b2692 Doing Science: In the Light of Philosophy

C hapter 1

IN THE BEGINNING WAS THE

PROBLEM

To engage in research of any kind is to work on a problem or a cluster of

problems of some kind cognitive, technological, social, artistic, or
moral. In imitation of Johns gospel, we may say that in the beginning was
the problem. So, those wishing to start doing science must find or invent a
problem to work on, as well as a mentor willing to guide them.

1.1 At the Source

Free agents prefer to work on problems they like and feel are equipped to
tackle. But of course most budding scientists are not fully free to choose:
their supervisors or employers will assign them their tasks for one does
not know ones own ability before trying and, above all, because finding a
suitable problem is the first and hardest step.
However, problem choice is only part of a whole package, which
includes also such noncognitive items as advisors suitability and availabil-
ity, research facilities, and financial assistance. In other words, the budding
scientist or technologist does not enjoy the luxury of picking his/her
favorite problem which is just as well because, given his/her inexperi-
ence, that choice is likely to be either too ambitious or too humble. In
sum, aspiring investigators are given to choose among a set of packages
offered by his/her prospective advisor or employer.
For better or for worse, there are no recipes or algorithms for generat-
ing problems other than reviewing the recent literature. In particular,
computers cannot pose problems, for they are designed, built, and sold to
help solve well-posed problems, such as curve fitting a given set of data

b2692_Ch-[Link] 1 11-Nov-16 [Link] PM

 b2692 Doing Science: In the Light of Philosophy

points. After listening to Stanislav Ulams panegyric of the abilities of com-

puters, I left him speechless by asking him, at a congress packed with sages,
whether such marvels might invent new problems. He paused for a long
while and finally admitted that this question had never occurred to him.
Such is the power of raw data and data-processing devices.
Half a century ago, Alan Turing proposed the test that bears his name
as the way to discover whether ones interlocutor is a human or a robot.
Later work in AI showed that Turings test is not foolproof. There is an
alternative: ask your interlocutor to pose a new and interesting question.
Computers will fail this test, because they are designed to operate on algo-
rithms, not to deal with questions that demand invention, in particular
problems, such as guessing intention from behavior. This test is therefore
one about natural intelligence, or thinking out of the digital box.

1.2 Types of Problems

The logical positivists like Philipp Frank, as well as their critic Karl Popper,
banned questions of the What-is-it? type. By contrast, the great physi-
ologist Ivan Pavlov (1927:12) held that they exemplify what he called the
investigatory reflex, for they elicit an animals response to environmental
changes. Indeed, they constitute existential dilemmas, hence the most basic
of all, for they include Friend or foe?, Safe or risky?, Edible or ined-
ible?, and the like.
Admittedly, only humans and apes capable of communicating with us
via sign language or computers will formulate problems in a sentence-like
fashion. But this is a moot point: what matters mostly is that the animals
that do not solve their existential dilemmas are unlikely to survive
unless they are tenured philosophy professors.
The importance of problems in all walks of life is such, that someone said
that living is basically tackling problems. For those who have solved the sub-
sistence problem, to live is to fall in and out of love with cognitive, valua-
tional, or moral problems. Those of us who ask Big Questions, such as How
and why did civilization start? are called bold scientists. And the few who ask
the biggest questions of all, such as What exists by itself?, What is truth?,
and Is science morally neutral? are called philosophers. These and similar
questions are transdisciplinary, whereas all the others are unidisciplinary.

