevant evidence or the probability
of truth, but
cognitive utility. Section 6 answers two plausible
objections to this view, and Section 7 outlines some of its
implications.
Section 8 recapitulates the argument. Although the concept of
utility is not
itself new, the critique of conceptual change research as mired
in normative
ideas is stated here for the first time, and the conjecture that
utility can
replace probability of truth as the key theoretical variable in
conceptual
change has not been previously proposed.
1.
Normative versus descriptive
inquiry
Descriptive
sciences aim to
provide accurate theories about the way the world is. As I use
it, the word
“descriptive” does not stand in contrast to “theoretical” or
“explanatory” but
encompasses all the empirical and theoretical practices of the
natural and
social sciences as we now conceive them; the term “empirical”
could have been
used instead, but is sometimes understood as standing in
contrast to
“theoretical.” The goal of descriptive science is to provide an
account of
reality that is intersubjectively valid and reflects the world
as it is,
independent of human judgments or wishes. Descriptive sciences
are essentially
concerned with adapting theories and concepts to data. The descriptive
sciences constitute what we
today call “science.”
Normative
disciplines, in
contrast, investigate how things ought to be. They are
essentially concerned
with conformity to standards of goodness. A large portion of
what we have in
our heads consists of more or less explicit normative knowledge.
Aesthetics,
epistemology, ethics, etiquette, law, literary criticism, logic,
rhetoric, and
several other disciplines ask what the appropriate standards
are, or should be,
in some area of human endeavour; how one decides whether some
instance does or
does not conform to the relevant standards; and why particular
instances
conform, or fail to conform. The state of theory in these
disciplines varies
widely, from formalized theories of valid inferences in logic to
the obviously
culture-dependent rules of good manners, and the highly
controversial theories
of literary criticism.
As these
nutshell definitions
are meant to illustrate, descriptive and normative disciplines
are so different
from each other that the distinction seems impossible to
overlook. But the
separation of normative and descriptive inquiry was in fact long
in coming; for
example, psychology was included among “the moral sciences” well
into the 19th
century. The distinction was not fully articulated and accepted
in Western
thought until the 20th century, supported by, among
other
influences, the logical positivists’ emphasis on the distinction
between fact
and value. (For criticisms of the distinction, see, e.g.,
Köhler, 1938/1966,
Putnam, 2002, and others.). It is nevertheless anachronistic to
think of
earlier generations of scholars as having confused
descriptive and normative inquiry. The situation is better
described by saying
that they had not yet distinguished them.
Astronomy
provides an example
of research in an era when the distinction was not yet fully
articulated (Kuhn,
1957; Margolis, 1987, 1993). Some ancient astronomers adopted
the normative
idea that planets ought to move in perfectly circular orbits,
because the
heavenly bodies were perfect beings and perfect beings ought to
move in perfect
orbits and the circle is the most perfect geometric figure.
Astronomers then
spent two millennia explaining the deviations of the observed
planetary orbits
from the normatively specified orbits using the Ptolemaic
construct of
epicycles, instead of exploring other hypotheses about the
geometric shape of
the orbits (Frank, 1952). Research in the natural sciences is no
longer
constrained by normative ideas in this way. Section 2 shows that
psychology, in
contrast, has not yet outgrown its normative inheritance.
2.
The
Ideal-Deviation Paradigm
Normative
principles have
generated a psychological research paradigm that I refer to as
the Ideal-Deviation
Paradigm. Although only
a subset of psychological research conforms to this paradigm,
the paradigm has
had a strong and largely negative impact on research in
cognitive psychology in
general and research on conceptual change in particular. A line
of research
that follows this paradigm proceeds through the following
general steps
(examples to follow):
(a)
Choose a normative theory. How ought the mind carry out
such-and-such a
process, or perform such-and-such a task?
(b)
Construct or identify a situation or task environment in which
that theory
applies, and derive its implications for normatively correct
behaviour.
(c) Recruit
human subjects and
observe their behaviour in the relevant situation.
(d) Describe
the deviations of
the observed behaviour from the normatively correct behaviour.
(e) Hypothesize
an explanation
for the observed deviations.
(f) Test the
explanation in
further empirical experiments.
Readers who are
familiar with
cognitive psychology will have no difficulty in thinking of
instances of the
Ideal-Deviation Paradigm. The prototypical example is research
on logical
inference (Evans, 2007). In this area, researchers originally
used logic as
developed by logicians – primarily the logics of syllogistic and
propositional
inferences – as the relevant normative theory. The reasoning
problems presented
to human subjects include Wason’s famous 4-Card Task (a.k.a. the
Selection
Task; Wason & Johnson-Laird, 1972). Propositional logic
prescribes a
particular pattern of responses to this task, and the deviations
of human
responses from the prescribed pattern have been replicated in
dozens, perhaps
hundreds, of experimental studies. Researchers have proposed and
debated a wide
range of explanations for the observed deviations
(Johnson-Laird, 2006; Klauer,
Stahl, & Erdfelder, 2007).
Research on
decision making is
a second instance of the Ideal-Deviation Paradigm. The
Subjective Expected
Utility (SEU) theory and the mathematics of probability provide
a normative
theory for how to choose among competing options. When people
are confronted
with choices that involve probabilistic outcomes in laboratory
settings, they
deviate from the normatively correct behaviour in a variety of
ways. Errors
like availability and
representativeness are
examples. In the
former, human judgments about the probability of an event (e.g.,
an airline
crash) are influenced by the ease with which the person can
retrieve an example
of such an event from memory. In the latter case, human
judgments are
influenced by the similarity of a sample to the population from
which the
sample was drawn. In this field, too, researchers have proposed,
debated, and
experimentally investigated multiple explanations for these and
other observed
deviations (Kahneman, 2011).
The key point
for present
purposes is that the Ideal-Deviation Paradigm mixes normative
and descriptive
elements in a way that is foreign to the way we now think of
scientific
research. To highlight this point, imagine biochemists in the
1950s deciding
that a particular protein molecule ought
to fold itself into such-and-such a three-dimension structure,
for, say,
aesthetic reasons. Imagine also that they observe that the
actual shape of the
molecule deviates from this normatively specified structure, and
then spend
their time and theoretical energy explaining why the protein
deviates in
such-and-such a way from the supposedly correct structure,
instead of
explaining why it folds together the way it actually does. No
such
investigation could survive peer review for a contemporary
chemistry journal.
