EPISTEMOLOGY – lec 4 Flashcards
Holism
about justification for a given kind of belief is the view that a belief of the kind is never
justified in isolation: any factor that contributes to justifying it does so only in relation to an
array of other beliefs, so that the factor really contributes to justifying a whole network of beliefs
– the new belief together with the background array.
Validity
An argument is valid iff it is not possible for its premisses to be true while its conclusion is
false.
non-ampliative reasoning
To reach a conclusion by a valid argument is to reach it by non-ampliative reasoning (the reasoning is ‘non-ampliative’ in that it takes you to a conclusion that is already guaranteed to be true given the
premisses).
Ampliative reasoning
Ampliative reasoning is reasoning that takes you to a conclusion whose truth is not guaranteed by the premisses. There are hard questions about the criteria of legitimacy for reasoning of this kind. The two most famous patterns of ampliative reasoning are inference to the best explanation and inductive
generalisation:
inference to the best explanation
X
The best explanation for X is Y
Y
inductive generalisation
All observed F’s (all F’s in the observed sample)
have been G.
All F’s are G
One argument for holism about justification in scientific theory-building concerns the reasoning
by which the empirical data might be held to support the theory
1 The empirical data cannot be held to provide non-ampliative support for the theory.
So
2 The empirical data can be held to provide only ampliative support: the data must be held to
support the theory by inference to the best explanation and/or inductive generalisation.
But
3a The justification conferred by an instance of inference to the best explanation is (at best)
holistic.
3b The justification conferred by an inductive generalisation is (at best) holistic.
So
4 The means by which the empirical data can be held to provide support for the theory confer (at
best) holistic justification.
Validity vs. soundness
It follows from the definition of ‘validity’ that a valid argument may very well not prove its conclusion.
For example, consider an argument with contradictory premisses. There are no possible circumstances in which all of this
argument’s premisses are true. So there are no possible circumstances in which both all of its premisses are true and its
conclusion is false. So the argument is automatically (‘trivially’) valid.
An argument proves its conclusion iff (a) it is valid, and (b) its premisses are all true. In this case the
argument is sound.
In general, working out whether a (deductive) argument proves its conclusion will have three steps:
1 Work out what the premisses and conclusion actually are.
2 Decide whether the argument is valid.
3 Decide whether its premisses are true.
Premiss 1 – The claim that the data cannot provide non-ampliative support for the theory
Where a body of empirical data supports a theory, it does so because the theory predicts the data:
the theory entails that X will be observed, and X is observed. So to claim that the body of data
provides deductive/non-ampliative support for the theory is to argue like this:
If the theory is true, X will be observed. X is observed, so the theory is true.
And this is a bad argument. (Compare ‘If I’m in Toronto, I’m in Ontario. I’m in Ontario, so I’m in Toronto.’)
Premiss 3a Inference to the best explanation and holism
A claim that Y is the best explanation for X assumes
a) a standard against which to measure what counts as a good explanation;
b) an account of which other possible explanations are available.
So data X can support theory Y only against the background of other beliefs that combine to
provide (a) and (b) in the overall pattern of reasoning. These beliefs will, in turn, owe their
justification to the fact that the theory they favour (Y) predicts a phenomenon that is actually
observed.
Philosophical tools 5 – Validity vs. soundness
It follows from the definition of ‘validity’ that a valid argument may very well not prove its conclusion.
For example, consider an argument with contradictory premisses. There are no possible circumstances in which all of this
argument’s premisses are true. So there are no possible circumstances in which both all of its premisses are true and its
conclusion is false. So the argument is automatically (‘trivially’) valid.
An argument proves its conclusion iff (a) it is valid, and (b) its premisses are all true. In this case the
argument is sound.
In general, working out whether a (deductive) argument proves its conclusion will have three steps:
1 Work out what the premisses and conclusion actually are.
2 Decide whether the argument is valid.
3 Decide whether its premisses are true.
Premiss 3b Inductive generalisation and holism
An inductive generalisation assumes that the initial sample (the group of observed F’s) is
representative of the population to which the generalisation reaches (all F’s).
But claims about what makes a sample representative are claims about the kinds of properties
that contribute to determining various kinds of difference between particulars. So inductive
generalisation from observation about observed F’s to a theory about all F’s will itself depend on
background beliefs about the kind of difference that makes a difference, and this kind of
background belief is itself justified partly by the role it plays in generating theories that are
successful in predicting observations.
The argument from the nature of scientific observation
Scientific observation is not just a passive matter of absorbing observable aspects of the world.
Rather, it is a practice that is governed by scientists’ background theories. In a given case,
background theory determines which experiments are done and how the experimental apparatus
are set up; it determines which effects produced by the running of the experiment are counted as
relevant observations; it determines what is counted as ‘repeating’ the experiment.
Given that scientific observation is theory contaminated in this way, the claim that observation X
arrived at by experiment e should be counted as data to be explained by our overall scientific
account of the world depends for its justification on background theory.
Underdetermination
Definition To say that a theory is ‘underdetermined’ by the data that are claimed to support it is
to say that these data could be explained just as well by rival theories.
Once it is allowed that justification in scientific theorising is holistic, it is hard to resist the
conclusion that scientific theories are underdetermined by their supporting data.
But it is important to distinguish between two different underdetermination claims:
The strong underdetermination claim Where Y and Y* are rival revisions of an initial scientific
theory in the face of observation X, there is never any rational reason to prefer Y to Y.
The weak underdetermination claim Where Y and Y are rival revisions of an initial scientific
theory in the face of observation X, there is sometimes no rational reason for prefer Y to Y*.
The arguments for holism that we have seen do not generate a version of holism that warrants the
strong underdetermination claim. There is a very hard question about how weak an
underdetermination claim we can get away with.
Underdetermination and the realism/anti-realism debate
3.i Strong underdetermination and scientific realism
If a strong underdetermination thesis is true, the scientific realist’s conception of correctness of
scientific theorising in terms of generating a theory that matches reality cannot be sustained. (If
strong underdetermination is true, the development of a scientific theory is not responsive to
what the world is actually like in the way that would be required to sustain the claim that
scientific theorising has ‘matching reality’ as its standard of correctness.
Underdetermination and the central argument for constructive empiricism
We saw the class before last that the central argument for constructive empiricism involves an
underdetermination claim – the claim that where T is a scientific theory and T* the theory that T
is empirically adequate, the only grounds for preferring T to T* are practical/pragmatic (they are
to do with the fact that T* is less cumbersome than T is).
Though T and T* are distinct theories, they will generate the same predictions in structurally
similar ways. So to say that T and T* are (as far as justification goes) equally good rival theories
is to endorse a comparatively weak underdetermination claim.
