Você está na página 1de 9

Lakatos

The big Lakatosian idea is that the difference between the Newtonians theory-saving manoeu-
vres and the flat-earthers is that the Newtonians programme is progressive, the flat-earthers is
degenerating. The suggestions and the terminology is due to Lakatos: and now is the moment
to read his great paper The Methodology of Scientific Research Programmes (or at least the
first half of it).
At the risk of spoiling some of the excitement!Ill indicate straight away the view of Laka-
tos that I take, so that youll have some sense, as we go along, of how various remarks fit into
the overall story. I want to argue that

1

Lakatoss

starting point

is broadly correcttheories

are

complex affairs, involving
something like a hard core we try to preserve, a protective belt of auxiliaries, and so on
(what the and so on amounts to, well see in just a moment). Thus, when we learn a
theory, we dont just learn a bunch of sentenceswe learn something more like a theo-
retical tradition, learn what is essential and what peripheral, gain some sense of how to
apply the theory to new exemplars, etc. So far, so good. However, when Lakatos moves
on from an

intuitive

understanding of (say) what it is for a theoretical traditiona
research programmeto be in a progressive or degenerating phase, and tries to give
more

formal

characterizations, then his account becomes much more questionable.

2

Although much of Lakatoss rhetoric is Popperian (indeed he presents himself as simply
drawing out themes already there, albeit not always in sharp form in Popper), in fact his
position is in some ways a lot closer to Kuhns than he acknowledges. The title of his
key paperThe

Methodology

of Scientic Research Programmespromises us a
methodology for science. We might expect, then, to nd Lakatos advancing some princi-
ples for rational theory choice (and after all, he does represent himself as aiming to
counteract Kuhnian irrationalism about theory choice). But it fact, when the wraps are
off, he does

not

have a contentful overall methodology for rational theory.

3

There remain very important respects in which Lakatos does differ from Kuhn (these
certainly amount to more than mere matters of emphasis, though they are fall well short
of giving Lakatos a methodology):
He castigates the vice of

ad-hockery

(this is a theme common to Lakatos and
Popper).
He emphasizes the virtue of

proliferation

(this is a theme common to Lakatos and
Feyerabend);
What I mean here will become clear (I hope!) as we proceed.

The Lakatosian descriptive apparatus, in a little more detail

Lakatoss core idea is that, in a key sense, the unit of appraisal is the research programme (which
generates a sequence of theories), not an individual theory. So what, more carefully, does he
mean by a research programme?
The basic components of a research programme are its


hard core

a central body of laws and principles that are held xed, and kept immune
from falsication. This core is surrounded by a protective belt of


auxiliary hypotheses and assumptions

a shifting, evolving, body of additional theoret-
ical propositions that are combined with the hard core (and statements of initial condi-
tions) to yield testable predictions. So it is the hard core plus the auxiliaries which form

a testable, potentially refutable, specic variant of the programme. (These specic
variants are likely, one after another to prove false in the end. But the resulting pessimis-
tic induction neednt show that there is anything wrong with the hard corethe blame
may be put to the account of the auxiliary hypotheses and assumptions.)
The research programme also embodies constraints on what are allowed as auxiliary
hypotheses: this is the


negative heuristic

the basic negative rule is dont change the hard core laws and prin-
ciples. So, more specically, a research programme might embody rules like dont
allow unmediated action at a distance, dont allow violations of the conservation of
energy, dont allow violations of the conservation of electric charge, dont allow
backwards causation, and so forth.
An example: it was found that


-decay (when a nucleus decays into a lighter nucleus
by emitting an electron) apparently violates the conservation of energy. In a particular
type of decay, the emitted electrons can have widely varying energies. So Pauli in the
1930 postulated that there must be another particle involved in the decay, to carry how-
ever much energy is required to conserve total energy. This would have to be a charge-
less particle to avoid violation of the conservation of chargewhich is why the particle
is difcult to detect experimentally.
Note that, in principle, there will be an innite range of auxiliary hypotheses that are
available to us to accommodate some phenomenon in accordance with the sort of nega-
tive heuristic rules mentioned so far. For example, in the neutrino case, we could have:


-decay involves the emission of an electron plus a neutrino


-decay involves the emission of an electron plus two tweutrinos


-decay involves the emission of an electron plus three threutrinos
and so on and so forth, not to mention


