[1] Theories Methods Research.ws2010.Slides.I

Prof. Paul Hoyningen-HueneUniversity of Hannover

Theories and Methods of ResearchWinter term 2010-11

1

Theories and Methods of

Research

Prof. Paul Hoyningen-Huene

Institute of Philosophy

Formalities

• Structure of the course:

– 11 sessions with lectures (20.10.2010 – 12.1.2011)

– 3 sessions with oral presentations of research projects by M.Sc. 13 students from International Horticulture (mandatory), 19.1., 26.1. and 2.2.2011

– Presentations: 15 min each: 5 min project, 5 min connection to lectures, 5 min discussion

• Second year students from Water Resources: Written presentation of research project (max 1 page) and connection to lectures (max 1 page), to be handed in by 2.2.2011, 12 noon

22

Formalities (2)

• Printout of PowerPoint slides can be downloaded at stud.IP

• Recommended book: Alan F. Chalmers (1999): What is This Thing Called

Science? Third ed. St. Lucia: University of Queensland Press.

• List of oral presentations for MSc 13 students will be circulated next week

• Written examination at the end: week following the end of term 7 Feb - 11

Feb 2011 (probably 9 Feb 2011, 10:15)

• Grading: 20% oral/written presentation of research project, 80% written

examination

• My email address: [email protected]

• My official office hours: Tue 16-17, Im Moore 21, back building, 4th floor.

33



2

1. Introduction

The fundamental questions of general philosophy of science:

What is science?

Why is scientific knowledge different from other forms of knowledge, e.g., more reliable?

What are the characteristic features of scientific knowledge?

44

Introduction (2)

The course will present four specific answers to these questions which have been given during the last one hundred years:

• Inductivism

• Deductivism

• Paradigm theory

• Systematicity theory

After an attempt to present these answers as persuasively as possible, criticism of them is raised which forces the development of new answers.

55

2. Inductivism

2.1 The basic idea of inductivism

In science and engineering, we find singular statements and general statements

• Singular statements

• This lake is contaminated by pesticides and heavy metals

• This plant has nine leafs

True singular statements express singular facts

In science and engineering, singular statements often express observable (and/or measurable) facts

In this case, it is easy to verify (to find out the truth of) the statement : observe or measure and you will know

66



3

Inductivism (2)

• There are different kinds of general statements in the sciences

• natural laws like Newton’s law of gravitation:

F = γ m1 m2 / r2 : they are supposed to be certain

• theories: this may mean

• very well established very general statements like Darwin’s theory of evolution or quantum theory

• General statements about which you are unsure: “it’s just a theory”; also the term “hypothesis” is used in this sense

Example: “My theory (hypothesis) is that plants of species X are not seriously affected by a change in the environmental condition Y”

77

Inductivism (3)

• (empirical) regularities: a general, regular relation between different variables

• models: typically a simplified image of a general relation among different variables (e.g., model organism, model of an economy, model of a lake)

General statements (laws, theories, general hypotheses, regularities, models) can not be verified in the same way as singular statements as they refer to indefinitely many facts

Thus, the “generalization” of singular statements to general statements is problematic

For example, consider the generalization from the observation of some finite number of swans that are white to: “All swans are white”

88

2.1 The basic idea of inductivism

The problem how to correctly generalize from singular statements and the justification of this step is called "the problem of induction".

In dealing with this problem, inductivism states that scientific knowledge is gained in a two step process:

Step one: Carefully observe singular facts without any theoretical prejudice ("theory-free facts") and articulate them in singular observation sentences. These singular observation statements can deliver an objective, and therefore also inter-subjective, description of facts.

99



4

The basic idea of inductivism (2)

Step two: Generalize singular observation statements to

general hypotheses (or "laws", "theories", "models").

This step, the "inductive generalization", is allowed if the

following three conditions are met:

C1: The number of observations must be large.

C2: The observations must be made under very different

conditions.

C3: None of the observation statements must contradict the

general hypotheses.

1010


General statements (theories, etc.) are needed for scientific

explanations, predictions and technical applications

a) Explanations and predictions

Both activities follow the same pattern:

Here is a particular event S0 of type S (description of a situation).

General statement: Whenever an event of type S, then an event

of type K (general hypothesis).

Therefore: K0 (conclusion).

