Transcript Prediction in Context: On the Comparative Epistemic Merit of

Prediction in Context:
On the Comparative Epistemic Merit of
Predictive Success
Martin Carrier
(Bielefeld University)
1.
Predictive Novelty vs. Coherence as Epistemic Merits
Prediction stands out among the
traditional criteria of epistemic
merit in science.
Heather Douglas (2009) demanded
to supplement the emphasis on
explanation with an equal emphasis
on prediction.
Her claim is that ceteris paribus prediction is epistemically
superior to accommodation.
The epistemic role of predictions is to provide demanding test
instances for theories; predictions serve to test explanations.
However: conflict between the requirement that a theory produce
novel predictions and the demand that it match the state of
knowledge.
Novel predictions anticipate surprising
phenomena that go beyond the system of
knowledge and tend to be inconsistent
with it.
Thus predictive power and coherence with
the background knowledge are conditions
that tend to point in opposite directions.
The existence of competing standards suggests that the emphasis
on prediction shouldn’t be exaggerated.
What is the merit of predictions if science leaves the controlled
conditions of the lab and struggles with the intricacy of the world?
2.
Prediction in Application-Oriented Research Endeavors
In application-oriented research, prediction is assumed to play a
key role.
Targeted intervention in natural
processes requires
the ability to anticipate the results of
one’s action.
=> Ambivalent views on the proper role of prediction in science:
It is complained, on the one hand, that explanation has eclipsed
prediction, whereas it is contended, on the other, that prediction
has eclipsed explanation.
=> Question as to the relationship of predictive strength to
explanatory power.
Biotechnology of the 1990s: the ability to
anticipate the outcome of one’s interventions is decoupled from any deeper
understanding.
Predictive strength and theoretical understanding are assessed as contrasting
virtues.
In contrast: the power to anticipate is more
dependent on the ability to explain than
many practitioners believe.
Use of trigger genes for getting a complex
activity of gene expression underway.
Trigger genes act as starters so that switching them on allows one to set a cascade of
processes in motion.
=> There is no need to disentangle this chain of events in order to
anticipate the effect.
The activation of eyeless initiates
eye formation in flies:
appropriate stimulation produces
eyes in the legs or wings of flies.
The identification of trigger genes
was assumed to enable biotechnologists to anticipate the outcome of their intervention without being able to follow out the
underlying causal chains.
Artificially induced eyes in drosophila
Biotechnologists: prediction does not always rely on understanding.
The transition from genomics to proteomics involves a conceptual
revolution which does not affect, however, the technical role of the
gene.
Biotechnologists: Genes can still be used as tools for bringing about
effects and anticipating the outcome of an intervention.
=> The correctness of predictions is independent, in large measure,
from the truth of more fundamental, high-brow theories.
However, this one-sided
emphasis on prediction and
the corresponding neglect of
explanation turned out to be
a failure.
In general, the genetic and
non-genetic contexts need to
be taken into account.
“Distalless” gene: acts
in a more specific way
and affects embryonic
development differently.
Caterpillar embryo:
formation of legs.
developed butterfly:
wing pattern.
The strong emphasis on prediction in application-oriented research
can be seen as a bias that may hurt the epistemic culture of
science.
Worse yet, neglecting explanation may hurt predictive strength.
Douglas requires to bring prediction back into philosophical
accounts of explanation.
But: converse demand to
bring back explanation into
accounts of prediction.
Predictive strength needs to
be placed in a network of
other merits and achievements.
Interconnection Processes
3.
The Role of Prediction in Expert Judgment
So far: In some practical contexts predictive power does not play
the outstanding role sometimes accredited to it.
Prediction is a great team player but a lousy soloist.
Now: scientific expertise: policy advice on the basis of scientific
knowledge.
Problems in need of expertise are
often complex.
But often the details can be ignored in favor of expounding the
distinctive features.
“Epistemic robustness”: the
recommendation remains unchanged if the pertinent causal
factors and factual conditions
fluctuate or are unknown.
Epistemic robustness is an important objective for scientific
expertise since addressing the
minute particulars is often
immaterial for deciding about
how to respond to a practical challenge.
Robustness—and what it‘s worth
Robustness is a quality standard characteristic for expertise. It replaces precision as a chief virtue of predictions in epistemic science.
The commitment to epistemic robustness tends to reduce the
importance of accurate predictions.
Widespread belief: expert recommendations may be based on
observational regularities alone.
=> Explanation and prediction are severed from each other.
However: observed
regularities are paradigm instances of nonrobust knowledge.
Tying prediction to
explanation is sometimes a good guide for
enhancing the robustness of the prediction.
4.
The Significance of Prediction in Pursuing Research
Policies
Possibility to anticipate the success of certain types of research
projects.
Prospects of demanddriven research policies:
outcome-focused projects guided by practical
requirements.
The successful pursuit of demand-driven
research programs requires to anticipate
probable novel results emerging along
certain research paths.
However: it is often deemed impossible to foresee the outcome
of research projects.
Vannevar Bush (1945): science policy for
stimulating progress in practical matters:
funding of basic research.
The assumed impossibility of anticipating
research outcome is taken to make any
demand-driven research policy fail; only a
knowledge-driven policy can be expected to
yield fruitful results.
However, this unpredictability claim is in need of serious
qualification.
Successful demand-driven
research:
The
Human
Genome
Project
The Manhattan Project
The empirical
record of attempts
to produce research
outcome on
demand is mixed.
The relevant variability can be accounted for to some
extent by the degree to which the
pertinent theoretical framework was
staked out.
The Discovery and Use
of Giant Magnetoresistance
Yet incomplete knowledge of the fundamentals does not
always thwart practical research endeavors.
The “X1,” the first
supersonic aircraft,
was designed in 1947
by ignoring fluid
dynamics and by
relying instead on
practical knowledge
and on experience
with high speeds.
=> It is a precarious endeavor to predict possible research
outcome and to assess the prospects of research projects
Prima-facie plausibility: the successful stimulation of demanddriven research requires understanding the epistemic processes in
science.
Bringing Bacon’s claim about the
relationship between understanding and
intervention to bear on science studies
produces the assertion that a targeted
intervention in science presupposes a
thorough understanding of science.
Yet Bacon’s claim is not universally true.
Some technological breakthroughs rely
on fundamental understanding but
others don’t.
5.
Conclusion
Prediction plays a less than
pronounced role in the context
of practice.
It is true, predicting the results
of human action is a key feature
of science turned practical.
However, bare predictions do
not count for much.
First, predictions need to be integrated into an explanatory framework if they are supposed to guide actions reliably.
Second, the preference for precise predictions in epistemic research
is supplanted with the objective of specifying a robust corridor of
estimates.
Finally, it is highly uncertain to predict the success of research
projects.