In short, given
our current
conception of scientific research, there is no justification for
the normative
element in the Ideal-Deviation Paradigm. A descriptive theory of
how people
think or learn must be based on accounts of the actual processes
occurring in
people’s heads when they draw inferences, make decisions, and
revise their
knowledge, regardless of whether those processes are similar to,
or different
from, normatively correct processes. Comparing empirical
observations to a
normative theory contributes nothing to that enterprise. Section
3 argues that
normative conceptions are nevertheless at the centre of
contemporary research
in conceptual change.
3. Ideal-Deviation
in
conceptual change
The
Ideal-Deviation Paradigm
has strongly impacted psychological research on belief revision,
conceptual
change, theory change, and related processes. The impact is not
immediately
obvious, because the relevant normative theory is less precise
and less
explicit than the normative theories that underpin studies of
logical reasoning
and decision making. The normative theory of belief management
can be summarized
in four principles:
Principle 1:
Grounding. Beliefs and concepts ought to be based on
evidence. In this context, “based on” means “derived from.” The
derivation is
typically understood to be some form of induction across
qualitative
observations and/or aggregation of quantitative data. To adopt a
belief for
which one has no evidence is deplorable, even irresponsible, and
a belief that
is not grounded in evidence is dismissed as a guess, prejudice,
or mere
speculation.
Principle 2:
Graded
conviction. Beliefs ought
to be held with a conviction that is proportional to the
strength of the
relevant evidence. For example, hearsay provides weaker evidence
than direct
observation; a anecdote provides weaker evidence than a study
based on a
representative sample; and a correlational study provides weaker
evidence for a
causal relation than an experimental study. The strength of
one’s convictions
ought to reflect such differences in the nature and extent of
the relevant
evidence. For purposes of quantitative comparisons, the
conviction with which a
belief is held can be conceptualized as an estimate of its
probability of being
true.
Principle 3:
Belief-belief
conflicts. When two
beliefs or informal theories contradict each other, the person
ought to choose
to believe the one that is backed by the stronger evidence. The
theory with the
strongest support ought to have priority in the control of
behaviour, including
both discourse and action. To hold contradictory beliefs (P
& not-P) is to
be inconsistent and hence irrational.
Principle 4:
Belief-evidence conflicts.
When beliefs are contradicted by new evidence, they ought to be
revised so as
to be consistent with both the old and the new evidence. Failure
to do so makes
a person “closed minded”, “irrational”, “rigid minded”, or a
victim of “robust
misconceptions.”
These four
principles are mere
common sense; this is how a rational agent ought to manage his
or her beliefs.
There seems to be little gain in giving such vacuous verities
the status of
principles. But my purpose is to make explicit what is normally
too embedded in
our conceptual infrastructure to be visible.
Elements of the
normative
theory of belief management, masquerading as descriptive
statements, can be
found throughout the cognitive sciences. For example, Allport
(1958/1979)
proposed the contact theory of racial prejudice. The key idea
was that negative
racial stereotypes would be diminished if a person with such a
stereotype were
subjected to frequent contacts with members of the relevant
ethnic group. The
hypothesis was that the contacts would provide evidence against
the negative
stereotypes and pave the way for other, more positive opinions.
In the
philosophy of science, Kuhn (1970) described theory change as a
consequence of
the accumulation of anomalies. In educational psychology, Posner
et al. (1982)
hypothesized that students have to be dissatisfied with their
current beliefs
about scientific phenomena before they are prepared to revise
them, and that
being confronted with evidence to the contrary is the key source
of
dissatisfaction. In developmental psychology, Gopnik and
Meltzoff (1997)
embraced Principle 4, designating belief-evidence conflicts as
the main drivers
of cognitive change: “Theories may turn out to be inconsistent
with the evidence,
and because of this theories change.” (p. 39) Although other
processes are
involved as well, the processing of counterevidence is the most
important:
“Theories change as a result of a number of different
epistemological
processes. One particularly critical factor is the accumulation
of
counterevidence to the theory.”
(p. 39)
Paradoxically,
cognitive
scientists confidently assert these variations of Principle 4,
while they
simultaneously and in parallel assert that people deviate from
Principle 4. In discipline
after discipline, researchers have observed that people do not
always and
necessarily revise their beliefs when confronted with
contradictory evidence.
The predictions of the contact theory of racial prejudice were
not verified and
the theory had to be reformulated (Pettigrew, 1998). Likewise,
Strike and
Posner (1992) found that students retain their misconceptions
even after
instruction that is directly aimed at confronting those
misconceptions with
contradicting evidence. “One of the most important findings of
the
misconception literature…is that misconceptions are highly
resistant to change”
(Strike & Posner, 1992, p. 153). In a review paper, Limón
(2001) wrote that
“…the most outstanding result of the studies using the cognitive
conflict strategy
is the lack of efficacy for students to achieve a strong
restructuring and,
consequently, a deep understanding of the new information.” (p.
364) Indeed,
the deviation of student behaviour from Principle 4 is the very
phenomenon that
created conceptual change as a distinct field of research, at
least within
educational research.
Consistent with
the
Ideal-Deviation Paradigm, researchers have responded to the
finding that people
do not (necessarily) adapt their beliefs to contradictory
evidence by proposing
various explanations for this deviation. For example, Rokeach
(1960, 1970)
proposed that belief systems have a hierarchical structure, and
that change
becomes more and more difficult as one moves from the periphery
to the centre.
As a result, most changes are peripheral and central principles
are hardly ever
affected by evidence. Political and religious principles are
cases in point.
The philosopher Imre Lakatos has proposed a similar theory to
explain theory
change in science (Lakatos, 1980). Festinger (1957/1962)
launched a
long-lasting line of research in social psychology that centred
on a set of
mechanisms for reducing what he called cognitive dissonance.
Cognitive
mechanisms for dissonance reduction process contradictory
evidence without any
fundamental revision of the relevant beliefs.
More recently,
cognitive
psychologists have added yet other explanations. The category
shift theory of
Chi (2005, 2008) and co-workers explains the robustness of
misconceptions as a
consequence of the inheritance of characteristics from the
(frequently
inappropriate) ontological category to which a phenomenon has
been assimilated.
Vosniadou and Brewer (1992) and Vosniadou and Skopeliti (2013)
explain the
deviations as a consequence of the synthesis of prior (and
frequently
inaccurate) mental models into more comprehensive (but sometimes
equally
inaccurate) mental models. Sinatra and co-workers have added
motivational and
emotional variables as additional sources of explanation
(Broughton, Sinatra,
& Nussbaum, 2013; Sinatra & Pintrich, 2003). Yet other
perspectives on
conceptual change have been proposed (see, e.g., Rakison &
Poulin-Dubois,
2001; Shipstone, 1984). Ohlsson, (2011, Chap. 9) provides a more
extensive
comparative analysis of these and related types of explanations.