1111


Example: Predict how far a mass falls within one second.

Law of free fall: d = ½ g t2; g = 9.8. Situation S0: t0 = 1

Therefore: d0 = 4.9.

The conclusion follows "deductively" (or "by logical deduction")

from the premises

By means of this pattern, a K0 that has not yet happened can be

predicted if S0 and the general hypothesis are known

If K0 has happened, it can be explained by recourse to S0 and the

general hypothesis

1212



5


b) Technical application

Technical action is doing one thing in order to realize another thing

You want to generate an event of type K

General statement: Whenever an event of type S, then an event of type K (general hypothesis)

Thus: realize S0 and you will get K0

Example: You want to regulate the level L of a lake

Find the general relation among L and the relevant variables: outflow, inflow, evaporation, etc. (a model of the lake)

Manipulate outflow such that the desired L results

1313


Thus according to inductivism, scientific research consists of an inductive procedure by which scientific hypotheses are gained, and logical deductions from these hypotheses by which predictions and explanations are given.

Scientific knowledge is different from and more reliable than other forms of knowledge because the two steps that generate it are much more rigorously controlled than the steps that lead to other forms of knowledge.

1414

2.2 Historical remark

From about the beginning of modern science in the

17th century and until the 20th century,

inductivism was so wide-spread that the natural

sciences were even called "inductive sciences".

1515



6

2.3 Normative versus descriptive

philosophy of science

There seem to be two principal ways of answering the

questions posed in section 1: What is science?

• One can make an empirical investigation into the

peculiarities of science by observing what scientists

really do.

This would be a descriptive procedure.

• One may consider how science should proceed in order

to produce reliable knowledge.

This is a normative procedure.

1616

Normative versus descriptive

philosophy of science (2)

In section 2.1, we obviously followed a

normative procedure.

The resulting philosophy of science is therefore

called normative.

1717

2.4 Problems of inductivism

There are six main problems that inductivism faces

1. Condition C2, which constrains the possibility of inductive

generalizations, is vague and if taken literally, can never

be fulfilled: it is not clear which conditions have to be

varied and which do not.

If no directives are given, infinitely many conditions can be

varied such that one is never allowed to proceed to the

inductive generalization.

Example: Influence of garlic on the function of compasses –

was a real question in the 16th century!

1818



7

Problems of inductivism (2)

2. The observation of facts must be somehow directed: only

relevant facts should be observed.

But the evaluation of facts as relevant presupposes

theoretical elements.

Thus, it is impossible to really gather theory-free facts

that are relevant to the hypothesis in question.

Example: Without the difference between (physical)

solutions and (chemical) compounds, the inductive

generalization to the law of constant proportions is

impossible.

1919


3. Even a singular observation statement is not theory-free.

For instance, a statement like "This is a plant of species X" uses the concepts of "plant" and "species".

These concepts are highly theory-laden: for instance, plants are assumed to be distinct from animals (among other things), and the concept of species assumes that biological entities can be ordered into fairly well-defined classes.

2020


4. The justification of the inductive

generalization is probably impossible.

Most attempts to justify it end up in circularities.

Example: Inductive generalizations are justified

because Nature is ordered by regularities

(laws). But how do you know that these

regularities will hold in the future?

2121



8


5. According to inductivism, disagreement

between scientists can only be the result of

mistakes; a “rational disagreement” is

impossible

This is highly implausible

2222


6. Inductivism is certainly not generally valid.

For instance, it is impossible to gain knowledge

of the structure of DNA (or any other highly

theoretical scientific conception) by inductive

generalization from phenomena.

2323


Note that difficulties 2 and 3 are not deadly to inductivism.

They show, however, that a build-up of scientific knowledge from a theory-free basis is an illusion.

Thus, inductivism looses much of its attractiveness because theoretical elements enter already at the level of observation.

The position dealt with next, deductivism, tries to draw consequences by allowing theoretical elements to enter science at all levels, and by avoiding the inductive generalization because it seems unjustifiable.

2424



9

Literature

Chalmers, A.F. (1999): What is This Thing Called Science?

Third Edition. Indianapolis: Hackett (especially chapters 1 - 4)

Losee, John (1993): A Historical Introduction to the

Philosophy of Science. 3rd ed. Oxford: Oxford University Press (an overview over the historical development of philosophy of science).

