Peirce's theory of abduction | Arthur W. Burks

We ask: What is so distinctive about abduction w.r.t. induction and deduction?

Temporal development of the concept of abduction in Peirce's work

Early

In his early work, Peirce condidered all types of inferencing as part of the evidencing process.

He distinguishes induction from abduction as follows:

Induction is:

  • Descriptive and summarizing
  • Inference from a sample of data to the whole data

Abduction is:

  • Explanatory and theoretical
  • Inference from body of data to an explaining hypothesis

Late

In his later period Peirce widened the concept of inference to include methodological processes as well as evidencing processes: induction is the method of testing hypotheses, and abduction includes the method of discovering them.

In Peirce's later work, abduction is more clearly separated as a method of discovering new ideas and has as such a place within the overall system Peirce was trying to build.

A logic of discovery would be distinct from the evidencing process itself (cf. early period), i.e. the matter of justifying a given hypothesis.

But is discovery itself actually an inference? Can there be a logic to it? Or is simply a matter of genius and psychology?

Logic of discovery

Can there be a logic of discovery, or is it a matter of genius and thus of psychology? Or is abduction merely the justification of what was discovered?

Burks distinguishes two different routes to tackle this problem.

An "external" criticism would be that the discovery of an inference is not itself an inference. (302)

An "internal" criticism asks whether Peirce can in fact justify calling the discovery of a hypothesis an inference based on his own account of logic and inference?

The paper takes the second approach.

Logic vs psychology

In his later works Peirce distinguishes psychology from logic.

  • Psychology describes and studies how people think
  • Logic studies how people ought to think

Initially this distinction is not so clear in Burks text, because Peirce conceives "of logic as a study of habits of inquiry" (303)

It thus sometimes seems as if Peirce is talking about psychology after all even though he uses the word "logic". p. 302 contains examples of this ambiguity, when he nevertheless refers to abduction in terms of instinct.

Burk's explanation is that Peirce distinguishes between logica utens and logica docens.

The logica utens of a reasoner is the undeveloped theory of logic that he uses as a basis of his reasoning. After careful and systematic study of the process of reasnoing one arrives at a logica docens, which is an improved and scientific theory of logic. (303)

And thus:

man's natural insight into truth is a logica utens which can be developed by abductive logic. It follows that it is the task of logic to bring into critical consciousness one's instinctive habits. (303)

Belief-habits

Peirce has an account of what he calls "belief-habits", which amounts to a sort of implicit logic, a logica utens

These are the reasoning habits of people that are "conscious habits of action". A "belief" is exactly this in Peirce's pragmatic account.

People are in doubt when these belief-habits are interrupted. When they are, inquiry tries to resolve this interruption.

There are three levels of inquiry:

  1. Abduction invents or proposes hypotheses that account for the facts to be explained
  2. Deduction explicates hypotheses; it deduces consequences that may be subjected to tests.
  3. Induction tests hypotheses.

Logic is the (explicit) study of habits of inquiry.

Note that abduction is the only type of (logical) inquiry that can introduce a new idea.

The question now is: is abduction then really a type of inference?

Is abduction inference?

That abudctive discovery is itself regarded as a type of inference seems to be connected to the idea of logic as normative reasoning.

The distinctive feature for Peirce is that reasoning is conscious and deliberate.

A man is not reasoning until he deliberately approves and consciously applies this habits of thought. (304)

And:

Hence, on Peirce's view, logic is the study of how one cognition ought to determine another cognition. (304)

What Peirce called "habit-beliefs" are thus not an inference because they are implicit and not deliberate (habitual instead).

Burks explanation why abduction is inference for Peirce, is that it is inference insofar as it is the conscious determination of one cognition (hypothesis) from another (data of the problem at hand). (304)

Neither are "perceptual inferences" inferences in Peirce's book.

This strong emphasis is nowadays more uncommon.

Is computation inference?

This definition of inference implies that what computers do cannot be called inference!

In the first place, the procedure of a machine does not have the element of conscious approval and control; it cannot certify the validity of its own inferences. Secondly, a computing machine lacks originality. (305)

A machine can solve only the kind of problems for which it has been constructed; it cannot formulate these problems nor the principles embodied in it by means of which it solves them. Thus according to Peirce's normative definition of inference a computing machine fails to infer in two respects, conscious approval and originality, both of which are essential to abductive reasoning. (305)

And Burks also remarks that a machine cannot verify the correctness of its own inferences.

This is really interesting in the context of AI nowadays and you could argue that this is not really that clear anymore.

It also has a weird consequence for the "logic of discovery": this logic is in fact not really bothered by any logical rules! Because if it were just the application of rules, it would just be the application of existing knowledge.

Peirce's pragmatism and abduction

He refers to pragmatism as the logic of abduction and means by this that it is a logical doctrine to be used in determining the admissibility of hypotheses. For pragmatism is a criterion of meaning-it states that an hypothesis has meaning if and only if it has empirical or practical consequences, and that the sum total of those consequences is its total meaning. (306)

Due to this connection with pragmatism, Peirce's logic of abduction is committed to only "discovering" hypotheses such that they afford an explanation of the facts; where "explanation" means that the facts can be deduced by it:

An hypothesis must explain or account for the facts in the sense that they can be deduced from it. (306)

Is this an early version of the hypothetico-deductive view in philosophy of science?

To conclude:

Thus pragmatism "is nothing else than . . . the logic of abduction". That is, pragmatism proposes a certain maxim which, if sound, must render needless any further rule as to the admissibility of hypotheses to rank as hypotheses, that is to say, as explanations of phenomena held as hopeful suggestions; and, furthermore, this is all that the maxim of pragmatism really pretends to do, at least so far as it is confined to logic, and is not understood as a proposition in psychology" (5.196) (306)