2 Doing Science: In the Light of Philosophy

b2692_Ch-[Link] 2 11-Nov-16 [Link] PM

 b2692 Doing Science: In the Light of Philosophy

1.3Erotetics
Most philosophers have overlooked problems and their logic, namely ero-
tetics, which should be the subject of countless original philosophical
research projects. The next few pages will recall the authors erotetics
discussed in what is likely to have been the first treatise in the philosophy
of science to sketch it (Bunge 1967b, vol. 1).
Whatever the kind of cognitive problem, we may distinguish the fol-
lowing aspects of it: (a) the statement of the problem regarded as a member
of a particular epistemological category; (b) the act of questioning a
psychological subject; and (c) the expression of the problem by a set of
interrogatives or imperatives (the linguistic aspect). In the present section
we shall focus on the first of these aspects.
From an action-theoretic viewpoint, a problem is the first link of a
chain: Problem Search Solution Check. From a logical point of view,
the first link may be analyzed into the following quadruple: background,
generator, solution (in case it exists), and control or BGSC for short.
Let us clarify the preceding by way of a hot example in astrophysics,
namely What is dark matter? We start by reformulating the given prob-
lem as Which are the properties P of the Ds?, or (?P)Px, where x desig-
nates an arbitrary member of the class D of all possible pieces of dark
matter, and P a conjunction of known and new physical properties. The
BGSC components of this particular problem are

Background B = Contemporary astrophysics plus particle physics.

Generator G = P x, where P = a conjunction of first-order properties.
Solution S = The cluster P of properties assignable to any D.
Control C = The laboratory analysis of a piece of dark matter or
of the radiation (other than light) that it emits.

Let us close by listing the elementary problem forms.

Which-problems Which is (are) the x such that Px? (?x)Px

What-problems Which are the properties of item c? (?P)Pc
How-problems How does c, which is an A, happen? (?P)[Ac Pc]
Why-problems Which is the p such that q? (?p)(p q)
Whether-problems What is the truth-value of p? (?v)[V(p) = v]
Inverse problems Given B and A B, find A. (A?)[AB]

In the Beginning was the Problem 3

b2692_Ch-[Link] 3 11-Nov-16 [Link] PM

 b2692 Doing Science: In the Light of Philosophy

The direct/inverse distinction may be summarized thus:

Direct Input System Output

Inverse Output System Input

In the simplest case, the inputoutput relation is functional, and it can be

depicted as follows:

Direct x f f(x)
Inverse f(x) f-1 x

However, most real-life problems of are of the means-end kind, most of

which have multiple solutions, so they are not functional.
Whereas direct problems are downstream, or from either causes or
premises to effects or conclusions, the inverse ones are upstream, or from
effects or theorems to causes or premises. A common inverse problem is
that of conjecturing a probability distribution from statistics such as average
and mean standard deviation. A far less common inverse problem is the
axiomatization of a theory known in its ordinary untidy version (see
Chapter 7).
Like most inverse problems, axiomatics has multiple solutions. The
choice among them is largely a matter of convenience, taste, or philoso-
phy. For example, whereas an empiricist is likely to start with electric cur-
rent densities and field intensities, the rationalist is likely to prefer starting
with current densities and electromagnetic potentials, because the latter
imply the field intensities (see Bunge 2014; Hilbert 1918).
All the prognosis problems, whether in medicine or elsewhere, are
direct, whereas the diagnostic ones are inverse. For example, having diag-
nosed a patient as suffering from a given disease on the strength of a few
symptoms, checking for the occurrence of further symptoms is a direct
problem. But the problem of medical diagnosis is inverse, hence far harder,
for it consists in guessing the disease from some of its symptoms.
Ordinary logic and computer algorithms have been designed to handle
direct problems. Inverse problems require inventing ad-hoc tricks, and
such problems have multiple solutions or none. For example, whereas 2 +
3 = 5, the corresponding inverse problem of analyzing 5 into the sum of
two integers has four solutions.