-decay involves the emission of an electron plus a neutrino and a shmeutrino
and the like. (OK, there may be extra constraints that knock out some of these crudely
extravagant hypotheses. But the basic idea is right. We can always spin new hypotheses
by multiplying entities beyond necessity.)
So as well as negative rules specic to a particular research programme (like dont
allow action at a distance, dont allow inheritance of acquired characteristics) there
will be more general negative rules that are sustained across theories, like

dont multiply
entities beyond necessity

.
The neutrino hypothesis exhibits a certain positive pattern of thought. Here we have a
problemapparent failure of conservation of energy. And we look for a solution that
conforms to a certain patternpostulate a new basic sort of particle to go alongside the
electrons, protons, neutrons and so forth that have already been recognized. The new
particles strategy has worked before, so Pauli in 1930 is working to a pattern that has
already begun to prove its worth. In a phrase, in this area of research, there was already
developing a

positive heuristic

i.e. a body of positive models, analogies, strategies,
pointers to where to look for developments and solutions to problems.
Another example from modern physics:
Weve just seen how we protect conservation of energy by postulating new particles.
Conservation principles are very simple yet powerful, so we dont readily want to
give up good candidates. In fact, they are such nice things to have that an attractive
positive strategy for coping with new phenomena is

postulate new conservation
laws



where possible

. For example, we nd a certain pattern of events in the decay of
some heavy particlesknown as baryons in the tradeso that some decays are

allowed and others apparently never found. So we postulate a new property, the
property of having a certain strangeness (zero for neutrons, minus one for Sigma
particles, minus three for the omega-minus) and then we require total strangeness to
be preserved in an interaction. This works a charm! (Quark theory later uses the just
the same strategy, and postulates new propertiesincluding indeed charm!that
are preserved in interactions.)
Lakatos sums things up like this:
The negative heuristic species the hard core of the programme which is irrefutable by
the methodological decision of its protagonists; the positive heuristic consists of a partially
articulated set of suggestions or hints on how to change, develop the refutable variants of
the research programme, how to modify, sophisticate, the refutable protective belt. The
positive heuristic of the programme saves the scientists from being confused by the ocean of
anomalies the scientists attention is riveted on building his models following instruc-
tions which are laid down in the positive part of this programme.

Comments

1.

The obvious key question to raise is:

do

scientists really make a sharp methodological deci-
sion to isolate a hard core in the way that Lakatos alleges. Interestingly, he remarks that
the actual hard core of a programme does not actually emerge fully armed like Athene
from the head of Zeus. It develops slowly, by a long, preliminary process of trial and error
(MSRP, p. 133, fn. 4). But if there has been a lot of trial and error, why not say thatafter
due explorationour scientists simply come to think that certain core claims are unlikely to
be false (rather than assign them a methodological decision to try to hold them true come
what may). And why not allow different, more or less central, claims to be held with varying
degrees of condence rather than make a sharp distinction, hard core vs. the rest?

2.

Note that the just quoted characterization of a positive heuristic is reasonable because rea-
sonably modest (the heuristic is a

partially articulated

set of hints). Elsewhere, Laka-
tos is a lot bolder: he says the positive heuristic denes problems, outlines the construction
of a belt of auxiliary hypotheses, foresees anomalies and turns these victoriously into exam-
ples, all according to a preconceived plan.

That

looks much too strongas if a pro-
grammes defenders are forearmed against (most) difculties.

Progression, degeneration, and ad-hockery

So far, Lakatos can all be taken as giving a framework for the neutral description of what
scientists do (or at least do, some of the time). And it is a tolerably useful descriptive frame-
workat least in rough, general shapegiven we take those notions of hard core, positive
heuristic and so on fairly loosely and set aside the sort of worries of detail just indicated.
Rightly or wrongly, scientists do

not

behave like naive Popperians, and simply reject a
theory whenever they get into trouble. Rather they keep plugging away at their research
programme, despite the anomalies and refutations that beset their current best efforts,
encouraged by whatever partial successes they are having, in the hope that they will
continue to make sufcient advances. Our example last time: Newtonians faced with the
anomalous behaviour of Uranus.
And scientists beset with problems and anomalies usually dont despairingly gaze at the
unlimited range of possible repairs. Rightly or wrongly, they are usually not that open-
minded, but rather pursue a

very

limited range of options guided by past experience.
This description in fact looks pretty reminiscent of Kuhns description of the practice of normal
science. Scientists are, most of the time, in the business of puzzle solving (as he puts it), using

various standard paradigms for guidance. Indeed, in his contribution to

Criticism and the
Growth of Knowledge

, Kuhn in effect complains that Lakatoss descriptive apparatus just

is

a
reworking of his.
Well come back to that issue, the relation between Kuhn and Lakatos, in a moment. But lets
first press on in Lakatosian spirit
Suppose we have some core theory

