Novel Prediction

Four Key Types:

1. Predictive Novelty : A result not widely known, particularly to the person conducting the experiment. The result is unexpected based on existing theories or knowledge.

2. Temporal Novelty : A result that was not known by anyone at the time the theory was proposed.

3. Heuristic Novelty : A result that did not belong to the problem situation which governed the construction of the hypothesis.

4. Use Novelty : A prediction that is not used to fix the value of a free parameter in a theory or simply built into the theory.

Explanation From Curd and Cover C4 Commentary (512-515):

Hempel does not use the phrase novel prediction. Instead, he writes of “new” evidence and “new” phenomena. The word new is in quotes because the evidence or phenomena need not be new in the sense of not having been previously recorded or observed.

Consider, for example, the case of the Swiss schoolteacher Johann Jakob Balmer (1825-98), discussed by Hempel. On the basis of four observations made by the Swedish physicist Anders Jonas Angstrom (1814-74), Balmer published, in 1885, a general

formula giving the wavelengths of a series of lines—now called the Balmer series—in the visible range of the emission spectrum of hydrogen.

The original data set on which Balmer based his formula was already widely known by scientists at the time. Balmer then used his formula to predict further lines in the hydrogen spectrum, which were subsequently verified, thus confirming his formula.

As Hempel notes, some of the additional lines had in fact already been observed in 1885, when Balmer put forward his theory, but Balmer and most other scientists were unaware of this fact. Thus, to call the later predictions new or novel does not mean

that no one had previously observed them but rather that the results of these observations, were not widely known. In particular, the results were not known to Balmer at the time he proposed his formula.

- Note: This is what is essential, that the prediction does not have to be universally new, but only 'new' to the people conceiving the theory.

We shall introduce the phrase epistemic novelty to distinguish the restricted epistemic sense of predictive novelty —not widely known and not known by the person proposing the theory—from the wider sense of temporal novelty —not known by anyone at the time the theory was proposed.

- Note: In this case, do scientists benefit from

Some philosophers of science have also deemed two further senses of novelty important. We shall call these design-novelty and use-novelty. A result is design-novel so long as the scientist did not deliberately construct the theory to yield the result in question. For example, when Newton devised his gravitational theory, he was perfectly well aware of the precession

of the equinoxes, but this was not an explicit constraint on his theorizing.

He did not, for example, go through several preliminary versions of his theory, rejecting those that did not give the right answer for the rate of precession. Contrast this with Einstein, who did reject earlier versions of the general theory of relativity because they failed to imply the correct value for the rate of precession of the perihelion of Mercury.

Design- novelty is what Elie Zahar and others have called heuristic novelty : a result is novel in this sense if it “did not belong to the problem-situation which governed the construction of the hypothesis.’’1' Use-novelty is a special case of design-novelty. A prediction lacks use-novelty when that prediction is used to fix the value of a free parameter in a theory or when the result in question is simply built into the theory. (The relevance of this sense of novelty for confirmation has been emphasized by John Worrall.20)

Several different questions can be raised about novel predictions and confirmation. Here are two of the most important:

- If a theory is deliberately designed to accommodate or explain phenomena that are already known, do those phenomena confirm the theory that predicts them? (If T is designed to explain E, does E confirm T?)

- Note : Absolutely not, this falls back to the argument that fitting points with a curve once does not mean its the true curve. One set of evidence will never be sufficient to conclude anything meaningful, due to my incomplete evidence principle.

- Other things (such as quantity, diversity, and precision of evidence) being equal, does a theory derive greater confirmation from novel predictions than from the already-known phenomena that it explains or accommodates? (Is T confirmed more strongly by novel prediction D than it is by E?)

- Note : I'm inclined to believe not. Why should the existence of evidence affect the explanatory power of a theory.

The most extreme position would answer “No” emphatically to both questions, thus denying any form of inductive confirmation by evidence. This is Popperianism...Few philosophers of science regard this kind of complete rejection of inductive confirmation as plausible.

Two other extreme positions (but less extreme than Popperianism) are often referred to as explanationism and predictions:

Explanationism is the view that only the explanation of previously known and accepted results can confirm a theory and that novel predictions have no power to confirm. Thus the explanationist answers the first question with an emphatic.“Yes” and the second question with an emphatic “No.” Something close to this view has been defended by Stephen Brush and John Worrall.

Predictionism is the opposite view that only novel predictions can confirm a theory and that the explanation of previously known and accepted results has no power to confirm. Thus, predictionists respond to the first and second questions with an emphatic “No” and an emphatic “Yes,” respectively.

Between the two extremes of explanationism and predictionism are two more conciliatory positions:

One popular choice is to answer “Yes” to both questions. This position, which we shall call accomodationism , allows for the explanation or accommodation of old evidence to have some confirmatory value but insists that novel predictions must always have more.

formalism agrees that accommodating old evidence can confirm but holds that, other things being equal, novel predictions are no better and no worse than old ones...Formalism is driven by the insistence that confirmation depends solely on the formal, inductive, logical relation between propositions, regardless of when and by whom those propositions were known or to what purposes they were put.

- Note : Essentially, formalism is a scapegoat, avoiding any direct answer.