In short,
although the
normative theory of belief formation is less explicit than the
normative
theories that underpin studies of logical reasoning and decision
making,
research on conceptual change follows closely the
Ideal-Deviation paradigm. The
basic structure of conceptual change research is that (a)
students ought to
revise their misconceptions when confronted with contradictory
evidence, (b)
the empirical evidence indicate that they do not in fact do so,
and therefore
(c) we need to explain why they do not do so. But this research
enterprise is
only meaningful if one accepts Principles 1 - 4 as relevant for
the study of
conceptual change. Section 4 prepares for a new approach to
conceptual change
by arguing that people do not base their concepts and beliefs on
evidence. If
so, Principles 1-4 are irrelevant for understanding conceptual
change.
4.
The
irrelevance of evidence
At first
glance, the normative
theory of belief management seems highly relevant for
understanding human
cognition. If people do not adapt their beliefs to reality, how
do they get
through their day? Surely the deviations from rational belief
management
uncovered in various areas of cognitive research are relatively
minor slips of
a fundamentally rational cognitive system for building and
maintaining a
veridical belief base? Such slips might be due, for example, to
cognitive
capacity limitations or emotional biases.
This view is
plausible but
difficult to evaluate. We know very little about how people form
and revise
beliefs in natural settings, because there are few relevant
empirical studies.
What follows are some informal observations and examples. In
conjunction, they
suggest a radical conclusion: The principle that people base
their beliefs on
evidence might be fundamentally incorrect rather than an
optimistic
idealization or a partial truth.
An adult person
has a large
belief base in memory, at least if the term “belief” is applied
broadly enough
to include not only the deep principles that tend to be the
object of analysis,
but also local, concrete facts. For example, I have multiple
beliefs about the
public transportation system in the city where I live: that
there are buses and
subway trains; that there are multiple subway lines; where they
go; how long a
trip is likely to take; how much it costs; the location of
stations; and so on.
This small domain of experience is likely to encompass several
hundreds,
perhaps even thousands, of beliefs, most of which are likely to
be accurate.
The view that a belief ought to be derived from, or based on,
observational
evidence works well with respect to such concrete, particular
matters. For
example, the belief that there is a subway station at the corner
of X and Y
streets might very well be acquired by no more complicated a
process than
walking down X street and encountering that very station at the
crossing with Y
street. Such routine
belief formation
events can plausibly be attributed to direct observation and in
that sense
conforms to Principles 1 - 4.
The direct
observation account
of belief formation quickly runs into difficulties when the
belief is general.
For example, most adults have a variety of beliefs about
economical, political,
and social affairs. Informal observations indicate that a
significant
proportion of such beliefs are not based on any evidence
whatsoever. Will
austerity economics stimulate the economy or depress the markets
by robbing
consumers of their ability to consume? Quite a few adults are
prepared to offer
a point of view about this issue, and, just as obviously, very
few of them have
access to relevant quantitative data or other observational
evidence.
This is not an
isolated
instance. Consider the range of controversial socio-political
and economic
issues in the public discourse: gun control, surveillance by
intelligence
organizations, same-sex marriage, drone strikes on foreign soil,
the benefits
of universal health care – a large proportion of adults have
beliefs regarding
many of such issues, but almost none of those beliefs are based
on evidence.
That is, a person who holds a belief on such an issue did not,
as a rule,
induce it from multiple historical examples or derive it from
statistical data
or other types of observational evidence. Most people cannot
give any coherent
or detailed account of why, how, or even when they adopted any
particular
belief. If people operated with Principle 1, they ought to
answer almost every
question about socio-political and economic issues by saying, “I
don’t know; I
don’t have an opinion on that; I don’t have enough information.”
If general
beliefs are not
formed by induction from observations, how, by what processes
are they formed
instead? Informal reflection on everyday life suggests that we
form general
beliefs by accepting what someone else tells us, either in
face-to-face
conversation or via media. The notion of evidence does not enter
into this
belief formation process in any prominent way, because we do not
normally and
as a rule question or doubt what we are being told. It is enough
to hear
someone say it for us to encode it as veridical. Gabbay and Woods
(2001) call this the Ad
Ignorantiam Rule: “Human agents tend
to accept without challenge the utterances and arguments of
others except where
they know or think they know or suspect that something is
amiss.” (p. 150) The
reason for this rule is probably that we tend to communicate
with people we
trust, and access sources that we have already judged as
reliable. But this
does not support the normative principle, because it is not
obvious that we
base our judgments about the trustworthiness of a source on
anything that would
qualify as evidence.
Another
hypothesis is that
belief formation is internal to the cognitive system. Many of
our beliefs
appear to arrive in the belief base as consequences of already
adopted beliefs.
For example, I believe that public education is an essential
social
institution. I also believe that nations that invest in
education will fare
better than those that do not. It would be an exaggeration to
say that I have
evidence for the second belief. After all, what counts as
evidence as to what
will happen in the future? It seems more accurate to say that I
have adopted
the second belief because it follows from the first. If public
education is
essential, nations underfund it at their peril. Intra-mental
derivations of
this sort can hardly be characterized as evidence-based. First,
the question of
evidence is merely pushed one step backwards, because the
derived belief cannot
be said to be evidence-based unless the beliefs it is derived
from are themselves
evidence-based. Second, the internal derivations are influenced
by factors that
are themselves unrelated to truth, such as a desire for
consistency,
instrumental gain, and various types of biases.
Consider next
Principle 2,
that beliefs ought to be held with graded convictions that
reflect the strength
of the evidence. Are people sensitive to the relative strength
of the evidence?
That is, do people in general and as a rule hold their beliefs
more strongly
when they are supported by more evidence and less strongly when
the support is
weaker? A thorough answer to this question would require
extensive data
collection and some way of measuring the strength of the
relevant evidence.
However, it is noteworthy that the beliefs that people hold with
the greatest
conviction tend to be their religious beliefs, and church
leaders and followers
alike insist that religious beliefs are, and should be, based on
faith, not
evidence. The fact that faith-based beliefs are held more
strongly than other
classes of beliefs is inconsistent with the idea that our brains
are
programmed, in some deep and fundamental way, to base beliefs on
evidence.