T

and current auxiliaries

A

which (together with
statements of initial conditions, but well take those for granted) imply

O

predicted

. And
suppose

O

p

is false; and what we nd instead is

O

actual

. We need a repair.
Easy! Introduce a special mirage hypothesis that explains why it seems to be the case
that

O

actual

when the relevant initial conditions are in play.
Easy, but a complete cheat, of course The new auxiliary is purely ad hoc: it has no
virtue other than that of rescuing

T

in the face of

O

a

. It implies no other testable
predictions. The kind of ad-hockery is surely a

VERY



BAD



THING

, on any view.
Lakatos has a number of stabs in different places at giving a clear characterization of
what we want to avoid here. But the basic idea seems reasonably clearwe want new,
improved, auxiliary hypotheses to predict novel facts. That is to say, we want new
auxiliaries to make predictions over and above those they are custom-designed to pre-
dict; and we want some of these additional predictions to be conrmed. (Example:
Hypothesizing the existence of Neptune with a certain orbit is

designed

to predict the
observed orbit of Uranus; but it

also

predicts other, novel, factse.g. that we will tele-
scopically observe a slowly moving point of light at a certain place in the heavensand
some of these novel facts are conrmed.)
Lets say that if we can cope with an anomaly for a theory (hard core plus all the surrounding
auxiliaries) by adjusting the auxiliaries in a way that predicts new factsif we can, so to speak,


do a Neptune

then we have made a

positive adjustment

. And if, as the research programme
evolves, we can make a series of positive adjustments, Lakatos will say that the programme is

progressive

. Thus:
Let us take a series of theories,

T

1

,

T

2

,

T

3

, , where each subsequent theory results from
adding auxiliary clauses to the previous theory in order to accommodate some anomaly,
each theory having at least as much content as the unrefuted content of its predecessor, Let
us say that such a series is theoretically progressive if each new theory predicts some
novel, hitherto unexpected fact. Let us say that a theoretically progressive series of theories
is also empirically progressive if each new theory leads to the actual discovery of some
new fact. Let us call a [programme]

progressive

if it is both theoretically and empirically
progressive, and

degenerating

if not.

Comments

1.

Lakatoss remarks dont cover the general case, because successive theories in a theory-
series in a research programme dont have to be related just by the

addition

of new auxilia-
ries: often we

replace

auxiliaries. But thats not a trivial emendment: if we drop old auxilia-
ries while adding new ones, we lose as well as gain predictionshow do we characterize
progress in this mixed case?

2.

And waiving that, Lakatoss explication of the idea of novel facts as hitherto unexpected
facts certainly wont do as it stands. Suppose the moving point of light in the right part of
the sky had already been observed (but thought to be a distant comet). Then the postulation
of Neptune to explain the orbit of Uranus would have predicted an

already known

fact about
what would be telescopically observed. But the hypothesis would have still counted as dra-
matically conrmed, because the prediction is of something novel in the sense of something

that it wasnt custom-designed to predict.

3.

Again, do we really want to say that a theoretical programme is

degenerating

once it is no
longer progressing? What if it is a limited-aim programme, which has successfully been
worked out in sufcient detail (or where the remaining problems are developing the mathe-
matics to enable us to deal with more and more complex cases, but there is no addition of
empirical auxiliaries to be doneexample, ??maybe the classical theory of isolated electri-
cal circuits??). In fact, Lakatos goes on to say that we should reject degenerating research
programmes as pseudo-scientic. But thats just absurd on his denition of degeneration
(which attens the distinction between lack of further progress and the at-earth case,
where adhering to the core theory leads to an ever spiralling growth of problems).

4.