4 Doing Science: In the Light of Philosophy

b2692_Ch-[Link] 4 11-Nov-16 [Link] PM

 b2692 Doing Science: In the Light of Philosophy

Inverse problems may be restated thus: given the output of a system,

find its input, mechanism of action, or both. That is, knowing or guessing
that A B, as well as the output B of a system, find its input A or the
mechanism M that converts A into B. For example, given a proposition,
find the premises that entail it; design an artifact that will produce a desired
effect; and given the beam of particles scattered by an atomic nucleus, guess
the latters composition, as well as the nature of the scattering force (For
the pitfalls of this task see, e.g., Bunge 1973a).
A fever may be due to umpteen causes, and its cure may be achieved
through multiples therapies, which is why both biomedical research and
medical practice are so hard (see Bunge 2013). As a matter of fact, most
inverse problems are hard because there are no algorithms for tackling
them. This is why most philosophers have never heard of them. The ref-
erees of my first philosophical paper on the subject rejected it even while
admitting that they had never encountered the expression inverse prob-
lem (Bunge 2006).
Finally let us ask whether there are insoluble problems. Around 1900
David Hilbert stated his conviction that all well-posed mathematical prob-
lems are soluble in principle, not just unsolved up to now. Here we shall
disregard unsolvable mathematical problems because they are arcane ques-
tions in the foundations of mathematics, and anyway they have raised no
philosophical eyebrows. We shall confine ourselves to noting that some
seemingly profound philosophical problems are ill posed because they pre-
suppose a questionable background.
The oldest and most famous of them is, Why is there something rather
than nothing? Obviously, this question makes sense only in a theodicy
that supposes that the Deity, being omnipotent, had the power of inaction
before setting out to build the universe: why bother with real existents if
He could spend all eternity in leisure? Taken out of its original theological
context, the said question is seen to be a pseudoproblem, hence not one
that will kindle a scientific research project. In a secular context we take
the existence of the world for granted, and ask only particular existence
problems, such as Why do humans have nails on their toes?, which is
asked and answered by evolutionary biologists. The answer is of course
that toenails descend from the fingernails that our remote ancestors had on
their hind legs, which worked as hands.

In the Beginning was the Problem 5

b2692_Ch-[Link] 5 11-Nov-16 [Link] PM

 b2692 Doing Science: In the Light of Philosophy

Yet it is often forgotten that all problems are posed against some con-
text, and that they vanish if the context is shown to be wrong. Let us recall
a couple of famous games of this kind.
Pseudoproblem 1: What would happen if suddenly all the distances in the
universe were halved? The answer is in two parts: (a) nothing at all would
happen, for all distances are relative, in particular relative to some length
standard, which would also shrink along with everything else; and (b)
since no universal shrinkage mechanism is known in physics, the said
event should be regarded as miraculous, hence conceivable but physically
impossible.
Pseudoproblem 2: What is the probability that the next bird we spot is a
falcon, or that the next person we meet is the pope? Answer: neither
belonging to a given biospecies nor holding a particular office are random
events, so the given questions should be completed by adding the clause
picked at random from a given population (of birds or people respec-
tively). No randomness, no applied probability. In conclusion, unless the
background of a question is mentioned explicitly, it wont start a research
project.
A final warning: genuine cognitive problems are not word games
played just to exercise or display wit. The best known of these games is
perhaps the Liar Paradox, generated by the sentence This sentence is false.
If the sentence is true, then it is false; but if it is false, then it is true.
The paradox dissolves either on noticing that the sentence in question
conflates language with metalanguage; or that it does not express a propo-
sition, for propositions have definite truth-values.
The first interpretation warns against such confusions, and the second
reminds us that only propositions can be assigned truth values, whence it
is wrong to call first-order logic sentential calculus, the way nominalists
do just because of their suspicion of unobservables. In sum, avoid barren
paradoxes when stating a cognitive problem, for truth is not a toy.

1.4 The Search for Research Problems

How does one find a suitable research problem? The answer depends on
the kind of problem: is it a matter of survival, like finding the next meal
ticket; a technological problem, such as increasing the efficiency of an