Other examples
of conviction
levels that do not seem to reflect the strength of the available
evidence
include those that pertain to beliefs regarding climate change
and the value of
vaccinations. At one time, it was rational to be sceptical
regarding the
reality of climate change; now, the evidence is overwhelming
(Oreskes, 2004).
Nevertheless, some people continue to believe that the climate
is not changing.
The controversy over vaccinations exhibits a similar pattern.
Although caution
was once rational, there are now multiple, large-scale studies
that show
conclusively that there is nothing wrong with the common
vaccines that are
given to children, or with the way they are administered. People
do not get
sick from vaccines; they get sick from germs. However,
anti-vaccine activists
continue to claim that vaccines are harmful and they have many
followers
(Offit, 2011).
Finally,
consider Principle 3,
namely that theory-theory conflicts are to be resolved with
reference to the
relative strengths of the evidence for the competing theories.
Do people
consistently side with the view that has the strongest evidence?
Consider the
issue whether human beings are fundamentally evil and require
discipline in
order to behave themselves, or fundamentally good, so all they
need is an
opportunity to blossom in a natural way. Every news story about
yet another
serial killer is evidence for the former view; every
heartwarming news story
about someone who goes out of their way to make a difference for
people around
them is evidence for the latter view. Every war produces novel
atrocities, but
every natural catastrophe – forest fire, hurricane, tsunami –
generates a fresh
batch of stories about individual heroism and self-sacrifice.
Anybody who
attends to the news has as much evidence for one view as for the
other. Given
that there is much evidence for either view, those of us who
hold a strong
opinion on the issue of human nature must have resolved the
conflict between
these two theories at least partially on the basis of something
other than the
evidence.
To summarize,
informal
observations suggest that people do not, in general, induce
their beliefs from
observational evidence. Although we often base concrete beliefs
about
particular objects and events on direct observation, we appear
to form general
beliefs through ubiquitous encoding of communications by trusted
sources and by
deriving them from other, already adopted beliefs. These
hypothetical but
plausible belief formation processes are not inductive in
nature, and the
contribution of what we normally call evidence to each is weak. In addition, people
show few signs of holding
their beliefs with a conviction that is proportional to the
strength of the
supporting evidence, resolve conflicts among competing beliefs
by comparing the
relative strength of the supporting evidence, or to revise their
beliefs when
they encounter contradictory evidence. These observations
suggest that the
normative theory of belief management, taken as a descriptive
theory, is
fundamentally wrong rather than merely an optimistic
idealization.
But if so, why
do the observed
deviations of human behaviour from Principles 1-4 deserve our
attention? The
consequence of abandoning Principles 1-4 is that conceptual
change researchers
no longer need to explain why misconceptions are robust in the
face of
contradictory evidence. If there is no reason to expect students
to revise
their beliefs when confronted with new evidence, then the
absence of such
revisions is not puzzling. Explanations for why misconceptions
are robust
become obsolete, not in the sense of being falsified, but in the
sense of being
answers to a question we do not need to ask. The problem of why
people do not
revise their beliefs is not so much solved as dissolved.
However, the task of
formulating a scientific theory of belief formation and belief
revision that
can support effective pedagogical practices remains. In Section
5, I propose
that some of the ideas put forward by the American pragmatist
philosophers can
serve as a starting point for a new approach to conceptual
change.
5.
A
Pragmatist Approach
At the end of
the 19th
century and the beginning of the 20th, American
scholars, lead by
William James, Charles Sanders Peirce, and John Dewey, tried to
reformulate the
classical philosophical problems about knowledge, meaning, and
truth in terms
of action instead of observation. They claimed that the meaning
of a concept or
belief resides in the set of actions or “habits” to which it
gives rise. “The
essence of belief is the establishment
of a habit, and different beliefs are distinguished by the
different modes of
action to which they give rise.” (Peirce, 1878, p. 129-130)
The
truth of a belief is tied to the outcomes
of executing those habits. They stopped short of claiming that,
“what works is
what is true”, but some of their contemporaries stated their
ideas even more
boldly than they did themselves (Schiller, 1905).
Pragmatism did
not flourish as
a philosophical, i.e., normative, theory. Its impact faded after
the demise of
its most charismatic leaders. Although it is once again
receiving serious
attention from philosophers (Stich, 1983), my purpose is not to
revive
philosophical pragmatism. Instead, I intend to mine this strand
of thought for
an approach to cognitive change that does not begin with the
assumption that
people decide what to believe by estimating the probability of
truth. The
question is what they estimate instead.
5.1 Cognitive
Utility As the Basis for Cognition
The pragmatist
emphasis on
action fits well with psychological theories of cognition. There
is broad
consensus on certain general features of what cognitive
psychologists have come
to call the cognitive
architecture,
i.e., the information processing machinery that underpins the
higher cognitive
processes (Polk & Seifert, 2002). At the centre of the
cognitive
architecture there is a limited-capacity working memory,
connected to separate
long-term memory stores for declarative and practical (skill)
knowledge. The
working memory receives input from sensory systems, and holds
information that
is being processed in reasoning and decision making. The purpose
of the
cognitive system is to generate behaviour that satisfies the
person’s current
goal. In the process, the system makes endless, lightening quick
choices: Which
goal to pursue next (planning); which part of the environment to
attend to next
(attention allocation); which interpretation of perceptual input
to prefer
(perception); which memory structure to activate next
(retrieval); which
inference to carry out next (reasoning); and which change, if
any, to make in
the system’s knowledge base at any given time (learning).
The pragmatist
stance invites
the hypothesis that the variable that guides the never-ending
choices is the cognitive
utility of the relevant
knowledge structures. To articulate this idea, imagine that each
knowledge
structure (concept or belief) in memory is associated with a
numerical value
that measures its past usefulness. When there is a choice to be
made among
knowledge structures, the one with the higher utility is
preferred and gets to
control discourse and action.
If a
knowledge structure is instrumental in generating a particular
action, and if
that action is successful, then the utility of that knowledge
structure is
adjusted upwards; if the action is unsuccessful, it is adjusted
downwards. Each
application of a knowledge structure is an opportunity for that
structure to
accrue utility (or to loose some of it, in the case of
unsuccessful action).
Over time, the value of the cognitive utility associated with a
knowledge structure
will become stabilized at some asymptotic value that estimates
its usefulness
in general. The distribution of utility values over the belief
base represents
the person’s experience of the world, as filtered through action
rather than
perception.