Still, ignoring the ne print, the basic idea that we should avoid ad-hockery in aiming for (in
some broad, intuitive sense) a progressive research programmes seems good. It helps pro-
vide what Kuhn leaves underspecied, i.e. (in his terms) an account of what amounts to
progress in normal science.

Appraisal and proliferation

Note, however, that a research programmes being progressive is, in itself, not a reason for
hanging on to it, and degeneration (certainly in Lakatoss weak sense) is not, taken by itself, a
reason for giving the programme up.
(To take the second half of that rst): a theoretical programme may be in trouble. But if
it is currently the only game in town and has empirical successes in some areas, then
(even if it is stagnant and we dont seem to be able to develop it further) it may be
rational to keep working at it. Rational policy will depend on how other theoretical
options are doing.
Likewise if a research programme is progressing. Another programme may be doing
sufciently much better that it wouldnt be rational to stick to the rst one. Heres a
mythical illustration. Suppose that while the northeners are developing Newtonian phys-
ics, the antipodean southerners develop relativistic physics (straight off, without going
though a Newtonian phase). And suppose our nineteenth century northern Newtonians
then encounter the southern theorists for the rst time. Our Newtonians programme is
still in a progressive phase (after all, they have just triumphantly discovered Neptune, as
have the antipodeans, similarly saving

their

hard core). But the Newtonians are faced
with the relativistic theorists who can (1) accommodate everything they can accommo-
date, (2) also deal with problems for the Newtonian (like the anomalous behaviour of
Mercury) and (3) predict new surprising new effects which are conrmed (like the grav-
itational effect on light). Our Newtonians should (it seems) be duly impressed and
change allegiance. Their programme stands criticized by a yet more successful one
which in some sense

incorporates

or

takes over

the Newtonian one
In short, then, the rational stance to take vis--vis a particular research programme depends not
just on how it is faring intrinsically but on how things stand

comparatively

. And the most telling
criticism of a programme P1the strongest reason for abandoning itis provided by a second
programme P2 that explains the successes of P1 but also has additional content which is not eas-
ily reconciled with P1 but which is also confirmed.
Now, Lakatos and Feyerabend are unreconstructed Popperians in at least this sensethey
stress the centrality of

criticism.

Rationality in science means, as a minimal core, being prepared
to subject your views to critical test, and being honest about the successes and failures. But if
criticism is the all-important engine of progress, and the most telling criticisms come from the
development of rival programmes, then (it seems) we should positively encourage the prolifer-
ation of theoretical programmes. (Its a sort of Darwinian picture:

proliferating mutation



of pro-

grammes

plus

natural selection by stern empirical test and criticism

will lead to the fittest
theories And evolution goes faster if there are more mutations.) Hence we arrive at the
Lakatos/Feyerabend

normative

claim that proliferation of research programmes is good.
Two points about this:
Note the claim that proliferation

ought

to happen is, of course, consistent with a
Kuhnian

descriptive

claim that proliferation is in fact a rarity, and only occurs in
revolutionary periods, while the normal condition of science in a particular eld is
indeed normal science where a single theoretical tradition holds the eld. But Kuhn, I
think, sees no reason to recommend against this state of affairsso there is a normative
disagreement.
Or at least, there

seems

to be disagreement. But it perhaps rather depends on how ne
you cut research programmes. Lakatos is very hazy about what makes for identity of
research programmes. (It will not do, for example, to say that programmes are [par-
tially] individuated by their hard cores given that, according to him, programmes early
in their career dont

have

determinate hard cores after all!) And it seems

possible

that
the relation between his research programmes and Kuhns episodes of normal science
is many-one. For example, perhaps at various stages in its development, different key
formulations of quantum theory have constituted different

research programmes

(in
Lakatoss sense, having somewhat different core principles) while belonging to one
overall developing episode of

normal science

(in Kuhns sense). I think there are indeed
signs that Lakatos is thinking in more ne-grained terms than Kuhn. So perhaps after
all, Kuhn

could

recognize the proliferation of ne-grained research programmes as a
good thing, for it promotes work within some overall paradigm. Ill not pursue this, but
just note again a real unclarity in the details of Lakatoss position (though to be fair,
there is an equal unclarity in Kuhn).

Method?