Salmon, M. H./J. Earman/C. Glymour/J.G. Lennox/P. Machamer/J.E. McGuire/J.D. Norton/W.C. Salmon/K.F. Schaffner (1992): Introduction to the Philosophy of

Science. Englewood Cliffs NJ: Prentice-Hall (a book on an advanced level).

2525

3. Deductivism

3.1 The basic idea of deductivism

Deductivism is an attempt to develop a position that avoids the difficulties that beset inductivism.

It is accepted that theoretical elements enter science at all stages and that inductive generalizations lack proper justification.

The basic idea of deductivism is that theories are not built bottom-up from theory-free data, but that they are deductively tested against data.

Inductivism and deductivism share the view of scientific explanation, prediction and technical application

2626

3.2 The demarcation criterion

In order to establish that deductive testing procedures are applicable, scientific hypotheses, theories, etc. must fulfill a specific condition.

This condition demarcates scientific hypotheses from others, like metaphysical, religious or pseudo-scientific ones.

In the given context, to be a "scientific hypothesis" does not imply that the hypothesis has been accepted or confirmed by science; it only means that it can be admitted to scientific testing procedures.

2727



10

The demarcation criterion (2)

Demarcation criterion:

A hypothesis is scientific if and only if it is empirically falsifiable, i.e., that there are empirical circumstances imaginable such that the hypothesis is refuted.

Thus, a scientific hypothesis is in principle empirically testable.

An example of a hypothesis that is not falsifiable is "The universe is governed by love and hate.“

Whatever happens empirically, it can be subsumed under this hypothesis. Thus, this hypothesis cannot be empirically refuted; and therefore, it is not a scientific hypothesis.

2828


In other words: scientific hypotheses must be empirically risky:

There is always danger that by means of empirical evidence they are shown to be false

2929


For inductivism, the process of producing a hypothesis is a rule-governed process.

By contrast, for deductivism the production of hypotheses is not constrained as long as the resulting hypotheses are empirically falsifiable.

Example: Kekulé’s idea of the ring structure of the benzene molecule, published in 1865

As legend has it, Kekulé came up with the idea of this structurein a dream of snakes. Each of the snakes bit into the tail ofanother snake. As a result, the snakes depicted a cyclicstructure.

Whether the hypotheses are (temporarily) accepted in science, however, will be determined by the testing procedures.

3030



11

3.3 Deductive hypothesis testing

The main tool for the empirical test of scientific hypotheses is logical deduction.

From a general hypothesis, specific sentences are deduced that can be compared with empirical data.

Example: Ohm's Law: R = U/I.

For a specific piece of matter, by simple mathematical manipulation, the following equation can be deduced:

U1/I1 = U2/I2.

This equation can be compared with the results of measurements.

If the measurement differs significantly from the deduced values, the hypothesis failed the test and it is falsified; otherwise it survived the test.

Hypotheses are kept in science as long as they have not been falsified.

3131

Deductive hypothesis testing (2)

Thus, the answer of deductivism to the fundamental questions of general philosophy of science (see sect. 1) is:

Science is the continual invention of falsifiable hypotheses and their critical empirical test.

Science does not try to confirm hypotheses, but to disconfirm (or falsify) them.

The spirit of science is thus fundamentally critical.

Science tries to be as unbiased and dogma-free as possible by testing all proposals as severely as possible.

This is well-expressed in the title of Karl Popper's 1963 book: Conjectures and Refutations.

3232

3.4 Consequence: fallibilism

In spite of all of the critical tests of a given hypothesis, it never becomes absolutely certain, even if it has passed all of the tests without being falsified.

That a new test will be devised which falsifies the hypothesis is never excluded.

As a consequence, scientific knowledge never becomes infallible, i.e. absolutely certain.

It is bound to be fallible.

This insight rejects the pre-dominant (Western) ideal of science from antiquity up until very recently, according to which scientific knowledge should be infallible.

3333



12

Consequence: fallibilism (2)

According to deductivism, scientific knowledge

consists of hypotheses that are accepted until

they are falsified.

Therefore, science is an intrinsically dynamic

enterprise as hypotheses have to be

continually tested.

To stop testing hypotheses is to stop doing

science.

3434


Deductivism was developed in the 20th century,

mainly in the work of Karl Popper (1902-

1994).