The cognitive
utility
hypothesis is not novel. A construct of this sort has been
incorporated into
the ACT-R model of the cognitive architecture proposed by John
R. Anderson and
co-workers (Anderson, 2007). In ACT-R, cognitive skills are
encoded in sets of goal-situation-action
rules (skill elements) that specify an action to be considered
when certain
conditions are satisfied by the current situation. In each cycle
of operation,
the architecture retrieves all the rules that have their
conditions satisfied. It
then selects one of those rules to be executed; that is, its
action is taken.
The action usually changes the current situation, and the cycle starts over
with a renewed
evaluation of which rules have their conditions satisfied in the
changed
situation.
In ACT-R, the utility u of
rule i determines
its probability of being selected for execution. In simplified
form, that
probability is given by
Eq. (1) Prob(i) =
u(i) / Σ u(1, 2,…i,…j),
where Σ u(1,2,…i,…j) is the sum of the
utilities of the
rules for which the conditions are satisfied by the current
situation. The
probability that a particular rule i will be selected is thus
proportional to
how much of the total utility represented by all the currently
satisfied rules
it accounts for. The probability of being chosen for execution
is thus a
dynamic quantity that depends on context and that changes from
moment to moment
as cognitive processing unfolds.
If rule i is
selected and
executed on operational cycle n, its
utility is adjusted upwards or downwards, depending on the
outcome. The
adjustment is given by the equation
Eq. (2) u(i,
n) = u(i, n-1) + α[R(i, n)
– u(i,
n-1)],
in which i is the relevant
rule, n is the
operational cycle, and R is the
reward or feedback from the environment about the success of the
executed
action (reinforcement in the behaviourist sense). The magnitude
R(i, n) – u(i,
n-1)
is the reward the rule
realized in cycle n,
R(i, n), over and above the utility it already
possessed in the previous
cycle of operation, u(i, n-1). The rate
parameter α
controls the proportion of that reward increment that is to be
added to the
current utility of the rule, u(i, n-1),
to compute the
utility of the rule in the following cycle, u(i,
n). The
reader is recommended to consult the original source for further
technical
details (Anderson, 2007, pp. 159-164).
In the ACT-R
theory, utility
values are associated with skill elements (rules), and there is
a separate
system of theoretical quantities that pertain to the learning
and application
of declarative knowledge elements. To make the utility construct
relevant for
belief formation we have to hypothesize that utility values are
associated with
declarative knowledge structures (beliefs, concepts, informal
theories) instead
of (or in addition to) skill elements. Furthermore, the relation
between
cognitive utility and belief (subjective truth) has to be
specified. One
possible hypothesis is that there is a threshold such that, when
the utility of
a particular belief rises above that threshold, the person feels
that the
belief is true. A cognitive system that operates in this way
would be
significantly different from ACT-R and other cognitive systems
described in the
cognitive literate (Polk & Seifert, 2002).
A key question
is what degree
of utility new information will be assigned when it is first
encoded into
memory. At the outset, the new knowledge structure has no track
record of
supporting successful action, so one might decide that its
initial utility is
zero. This causes a paradox: If it is zero, it will always have
lower utility
than any competitor with even a modest track record, so it will
never be
activated or chosen, and therefore never have an opportunity to
accrue utility.
To an outside observer, it will appear as if the learner did not
encode the new
information, because his or discourse and action continue to be
guided by other
knowledge structures.
There are
multiple solutions
to this theoretical problem. In Anderson’s ACT-R theory, the
initial value is
indeed set to zero, but a knowledge structure (rule) can be
created multiple
times, and each time the utility value is increased. Other
solutions are
possible. The initial value can be hypothesized to be random, or
equal to the
mean of the utility values of all knowledge structures in
memory. There might
be situations in which competing older knowledge structures do
not apply, but
the new one does, and those situations afford the newer
knowledge with
opportunities to accrue utility.
Many scenarios
that seem like
straightforward instances of truth-based processing are equally
or better
understood in terms of utility. For example, suppose that my
eyes itch. I might
have dry eyes, or I might suffer from an allergy attack. I
decide to take an
antihistamine pill. The itch disappears. In a logic-inspired
analysis, the
belief that I am suffering from an allergy outbreak is a
hypothesis the truth
of which is unknown. The connection between the belief that I
have an allergy
and the prediction that the itch will disappear is a
step-by-step chain of inferences.
The disappearance of the itch is an observation that verifies
the hypothesis,
and my estimate of the probability that I have an allergy
increases as
specified by, for example, Bayesian principles.
This account
has weaknesses.
One weakness is that I am not aware of any lengthy reasoning
process to arrive
at a testable prediction. The process that connects the belief
“I have an
allergy attack” with the fact that my itch stopped is a process
of problem
solving and planning (what
should I do
about my itchy eyes?), not a process of propositional
inference. Another
weakness is that the envisioned process is an instance of a
logical fallacy: If P,
then Q in conjunction with Q does not imply P. This is Popper’s classical critique of
verificationism. But if
the truth-based account has a logical fallacy at its core, how
can people
function? The utility-based account avoids this problem by
postulating a direct
link between the action outcome and the relevant belief: The
action of taking
the antihistamine worked, so my disposition to act on the
allergy belief in the
future is increased.
As the example
illustrates,
the difference between an account in terms of evidence,
inference, and truth,
on the one hand, and an account in terms of utility and action,
on the other,
can be subtle. How does that difference affect how we view
conceptual change?
The pragmatist stance focuses attention on action, the output
side of the
cognitive system, instead of perception, the input side. What
the learner does
matters more than what he or she hears or sees. Passive
reception of
information will not in and of itself have any cognitive
consequences. Unless
the learner retrieves a knowledge structure and uses it to
decide what to do
next, that knowledge structure cannot accrue utility and hence
might remain
dormant, even though the new information has been encoded
accurately.
In the
pragmatist perspective,
new information does not replace the old. In a logic-based
theory, two
different beliefs can be mutually incompatible, which implies
that a person
cannot embrace both. The Earth is either round or flat; it is
impossible to
believe both assertions at once. However, the fact that
knowledge structure i
has utility u(i) is not incompatible with the fact that
knowledge structure
j has utility u(j). The belief that the Earth is flat
might be useful for
mapmaking purposes, while the belief that the Earth is round
might be more
useful for the purpose of circumnavigation. Many tasks in real
life admit of
multiple solutions, varying with respect to goal satisfaction,
efficiency, and
range of applicability. Evaluating beliefs with respect to their
cognitive
utility is thus very different from evaluating them with respect
to their
truth.