6 Doing Science: In the Light of Philosophy

b2692_Ch-[Link] 6 11-Nov-16 [Link] PM

 b2692 Doing Science: In the Light of Philosophy

engine; a moral problem, such as how to help someone; or an epistemic

problem, such as to discover how dark holes arise or evolve?
The question of problem choice has mobilized psychologists, historians,
and sociologists. These experts have attacked what Thomas Kuhn (1977)
called the essential tension. This is the choice between a potboiler that
may inflate the investigators CV but wont alter anyone elses sleep, and a
risky adventure with an uncertain outcome that may alter an important
component of the prevailing worldview, as was the case when Michael
Faraday assumed that electric charges and currents, as well as magnets,
interact via massless fields rather than directly.
Familiar examples of the first kind are spotting a previously unknown
celestial body of a known kind, the chemical analysis of a newly discovered
wild plant, and computing or measuring a well-known parameter with
greater precision. In contrast, looking for evidence of the ninth planet of
our solar system, digging for hominid fossils in a newly found archaeo-
logical site, and searching for a better cancer therapy, are instances of long-
term and risky projects. They are risky in the sense that one embarks on
them even while fearing of wasting time and resources.
Unsurprisingly, in every walk of life traditionalists outnumber innova-
tors. However, though real, the conservative/innovative tension is tran-
sient, since the initial success of a groundbreaking research project is bound
to attract droves of researchers who inaugurate a new tradition.
Occasionally, even the news that an established scientist is trying a new
approach would have the same result, namely the sudden recruitment of
hundreds of young researchers working on the same project. This used to
happen in particle physics in the 1950s and 1960s, when certain new
theories became instantly fashionable for a few months. Some of the most
ambitious projects attracted exceptionally able investigators, and remained
in fashion even if they failed to deliver the promised goods. String theory
is one of them. Nowadays this theory resembles the aging hippie who
keeps wearing his jaded jeans.
However, the conservative/innovative distinction is best drawn once
the problems concerned have been worked out. The first question one
should tackle is about the main sources of problems, and this question is
tacitly answered the moment a problem typology is proposed. For exam-
ple, moral, political, and legal problems arise only in the course of social

In the Beginning was the Problem 7

b2692_Ch-[Link] 7 11-Nov-16 [Link] PM

 b2692 Doing Science: In the Light of Philosophy

interaction. Thus, Robinson Crusoe felt no moral qualms before meeting

Friday. As soon as this meeting occurred, each of them must have asked
himself how best to treat the other: as friend, foe, or neither; as competitor,
cooperator, or neither and so on.
In contrast, pure curiosity prompts us to asking epistemic questions,
such as whether the squared root of 2 can be expressed as the ratio of two
whole numbers; whether dark matter is anything other than clumps of mat-
ter whose constituent atoms have fallen to their lowest energy level; and
whether primitive living cells might soon be synthesized in the laboratory.
Collecting problems into a number of different boxes or kinds sparks
off yet another problem, namely that of the possible links among such
boxes. A familiar member of this kind is the relation between science and
technology. The standard answer to this question is that science generates
technology, which in turn poses scientific problems, so that each feeds the
other. Let us briefly consider the four most popular answers, and then a
fifth, namely whether philosophy too can meddle with science, now help-
ing, now obstructing it.
So far, the corresponding findings of the historians and sociologists of
science, technology, and philosophy on the above questions constitute a
motley collection of isolated items that are so many problems. Suffice it to
list the following famous items:

a. What led to the discovery of irrational numbers, that is, numbers

that are not ratios of integers like 2/3? Short answer: the wish to cor-
roborate Pythagorass postulate, that the basic constituents of the uni-
verse are whole numbers. According to legend, the member of his
fraternity who dared disputing this conjecture, to the point of proving
that the square root of 2 is irrational, was put to death. In short, his
research project had a philosophical motivation.
b. Why did some ancient Greek and Indian thinkers hold that all things
in the world are combinations of bits of fundamental or indivisible
things? Perhaps because carpenters, bricklayers, and other craftsmen
make artifacts by assembling or dividing things. Those were deep
thinkers, not problem-solvers, interested only in ordinary life issues.
Parallel: the metamathematician who asks not what the result of a
computation is, but what kind of animal a computation is.