Many scientists, including social scientists,

adhere to this position.

Deductivism is also called "falsificationism",

"critical rationalism", and the "hypothetico-

deductive account of science".

3535

3.6 Problems of deductivism

Deductivism is attractive because it seems successfully to exploit the "asymmetry between verification and falsification of general hypotheses“:

The verification of a general empirical hypothesis, i.e. a definitive proof of its truth, is impossible because for general statements one would have to make infinitely many tests (the problem of induction, see section 2.1).

The definitive falsification of a general empirical hypothesis, however, seems possible because a single fact that contradicts the hypothesis suffices.

3636



13

Problems of deductivism (2)

On closer inspection, however, the situation is more complicated.

The falsification of a hypothesis by, say, some measurement, assumes that the measurement is reproducible.

But this assumption makes use of an inductive generalization, namely, that the same measurement performed tomorrow will yield the same result.

Thus, a definitive falsification is also not possible.

3737


2. The rule that science should never stop its attempts to critically test hypotheses leads to unintended and unpleasant results.

Suppose you have a hypothesis that states a specific functional dependence of a variable y on a variable x: y = f(x).

Real life examples:

• Boyle’s gas law pV = k, or p = k/V

• Ohm’s law: I = V/R

The test of the hypothesis y = f(x) does not only include that y does indeed co-vary with the variable x in the way the function f states.

3838


It must also include the test of the assumption that y

depends on the variable x only, and not on additional

variables u, v, w, etc., too.

However, the list of potentially influential variables is

indefinite (it includes the haircut of the

experimenter, the position of Venus, etc.).

Thus, a systematic test of the potential dependence of

y on u, v, w, etc. cannot be carried out; testing

cannot even approximately be complete.

3939



14


In more general terms, the problem consists in this.

Any account of the development of science must

contain the following three elements:

a principle for the generation of hypotheses;

a principle for the elimination of hypotheses;

a principle for the (perhaps only temporary)

acceptance of hypotheses that allows for

(temporarily) halting tests and accepting a

hypothesis in order to be justified to apply it.

4040


The fundamental problem of deductivism consists in its lack of a principle for the (perhaps only temporary) acceptance of hypotheses.

Deductivism cannot stop testing hypotheses, thus never allowing the move forward to applications.

The paradigm theory to be discussed next will provide an answer to this problem.

4141

Literature

Chalmers, A.F. (1999): What is This Thing Called

Science? Third Edition. Indianapolis: Hackett

(chapters 5 - 7)

Popper, Karl R. (1968): The Logic of Scientific

Discovery. Second Edition. London: Hutchinson

(especially chapters 1 and 2)

4242



15

Paradigm theory

4.1 The basic idea of paradigm theory

Paradigm theory begins with a comparison of the normative positions (inductivism, deductivism) with the history of science which results in a strong discrepancy:

In many cases scientists do not behave in the way the normative positions posit.

Thus we have a discrepancy between norms (how science should be done) and facts (how science is actually done).

4343

4.1 The basic idea of paradigm theory

There are two main possibilities to explain this discrepancy:

• either the norms are wrong (for instance: unrealistic) or

• the actual practice of science is bad (for instance: dogmatic).

From the discrepancy alone, one cannot judge whether the

norms should be given up or if the practice of science should

be changed.

Before deciding this question, paradigm theory aims at a

description of the general characteristics of the sciences,

especially of their development over time.

4444

4.2 The developmental model of

scienceParadigm theory describes the general

characteristics of the development of the basic disciplines of the natural sciences (as opposed to the applied ones) by a phase model (or "developmental model").

In such a model, different phases of a science's development are distinguished and their temporal order is described.

In each phase, science is practiced in a specific form.

4545



16

The developmental model (2)

The phase model is a cyclical model (more precisely, a

spiral model) of scientific development. Its structure

is as follows:

P -------> N "P" stands for "pre-normal science"

"N" stands for "normal science"

R "R" stands for "revolutionary science"

4646


P N

R

The arrows in the model should be understood as follows.

From P, a first phase of normal science N1 is reached.

N1 leads to a first phase of revolutionary science R1.

From R1, a new phase of normal science N2

is reached, which in turn leads to a new phase R2, etc.

In the following, the different phases and the transitions between them will be explained.