Falsification
by contradictory
evidence is, in principle, a one-shot affair. A single
application of Modus
Tollens is logically sufficient to bring down a belief and even
an entire
theory. But utility-based belief revision is necessarily a
gradual matter. Once
the utility rises to the point where a new knowledge structure
is chosen to be
the basis for action on at least some occasions, belief change
is contingent on
the outcomes of the resulting actions. The utility of structure
i might be
steadily raising with each application, while the utility of
some competing
knowledge structure j is gradually dropping. Eventually, the
utility of the
newer knowledge structure will surpass that of the older,
competing structures,
and rise above the threshold of belief. If changes in utility
values are
incremental, then this process is necessarily gradual.
The most
radical difference
between a truth-based and a utility-based account of cognition
pertains to the
trigger of conceptual change. In the truth-based account, it is
the failure of
the older knowledge that drives belief revision. Change happens
because already
acquired concepts and beliefs have been found to be false,
triggering
dissatisfaction and a search for more veridical concepts and
beliefs to replace
them. If there is no failure, there is no push for change. In
the utility-based
account, on the other hand, new information need not wait for
falsification or
dissatisfaction with prior beliefs. It is the success of the
newer concepts and
beliefs that drives the change. No dissatisfaction with the old
belief is
required, only a recognition that the newer belief is an even
more useful basis
for action. Change is driven by success, not failure (Ohlsson,
2009, 2011).
However, before the utility-based perspective can be adopted,
some plausible
objections must be dealt with; this is the task of Section 6.
6.
Two
objections
The purpose of
this section is
to address two objections that must have occurred to the reader.
The first is
that evolution through natural selection ought to have pushed
human cognition
in the direction of the normative theory of belief management
(Principles 1-4),
and the second is that the behaviour of scientists appears to
conform to the
normative theory.
6.1 Natural
Selection for Truthfulness?
One might argue
that the shift
from estimates of the probability of truth to estimates of
cognitive utility is
unimportant. After all, how can a belief be useful unless it is,
in fact, true?
If only true beliefs are useful, then the selective pressures
that drove the
evolution of human cognition must have pushed the belief
management processes
in the learner’s head to conform at least approximately to
Principles 1- 4. How
could our hunter-gatherer ancestors have survived unless their
beliefs
corresponded to reality? The instrumental value of veridicality
in the struggle
for survival implies that the human cognitive architecture is
designed to
derive beliefs from evidence.
But natural
selection cannot
have operated directly on the truthfulness of beliefs. The
probability of
surviving long enough to mate and to raise the resulting
offspring to
reproductive age is a function of how the individual behaves,
not on how he or
she thinks. What mattered during human evolution cannot have
been the truth of
beliefs per se, but the effectiveness of human behaviour.
Consistent selection
in the direction of effective action would create a
utility-based rather than
truth-based system.
The distinction
between truth
and utility would be of minor importance, if the two were
perfectly correlated.
However, false beliefs can lead to successful action. For
example, it does not
matter what belief one has about the causes of severe weather,
as long as that
belief implies that when storm clouds gather, it is time to seek
shelter. The belief
that lightening is a sign of the anger of the gods and the
belief that it is an
electrical discharge are equally good reasons to get out of the
way. The belief
that a certain medical condition is caused by an evil spirit and
that the
spirit can be exercised by ingesting a certain herb can be as
successful as an
account of the disease in terms of bacteria, white blood cells,
etc., if the
relevant herb contains traces of, for example, an antibiotic
substance.
An even
stronger example is
provided by the 14th century physicist Buridan’s
impetus theory of
mechanical motion (Claggett, 1959; Robin & Ohlsson, 1989). A
central
principle in this theory says that to keep an object in motion
requires the
continuous application of force, the opposite of the principle
of inertia that
is at the centre of Newtonian mechanics. However, the impetus
principle holds
on the surface of the Earth due to the universal presence of
friction. If the
goal is to keep an object moving, or to make it move further or
faster, the
impetus concept is as useful a guide to action as the theory
that physicists
teach (apply more force).
In short, truth
and utility
are only partially correlated, and evolution has no way of
selecting for the
truth of beliefs directly, but only for the success of an
individual’s struggle
for survival. Evolutionary considerations thus support rather
than contradict
the hypothesis that utility is the key variable in belief
formation.
6.2 The
Behaviour of Scientists
The reader
would be excused
for thinking that the present author is engaged in a
self-defeating enterprise:
To use evidence and arguments to make the reader believe that
people do not use
evidence and arguments when deciding what to believe. This
article is itself an
attempt to base belief in this matter on evidence. More
generally, scientists
do base their theories on evidence and scientists are people, so
it seems
unreasonable to claim that this is not a common cognitive
capability.
Gopnik and
Meltzoff (1997) has
emphasized this connection between the procedures of scientific
knowledge
creation and individual belief formation: “The central idea of
[our] theory is
that the processes of cognitive development in children are
similar to, indeed
perhaps even identical with [sic], the processes of cognitive
development in
scientists.” (p. 3) Indeed, they have stated their
hypothesis quite
clearly: “…the most central parts of the scientific enterprise,
the basic
apparatus of explanation, prediction, causal attribution, theory
formation and
testing, and so forth, is not a relatively late cultural
invention but is
instead a basic part of our evolutionary endowment.” (pp. 20-21) The consequence is
that cognition can be
explained with normatively correct processes such as Bayesian
inference (Gopnik
et al., 2004).
The
utility-based view
explored in this article does not deny that people can acquire
the higher-order
cognitive skills needed to engage in the methods and procedures
of science. It
does claim that those methods and procedures are acquired.
Scientists are
professional theorizers; they engage in belief formation (a.k.a.
hypothesis
testing) deliberately and on purpose, with a high degree of
awareness. To be
able to do this, they undergo a multi-year training process
called graduate
school. They are supported by a wide variety of tools such as
special-purpose
statistical software that embody the principles of the normative
view of belief
management. Furthermore, scientific research takes place within
a social
context, the scientific discipline, that enforces adherence to
the normative
theory. For example, a scientist who revises his or her theory
to improve its
fit to empirical data (Principle 4) is more admired than someone
who continues
to advocate a favourite theory in the face of counterevidence.
The behaviour of
scientists shows that people can acquire the high-level skills
needed to
function at least approximately as prescribed by the normative
theory. But this
does not imply that the basic processes of the cognitive
architecture conform
to the normative theory.
To cast the
procedures of
science as a description of conceptual change in the individual
is to confuse
two levels of description: the level of the basic processes of
the cognitive
architecture (the “basic part of our evolutionary endowment”),
on the one hand,
and the level of acquired higher-order strategies and skills, on
the other. The
arguments put forth in this paper concern the basic processes. I
know of no
reason to believe that “the most central parts of the scientific
enterprise,
the basic apparatus of explanation, prediction, causal
attribution, theory
formation and testing” is part of our “evolutionary endowment.”