8 Doing Science: In the Light of Philosophy

b2692_Ch-[Link] 8 11-Nov-16 [Link] PM

 b2692 Doing Science: In the Light of Philosophy

c. Why did Erathostenes wonder about the shape and size of planet
Earth? Presumably, because he was curious and examined some of the
indicators that our planet was not flat, such as the seeming sinking of
westbound ships.
d. What led Empedocles to suspect that the biospecies had evolved instead
of being fixed? Perhaps he was led by religious skepticism jointly with
the finding of marine fossils on the top of certain mountains.
e. What led Olaf Rmer to design his ingenious device to measure the
speed of light in 1676, when light was generally regarded to travel at
infinite speed? Likely, it was his observation that Jupiters eclipses
were seen at different times at different times of the year, when that
planet was at two different places. So, light had to take some time to
travel from Jupiter to Earth. His was a case where a research project
sprang from a surprising observation. Likewise, in 1820 Johannes
Mller undertook to measure the speed of nervous pulses out of sheer
curiosity the main fountain of science according to Aristotle.
f. Why did Newton try out the inverse square hypothesis? As told by the
diarist Samuel Pepys, a number of amateurs who used to meet at a
coffee house surmised that our solar system was held together by an
unknown force, and one of the habitus offered a substantial prize to
the first man to propose a plausible solution. Newton had just crafted
the first theory that allowed the exact statement of this inverse prob-
lem, which he solved by transforming it into a direct problem: calcu-
late the planet trajectory assuming that the force binding it to the sun
is inversely proportional to the square of their mutual distance. This,
Newtons great inverse problem, rather than Humes problem of
induction, is the one that started a new epoch in theoretical science.
g. In 1688 the physician Francesco Redi put to the test the popular
hypothesis of spontaneous generation, by isolating a morsel of meat.
Contrary to expectations, no flies appeared; they only came from lar-
vae that deposited on rotting meat. Lazzaro Spallanzani and others
corroborated this result during the next three centuries: over that
period the maxim omne vivo ex vivo held sway. The materialist hypoth-
esis (abiogenesis) that the earliest organisms had been born spontane-
ously (without Gods help) from the synthesis of abiotic materials was
regarded as having been falsified once and for all.

In the Beginning was the Problem 9

b2692_Ch-[Link] 9 11-Nov-16 [Link] PM

 b2692 Doing Science: In the Light of Philosophy

h. In 1862 Louis Pasteur, the founder of microbiology, endeavored to

find out whether his microbes might develop into more complex
organisms. To this end he made a jar of sterile nutrient broth, which
he boiled. That is, he unwittingly made sure that no life germs
remained. Unsurprisingly, microscopic observation showed no signs
of life. Once again, experiment was regarded as having killed the
myth of spontaneous generation. At the time no one dared analyze
Pasteurs experimental design. He succeeded in falsifying a myth
for a while.
i. In 1953 the chemistry graduate student Stanley Miller and his mentor,
the seasoned physicist Harold Urey, synthesized some amino acids and
other organic molecules out of a mixture of methane, ammonia, water,
and hydrogen, which they subjected to electric discharges on the
assumption that the primitive atmosphere had been similarly electri-
fied. True, the outcome was inconclusive, for no living things were
produced. But suddenly the synthesis of life became a serious research
project, and it did so propelled by the philosophical hypothesis that life
might indeed have emerged from lifeless precursors as any good
materialist would surmise.

After more than one century, the MillerUrey project is still been pursued
even though their result had fallen into the crack between the two traditional
categories, confirmed and falsified, as shown in the following diagram.

Confirmed
Experimental outcome / Inconclusive
\ Falsified

We shall return to this problem in Chapter 3. Suffice it for now to note

that the standard philosophies of science, in particular confirmationism and
falsificatonism, take it for granted that (a) all experimental results are une-
quivocal, and (b) experiment is the undisputed umpire that arbitrates
between competing hypotheses. Clearly, these two pillars of the standard
philosophy of science are cracked. And assuming that hypotheses can only
be more or less probable wont repair those pillars, for assigning hypoth-
eses subjective (or personal) probabilities is unscientific (Bunge 2008).