We just imagined a rather simple case of comparison where we have a pair of theories, and T2
supplants T1 (relativistic and Newtonian dynamics, say). Lakatos describes such a case thus:
T2 explains the previous success of T1 (all the unrefuted content of T1 is contained
within the limits of observational errorin the content of T2)
T2 has excess empirical content over T1
some of the excess content of T2 is conrmed.
Actually there is trouble lurking here in the idea of one theory have excess truth-content over
another (well come back to this). But the intention is to describe a situation where T2 system-
atically does better than T1, and so provides a telling criticism of T1and we should thus clear-
ly prefer T2. Lakatos puts it this way: for his sophisticated form of Popperianism,

this

is what
falsification of a theory comes tothe production of a better theory, with excess corroborated
content. And the rational move, faced with a falsification in this sense, is perhaps clear
enough.

An aside on crucial experiments

: This reconstruction of the notion of falsication explains
why Lakatos, like Kuhn, is leery of the old idea of the crucial experimenti.e. a batch of
key observations that refute a theory. The anomalous behaviour of Mercurys orbit was
known for decades in the nineteenth century. But it wasnt treated as a refutation of Newto-
nian celestial mechanics. Then along comes Einstein, proposes an improved dynamics
whichit turns outgets it right (more or less) about Mercurys orbit. So the new theory
accommodates everything the old theory did, plus some. This phenomenon which the new
theory explains and the old theory cant is

then

promoted to the status of a crucial experi-

ment for the old theory. But it isnt the observations that refute the old theoryit is the
observations

combined with the new theory

.
Note, however, the case just described where T2 falsifies T1 in this sense is unusual when we
are making inter-programme comparisons. (