The late
arrival of science in human history, its invention by one
culture at one time,
and the extensive training individuals need to conduct
scientific research make
it highly implausible that anything like the “basic apparatus”
of science is
among our “evolutionary endowment.” Instead, the cognitive
apparatus of science
is precisely “a relatively late cultural invention.”
The relation
between the basic
processes of cognitive change and the procedures of science is
the opposite of
the one claimed by Gopnik and Meltzoff
(1997). Rather than scientific practices explaining how
cognitive change
happens in children and lay adults, the relationship should be
construed the
other way around: A theory of the basic cognitive processes
should explain how
it is possible to acquire the higher-order strategies for belief
management
that approximate the normative theory in Principles 1-4. The
utility-based
perspective have other implications as well, three of which are
discussed in Section
7.
7.
Implications
If we adopt the
utility-based
perspective, what follows? From the point of view of basic
research on
conceptual change, it implies a re-evaluation of existing
theoretical
constructs, methodologies, and applications. Traditionally,
research on
conceptual change and belief formation has been
perception-centric: The focus
has been on what the learner sees and hears, and how he or she
processes the
perceived information. The utility-based perspective, in
contrast, implies a
need to focus on what the learner does, when and where he or she
succeeds or
fails, and on what information is activated, retrieved, and used
to guide
action. A learning trajectory is primarily to be defined in
terms of tasks
undertaken, and only secondarily in terms of information
encountered. As a side
effect of such a re-focusing, the traditional concepts, tools,
and puzzles
regarding truth inherited from philosophy and logic will become
comparatively
less important.
The
perception-centric bias of
cognitive research in general and cognitive studies for
education in particular
is driven, in part, by the practicalities of psychological
experimentation. The
experimenter controls the subject’s task environment, so he or
she can create
complex but well-specified conditions and contrasting situations
by varying the
stimulus. Such variations can easily be described in research
reports. The
subjects’ behaviours, on the other hand, are only easy to report
and interpret
in an intersubjectively valid way if they consist of simple,
easy-to-record
events, like pushing a button or placing a mark on a rating
scale. The
pragmatist perspective implies that this style of empirical
inquiry runs the
risk of eliminating from the researchers’ consideration the
central subject
matter of cognition, namely complex, temporally extended,
hierarchically
structured, and dynamically coordinated sequences of actions in
the service of
human goals and objectives. The pragmatist perspective implies a
need for a
period of methodological innovation in which researchers develop
new techniques
to record and interpret complex behaviours.
From the point
of view of
instructional application, the utility-based account poses
multiple challenges:
How to stimulate students to encode knowledge that they have no
reason to
believe, and that is only tangentially relevant for their own
action? How to
design situations in which the new knowledge presented in the
course of
instruction, but not their prior knowledge, applies, so that the
new knowledge
can accrue utility? How to provide learners with multiple
opportunities to
apply new knowledge without resorting to mind numbing drill and
practice? These
questions are quite different from the questions of why
misconceptions are
robust, what evidence will convince a student that his or her
misconceptions
are in fact inaccurate, or how to train students to pay
attention to evidence,
so pursuing them will likely lead educational researchers in
novel directions.
8.
Conclusion
Throughout the
history of
science, interdisciplinary work has often been innovative and
path breaking. At
the beginning of conceptual change research, there was every
reason to believe
that drawing upon a variety of disciplines was a productive way
to proceed.
However, researchers (including the present author) overlooked
the distinction
between normative and descriptive disciplines, fell into the
Ideal-Deviation
paradigm, and spent their theoretical energies explaining the
main observed
deviations from the normative theory: that students do not
revise their prior
conceptions when confronted with counterevidence. But the
normative idea that
people ought to base their beliefs on evidence is irrelevant for
the empirical
study of cognition. There is little or no evidence that people
base any of
their beliefs on evidence, and considerable evidence that they
do not. If they
do not, then it is no surprise that science courses fail to
impact students’
beliefs about scientific phenomena, and efforts to explain this
supposed
phenomenon are unnecessary. To make progress in understanding
conceptual
change, researchers need to adopt a resolutely naturalistic
approach that makes
no normatively inspired assumptions about belief formation and
belief revision.
The pragmatist view that cognition evolved to support successful
action and
that beliefs are evaluated on the basis of their cognitive
utility instead of
their probability of being true is an alternative starting point
for conceptual
change research. The utility-based perspective implies that
action is necessary
for conceptual change, that old and new beliefs are not mutually
incompatible,
that conceptual change is necessarily gradual, and that change
is not driven by
the failures of misconceptions but by the successes of better
ideas. A research
program to articulate this perspective would replace the
traditional
perception-centric bias of psychological research with an
action-centric
approach that forefronts the cognitive consequences of complex
actions.
References
Allport, G. W.
(1958/1979). The
nature of prejudice (2nd ed.).
Reading, MA: Addison-Wesley.
Anderson, J.
R. (2007). How
can the human mind occur in the physical universe (pp.
159-165)? Oxford UP.
Broughton, S.
H., Sinatra, G.
M., & Nussbaum, E. M. (2013). “Pluto has been a planet my
whole life!”
Emotions, attitudes, and conceptual change in elementary
students’ learning
about Pluto’s reclassification. Research
in Science Education, 43, 529-550.
Butterfield,
H. (1957). The origins of
modern science 1300-1800
(revised ed.). Indianapolis, IN: Hackett.
Carey, S.
(2009). The origin of
concepts. New York: Oxford
University Press.
Chi, M. T. H.
(2005).
Commonsense conceptions of emergent processes: Why some
misconceptions are
robust. The Journal of
the Learning
Sciences, 14,
161-199.
Chi, M.T.H.
(2008). Three
types of conceptual change: Belief revision, mental model
transformation, and
categorical shift. In S.Vosniadou (Ed.), Handbook
of research on conceptual change (pp. 61-82). Hillsdale, NJ:
Erlbaum.
Claggett, M.
(1959). The science of
mechanics in the middle ages.
Madison, Wisconsin: University of Wisconsin Press.
Duit, R.,
& Treagust, D.
F. (2003). Conceptual change: A powerful framework for improving
science
teaching and learning. International
Journal
of Science Education, 25(6),
671-688.
Evans, J. St.
B. T. (2007). Hypothetical
thinking: Dual processes in reasoning and judgment. New
York: Psychology
Press.