10 Doing Science: In the Light of Philosophy

b2692_Ch-[Link] 10 11-Nov-16 [Link] PM

 b2692 Doing Science: In the Light of Philosophy

1.5 Problem System

It is well known that problems come in all kinds and sizes. There are
cognitive and moral problems, individual and social issues, scientific and
technological conundrums, political issues, and so. Then again, problems
can be either local or systemic, and they may be tackled either by indi-
vidual experts or by multidisciplinary teams.
Small problems call for the use of known tools found in circumscribed
fields, whereas big problems call for further research, which may require
breaching disciplinary walls. For example, whereas an experienced bone-
setter may fix a fractured bone, a problem concerning invisible entities may
require interdisciplinary research.
Another philosophically significant partition of the set of scientific
problems is the direct/inverse dichotomy. The problem of induction is
the best known of all the inverse philosophical problems. It consists in
leaping from a bunch of data to a universal generalization. For example,
given a bunch of scattered dots spread on a Cartesian grid, find a smooth
curve joining them. As every calculus student knows, the standard solu-
tion to this problem is the interpolation formula invented by James
Gregory in 1670. This is an nth power polynomial f(x) constructed from
a set of n + 1 values of f plus the smoothness assumption.
Gregorys formula is only good to handle low-level variables, such as the
stretching and load of spring scales (Hookes law), and Ohms law, which
relates the voltage and current intensity of a direct-current electrical circuit.
The higher-level law statements, such as those occurring in electrodynamics,
cannot be reached by induction because they go far beyond data.
The inception of Gregorys curve-fitting method was neither inductive
nor deductive: it was an invention or abduction, as Charles Peirce
might have called it. Only the problem that prompted it, namely going
from a bunch of data to a general formula, deserves being put under the
heading of induction.
Incidentally, nowadays Gregorys invention is usually called the Newton
Gregory formula. This is an instance of the Matthew effect, studied by
Robert Merton, whereby a minor scientist is hitched to a famous one in
order to highlight the findings importance. Another, far less important
case, is that of the FeynmanBunge coordinate.

In the Beginning was the Problem 11

b2692_Ch-[Link] 11 11-Nov-16 [Link] PM

 b2692 Doing Science: In the Light of Philosophy

Going back to Hume, he is unlikely to have heard of the problem in

question, much less of Gregorys solution to it. Nor did Popper, who three
centuries later claimed to have solved what he called Humes problem.
Note that Gregorys formula is only good to handle low-level variables,
such as the stretching and load of spring scales (Hookes law). Higher-level
law statements, such as those occurring in electrodynamics, cannot be
reached by induction because they go far beyond data.
The hardest problems are the so-called Big Questions, such as the one
concerning the origin of life. As reported above, in 1953 Harold Urey and
Stanley Miller approached this problem from scratch, that is, from simple
molecules such as hydrogen, water, methane, and ammonia, they did not
get living things but instead nucleotides-essential components of the DNA
molecule. Thus, Miller and Urey had produced the first solid result in
synthetic biology, after Aleksander Oparins bold and inspiring work in the
1920s. The same year 1953, help came from an unexpected quarter: the
molecular biology due to Francis Crick and James Watson. This break-
through suggested an alternative strategy for creating life in the lab, namely
starting from highly complex organic molecules, such as nucleosides,
instead of proceeding from scratch.
This new strategy soon yielded some sensational results, such as Har
Gobind Khoranas synthesis of a gene in 1972, and Craig Venters synthesis
of the whole genome of a bacterium in 2010. Although synthetic life is still
only a promise, no biologist doubts that it will be achieved in the foreseeable
future by following the method of building increasingly complex systems by
joining lower-level entities.
To conclude, note that the standard philosophies of science, in particular
confirmationism and falsificatonism, take it for granted that (a) all experi-
mental results are unequivocal, and (b) experiment is the undisputed umpire
that arbitrates between competing hypotheses. Clearly, these two pillars of
the standard philosophies of science are cracked: some experimental results
are inconclusive, and a research program may be pursued despite setbacks,
as long as it is backed by a strong philosophical hypothesis, such as that
ofabiogenesis.

12 Doing Science: In the Light of Philosophy

b2692_Ch-[Link] 12 11-Nov-16 [Link] PM