4747


Pre-normal science is a scientific practice which is not very

structured and in which various schools compete.

They have different fundamental viewpoints about the subject

matter.

There are even disagreements about what does, and what does

not, belong to the field.

Example: Research on electricity in the first half of the 18th century:

different schools:

• Electrostatic attraction and frictional generation of electricity

fundamental, electrostatic repulsion secondary

• Attraction and repulsion equally fundamental

• Conduction fundamental: result of an electric fluid

4848



17


The transition to normal science is called the maturation of the field.

It happens when one of the schools produces an extraordinarily convincing solution to some of the fundamental problems of the field such that the members of the other schools join this school.

This solution must be such that the fundamentals of the field now seem to be clear, and that further research can build upon them.

The decision for the theory associated with the outstanding solution is based on a set of scientific values which include scope, accuracy, predictive power, consistency, and fruitfulness.

Example: Bohr’s atomic model:

• describes distribution of positive and negative charges

• explains discrete emission spectra

• “explains” stability of atoms

4949


Normal science is a scientific practice that is characterized by a

consensus of its practitioners about the fundamentals of the

field.

More specifically, it can be described as follows:

1. The consensus about fundamentals of the field provides

scientists with a framework of re-search which they accept.

2. The basis of the consensus are paradigmatic solutions to

concrete scientific problems, the so-called paradigms.

3. The research activity is implicitly governed by the paradigms.

They function as exemplars of successful research.

5050


4. Research has analogies to puzzle-solving (in chess puzzles, or crossword puzzles, or Sudoku puzzles)

- There are an (implicit) set of rules which constrain the solution process and the criteria for the identification of completed solutions.

- There is an expectation of the existence of a solution. If one fails to achieve a problem solution, either the problem was badly selected or one failed to solve a solvable problem.

- The research activity does not aim at innovation of the guiding framework, i.e., it takes the paradigmatic solutions for granted.

- Contrary to inductivism and deductivism, the research activity cannot be seen as testing or confirming the guiding rules of science -- if anything is tested, it is the scientist! Rather, it is the production of new knowledge on the basis of already existing knowledge.

- The motivation for the activity is less the desire to posses the solution, but rather the exposition of one's ability to creatively solve problems.

5151



18


5. There is a quasi-dogmatic element: the

framework is not called into question.

This is called "quasi-dogmatic" because

- this sort of dogmatism does not hold forever

(see transition to revolutionary science), and

- it is not negatively evaluated.

5252


The transition to revolutionary science happens when significant anomalies turn up which hinder the practice of normal science.

Significant anomalies are anomalies that shed doubt on the prevailing framework of research.

In normal science, there are always anomalies, i.e., problems that should be solvable but aren't at the moment

Example: For Darwin’s theory in his time: sex ratio of 1:1 not explainable

Because there are always many possible causes for these anomalies, usually the guiding framework is not called into question.

But significant anomalies do just that.

Whether an anomaly is a mere anomaly or an significant one is a question of judgment on which scientists may disagree.

Factors influencing this judgment include a clustering of anomalies or the persistence of anomalies in spite of prolonged efforts of the best scientists.

5353


Revolutionary science aims at establishing a new framework for research which makes normal science possible again.

That is, it aims at new paradigmatic solutions.

In the new framework, some of the significant anomalies must be solvable, and the main achievements of the old framework must be preserved, or must at least be reproducible.

Competing theories are evaluated according the scientific values mentioned above (transition to normal science).

Revolutionary science differs from pre-normal science in that there is a consensus about the significant anomalies which must be solved, and there is the need to preserve or reproduce the results of an earlier accepted theory.

However, there are also similarities to pre-normal science which include a lack of general consensus and the existence of competing schools.

5454



19


The transition to a new phase of normal science is called a scientific revolution.

It happens in analogy to the transition from pre-normal to normal science due to an extraordinary achievement

Typically, the old phase of normal science is incommensurablewith the new phase of normal science (or the old and the new guiding theory).

This means that they lack a common measure in three senses:

- they use different sets of concepts, and some of the new concepts cannot be expressed in terms of the old concepts and vice versa

Example: wave function in quantum mechanics; concept of ether in field theory

5555


- slightly different sets of data are seen as relevant.