Festinger, L.
(1957/1962). A
theory of cognitive dissonance. Stanford, CA: Stanford
University Press.
Frank, P.
(1952). The origin
of the separation between science and philosophy. Proceedings of the American Academy of Arts and
Sciences, 80(2),
115-139.
Gabbay, D.,
& Woods, J.
(2001). The new logic. Logic Journal of the Interest Group
in Pure and
Applied Logics, vol. 9, pp. 141-174.
Gopnik, A.,
& Meltzoff,
A. N. (1997). Words, thoughts, and theories. Cambridge
MA: MIT Press.
Gopnik, A.,
Glymour, C.,
Sobel, D. M., Schulz, L. E., & Kushnir, T. (2004). A theory
of causal
learning in children: Causal maps and Bayes nets. Psychological Review, 111,
3-32.
Johnson-Laird,
P. N. (2006). How
we reason. New York: Oxford University Press.
Kahneman, D.
(2011). Thinking, fast
and slow. New York:
Farrar, Straus, and Giroux.
Klauer, K.
C., Stahl, C., & Erdfelder, E. (2007). The abstract selection
task: New data and an almost
comprehensive model. Journal
of
Experimental Psychology: Learning, Memory, and Cognition,
33, 680-703.
Kuhn, T. S.
(1957). The
Copernican revolution: Planetary astronomy in the development
of Western
thought. New York: Random House.
Kuhn, T. S.
(1970). The
structure of scientific revolutions (2nd ed.).
Chicago, IL:
University of Chicago Press.
Köhler, W.
(1938/1966). The place of
value in a world of facts.
New York: Liveright.
Lakatos, I.
(1980). Philosophical
papers (vol. 1): The methodology of scientific research
programmes).
Cambridge, UK: Cambridge University Press.
Limón, M.
(2001). On the
cognitive conflict as an instructional strategy for conceptual
change: a
critical appraisal. Learning and Instruction, 11,
357-380.
Margolis, H.
(1987). Patterns,
thinking, and cognition: A theory of judgment. Chicago,
IL: University of
Chicago Press.
Margolis, H.
(1993). Paradigms
and barriers: How habits of mind govern scientific beliefs.
Chicago, IL:
University of Chicago Press.
Nersessian, N.
J. (2008). Creating
scientific concepts. Cambridge, MA: MIT Press.
Offit, P. A.
(2011). Deadly choices:
How the anti-vaccine
movement threatens us all. New York: Basic Books.
Ohlsson,
S. (2009). Resubsumption: A possible mechanism for conceptual
change and belief
revision. Educational
Psychologist, 44,
20-40.
Ohlsson,
S. (2011). Deep learning:
How the mind
overrides experience. New York: Cambridge University
Press.
Oreskes, N. (2004). Beyond
the ivory tower: The
scientific consensus on climate change. Science,
306, 1686.
Peirce, C. S.
(1878). How to
make our ideas clear. Popular Science Monthly, vol. 12,
pp. 286-302.
[Reprinted in N. Houser and C. Kloesel (Eds.), The essential
Peirce: Selected
philosophical writings (vol. 1, pp. 124-141). Bloomington,
IN: Indiana
University Press.]
Pettigrew, T.
F. (1998).
Intergroup contact theory. Annual Review of Psychology,
vol. 49, pp.
65-85.
Polk, T.
A., & Seifert, C. M., (Eds.), (2002). Cognitive
modeling.
Cambridge, MA: MIT Press.
Posner, G. J.,
Strike, K. A.,
Hewson, P. W., & Gertzog, W. A. (1982).
Accommodation of a scientific conception: Toward a theory of
conceptual change. Science
Education, 66,
211-27.
Putnam, H.
(2002). The collapse of
the fact/value dichotomy;
and other essays. Cambridge, MA: Harvard University Press.
Robin,
N., & Ohlsson, S. (1989) Impetus then and now: A detailed comparison
between Jean Buridan
and a single contemporary subject.
In D.
E. Herget (Ed.), The
history and
philosophy of science in science teaching.
Proceedings of the First International Conference
(pp. 292-305).
Tallahassee: Florida
State University,
Science Education & Dept. of Philosophy.
Rakison,
D. H., & Poulin-Dubois, D. (2001). Developmental origin of
the
animate-inanimate distinction. Psychological
Bulletin, 127(2),
209-228.
Rokeach, M.
(1960). The
open and closed mind. New York: Basic Books.
Rokeach, M.
(1970). Beliefs,
attitudes, and values: A theory of organization and change.
San Francisco,
CA: Jossey-Bass.
Schiller, F.
C. S. (1905).
The definition of ‘pragmatism’ and ‘humanism’. Mind, 14,
235-240.
Shipstone, D.
M. (1984). A
study of children’s understanding of electricity in simple DC
circuits. European
Journal of Science Education, 6, 185-198.
Sinatra, G.
M., &
Pintrich, P. R., (Eds.), (2003). Intentional
conceptual change. Mahwah, NJ: Lawrence Erlbaum.
Stitch, S. P.
(1983). From folk
psychology to cognitive science:
The case against belief. Cambridge, MA: MIT Press.
Strike, K. A.,
& Posner,
G. J. (1992). A revisionist theory of conceptual change. In R.
A. Duschl and R.
J. Hamilton (Eds.), Philosophy of science, cognitive
psychology, and
educational theory and practice (pp. 147-176). New York:
State University
of New York Press.
Thagard, P,
(1992). Conceptual
revolutions. Princeton, NJ: Princeton University Press.
Vosniadou, S.,
Baltas, A.,
& Vamvakoussi, X., (Eds.), (2007). Reframing
the conceptual change approach to learning and instruction.
Amsterdam, The
Netherlands: Elsevier Science.
Vosniadou, S.,
& Brewer,
W.F. (1992). Mental models of the earth: A study of conceptual
change in
childhood. Cognitive Psychology, 24,
535-585.
Vosniadou, S.,
&
Skopeliti, I. (2013). Conceptual change from the framework
theory side of the
fence. Science &
Education, DOI
10.1007/s11191-013-9640-3.
Wason, P. C.,
&
Johnson-Laird, P. N. (1972). Psychology of reasoning:
Structure and content.
London, UK: B. T. Batsford.
Watson, J. D.,
& Crick,
F. H C. (1953). A structure for deoxyribose nucleic acid. Nature,
vol.
171, pp. 737-738.
[1] Indeed, many philosophers insist that unless knowledge is true and certain, it does not qualify as knowledge.