Within

a research programme, when things are pro-
gressing nicely, later theory versions will often falsify earlier onesi.e. be able to incorporate
or explain what was successful about the earlier versions, and add some additional confirmed
content. Indeed, thats Lakatoss official definition of progress. But now fix on the case where
we are making comparisons between theory-versions belonging to different programmes.) In
the inter-programme case, the Einstein trumps Newton scenario is exceptional. More com-
monly, there may be rival programmes that each have their own successes and failures T2 ex-
plains some of previous success of T1, and has confirmed excess content, but T1 also has some
successes (apparently) lacked by T2. In this sort of case, where we dont have a straightforward
falsification of T1 by T2, what makes for rational theory choice?
Lakatoss intentions here are elusive. Or at any rate, they are elusive enough for Ian Hacking
to write that Lakatos doesnt have a methodology in the sense of a set of recommendations for
theory-choice, while Bill Newton-Smith thinks that Lakatos did intend his so-called method-
ology of scientific research programmes to be used in making action-guiding decision with
regard to theory choice in contemporary science.
1. (The Newton-Smith line) Lakatos writes that normal sciencein the sense of developmen-
tal work within a research programmeis ne, so long as we keep in mind the Popperian
recognition that there is good, progressive normal science and there is bad degenerating nor-
mal science, and as long as we retain the determination to eliminate, under certain objec-
tively dened conditions, some research programmes. And Lakatos then talks of editors of
journals refusing to publish papers in degenerating programmes and says research founda-
tions, too, should refuse money. Which indeed sounds pretty prescriptive!
But how is the prescription to be spelt out? Suppose we try: if research programme R2 is
progressing and rival programme R1 is degenerating, we should stop working on R1. But
that plainly wont do. Maybe R1 is just temporarily stuck in a degenerating phase and on the
verge of another great leap forwardLakatos himself notes the possibility, and allows that
it is rational to stick to a programme going through a bad patch.
Indeed, part of his beef against more nave forms of falsicationism is that new theoreti-
cal frameworks do need to be allowed time to develop. Over-enthusiastic pruning of tender
shoots that look a bit sickly would lead to the elimination of growths that eventually prove
fruitful. One must treat budding programmes leniently. (To take a standard example,
Copernican cosmology was born refuted. If the earth is spinning, and the surface is moving
at over a thousand miles an hour, how come we are not spun off it; how come that a weight
dropped from a high tower doesnt fall some distance behind it as the earth rushes on? And
so forth. But further work on the programme led to triumphant vindication.)
2. So, at best, the idea is going to have to be if programme R1 gets bogged down over the long
term then we should abandon it. But then how long is the long term?
Feyerabend argues, in effect, that if long term is left entirely vague, then the methodo-
logical advice is empty. But if we try to ll in details, well either get something too restric-
tive again (there are plenty of cases were bad patches have been pretty long-lived), or with a
generous enough setting of the time limit, well get something useless (roughly: dont og a
horse that is dead, dismembered, and six foot under).
Lakatos, in History of Science and Its Rational Reconstructions notes this line of criti-
cism. He responds that Feyerabend conates methodological appraisal of a programme
with rm heuristic advice about what to do. It is perfectly rational to play a risky game;
what is irrational is to deceive oneself about the risk. But then he still goes on to talk about
research foundations refusing money to degenerating programmes. Thats far too quick.
Leave aside the provocation about funding (if it is rational to keep working on a risky pro-
gramme, why isnt it rational and proper to keep funding it?). And if methodological
appraisal is to have any content at all, dont we want it to be able to distinguish the risky
game from the hopelessly lost gamesay, pursuing hidden variable theories in QM (risky)
versus at earthism (a hopeless cause)? Both programmes are degenerating in Lakatoss
sense: so what makes the difference (surely any methodology should address this question)?
3. (The Hacking line) Hacking implies that this sort of question is misplaced for Lakatos, for
(despite those provocative remarks about funding and so forth) he means his talk of method-
ological appraisal to be purely backward lookinghe really is not trying to give action-
guiding advice, but only talking of after-the-event appraisal.
The idea, perhaps, is this; we can look back at episodes in the history of science and say
that these episodes of theory-change were scientically irrational and those episodes were
rational. Soviet scientists in the 1930s may have been externally rational in adhering to the
Lamarkian idea of the inheritance of acquired characteristicsbetter than being sent off to a
labour camp for deviance from the party line. But it wasnt a scientically rational choice:
the programme was degenerating. On the other hand, the widespread adoption of the quan-
tum theory by the 1930s (despite all its anomalies and apparent incoherencies) was
rationalthe programme was spectacularly progressive.
And the thought might be that here we have the makings of a riposte to Kuhnor at
least, the Kuhn of a common reading, the sceptic about the rationality of theory choice, who
holds that theory choice is not methodologically constrained (Thomas Kuhn invites us to
think of scientic progress as exercises in imitation interrupted by changes in fashion.
Sylvain Bromberger. And sociologists of science have leapt to that reading, and seen Kuhn
as advocating some kind of social relativism, that cant distinguish the Soviet Lamarkians
from the early quantum theorists).
4. But again this is too quick. Given Lakatoss recognition that it can be rational to stick to a
research programme even when it is in the doldrums, the adoption of quantum theory in the
1930s was a rational choice, but not the unique rational choice. The classical framework
had been so successful (and early quantum theory was in such a conceptual mess) that it
would have been quite rational for most people to have kept plugging on in the old ways for
a bit longer. So if there was a bandwagon effect in the 1930s, thenat least by Lakatoss
own standardsit wasnt simply down to considerations of rational theory-choice. Which is
a Kuhnian position.
The real Thomas Kuhn protests against misreadings of his position that make him out
hold that scientic change is simply non-rational (it is an exercise of rationality if anything
is). He insists that considerations of accuracy, scope, simplicity, fruitfulness normally
play their part in theory change. So the real Thomas Kuhn would agree that the Soviet
Lamarkians differ from the quantum theorists; he would emphasize, however, that change
after periods of crisis is not fully dictated by canons of reason. And it seems that Lakatos
will, by his own lights, have to agree. So is Lakatos, at bottom, just a Kuhn in Popperian
dress?
Summary
1. A competent theory of theories is going to have to recognize some complex structures (the-
ories arent just bunches of sentences). And something like Lakatoss categories look useful.
Though we saw that e.g. the way he spells out the idea of hard core or progressiveness is
open to question. And he certainly doesnt say enough about how you individuate hard
cores, programmes, and so on. So there is much work to be done on the descriptive catego-
ries (which doesnt mean that his story, crude though it is, isnt useful to set against Popper
and Kuhn).
2. Lakatos castigates ad-hockery (Kuhn says less about what makes for cheating in normal
science); and Lakatos praises proliferation (which certainly goes against the tenor of Kuhns
emphasis on the supposed monolithic character of normal science). But, in the end, Lakatos
apparently does not have a contentful overall methodology for rational theory-choice.

Você também pode gostar