This is because the choice of data and the methods of their measurement may be influenced by the theories which make use of them

Example: unaccounted rest of 42”/century of the total Mercury perihelion precession of 570”/century

- somewhat different scientific values may play a role in the evaluation of competing theories because these values may depend on the theories in question

Example: deterministic theories are highly valued in classical physics whereas probabilistic theories are accepted in quantum physics

5656


It is important to note that incommensurability does not make theory comparison impossible, but it makes it more difficult than commonly assumed.

This latter fact helps explain why theory choice in science is sometimes a lengthy process.

Incommensurability does not only apply to the relationship between normal science traditions of fundamental science, but also to the relationships between the schools of pre-normal science and, in somewhat modified form, to the relationships between different cultures.

5757



20

4.3 The argument against

normative philosophies of science

The most important tension between the normative

philosophies of science (inductivism and deductivism) and

the descriptive paradigm theory concerns normal science.

From the point of view of the normative philosophies,

normal science is a bad practice of science because of its

quasi-dogmatic element.

The question arises whether the normative philosophies are

right, in which case most of modern science is indeed bad

science, or whether the normative philosophies are wrong

in this respect, in which case normal science is a good

scientific practice (of course, both could be wrong).

5858

Against normative philosophies (2)

Paradigm theory defends the practice of normal science

• by showing that this practice serves vital functions for the goal of science (knowledge production), and thus

• that normal science is a reasonable scientific practice, and consequently,

• that the normative philosophies of science are wrong in this respect.

The argument runs as follows.

5959


The quasi-dogmatic element of normal science provides the

principle which was missing, especially in deductivism: a

principle for the (perhaps only temporary) acceptance of

hypotheses that allows for (temporarily) halting tests.

In this way, emerging difficulties with a theory are not

immediately taken as indicators of its failure in principle, but

as exposing the incapability of the respective scientist.

In this way, the real potential of a theory can be probed.

It has often happened in the history of science that problems

could eventually be solved in the existing research framework.

6060



21


However, if people work on a significant anomaly within the old framework, they may get a deeper understanding of where exactly the failure of the old theory lies, and this may prove extremely important for the invention of alternatives.

Without this principle for the acceptance of theories, scientist would tend to jump from one theory to the next without ever knowing their real strengths and weaknesses.

Thus, the quasi-dogmatic element of normal science serves the aim of science and it is reasonable.

6161


Paradigm theory was first articulated by Thomas

S. Kuhn (1922-1996) in his 1962 bestseller The

Structure of Scientific Revolutions.

There was a wide-spread discussion of paradigm

theory in many fields which involved many

misunderstandings.

6262

4.5 Problems of paradigm theory

The phase model of paradigm theory essentially makes a statistical statement of the form:

"Typically (or: In most cases), science develops in such a way . . ."

In the systematic social sciences such as psychology or sociology, such statistical statements have to be validated by statistical methods applied to historical evidence.

However, paradigm theory has not done that.

Thus, the validity of paradigm theory’s general statements about the development of scientific disciplines is uncertain.

6363



22

Problems of paradigm theory (2)

There are serious doubts that the phase model applies throughout.

Example: Development of General Relativity Theory (GRT) from 1915 onwards

There was very little normal science after 1919 (when the solar eclipse confirmed GRT’s prediction about the bending of light due to the Sun)

However, in the 1950s people started inventing alternatives to GRT although there were no significant anomalies!

6464

Problems of paradigm theory (3)

In answering this line of criticism, one might simply accept it with the remark that the difficulty mentioned is typical for many general statements about historical processes.

Very often, limited access to historical data prohibits the use of proper statistical tools.

The systematic social sciences are in a much better position in this respect because they can often produce the sort of data that are necessary for proper statistical analysis.

6565

Literature

Chalmers, A.F. (1999): What is This Thing Called Science?

Third Edition. Indianapolis: Hackett, chapter 8.

Hoyningen-Huene, Paul (1993): Reconstructing Scientific

Revolutions. Thomas S. Kuhn's Philosophy of Science. Chicago: University of Chicago Press.

Kuhn, Thomas S. (1970): The Structure of Scientific

Revolutions. 2nd ed. Chicago: Chicago University Press.

Preston, John, 2008: Kuhn's "The structure of scientific revolutions": a reader's guide. London; New York: Continuum Logo.

6666

Documents

[1] Theories Methods Research.ws2010.Slides.I