This paper deals with Charles Sanders Peirce's conception of “uberty” of an inference to a plausible hypothesis, i.e. its potential to lead to an original idea [30]. Still with other designations that imply the introduction of novelties, Peirce included this trait as the characteristic of abduction, the only type of inference to present it. But in his last writings, he distinguished two tasks that coexisted in abduction: (i) the formation of hypotheses and (ii) the selection or adoption of them, where the first task, the source of new ideas, was only admitted to be by Peirce, the product of instinct; while the second task was specifically inferential. This leads to the dilemma of trying to reconcile reasoning with instinctive insights. However, in contrast with this author, we will assume that the creative formation of hypotheses is much more sophisticated than mere instinct, subsuming all this complex task to what we have called “transduction”, a variant of Peirce's abduction, dominated by a cluster of expert-agential cognitive mechanisms and non-deductive activities based on similarity. Transduction, a modelling generation reasoning process, is achieved by means of three mechanisms stemming partly from abduction, partly from induction: (1) evocation of the idea by similarity, (2) analogical transfer, a mix of both processes, following here Aristotle's notion of paradeigma, and (3) reasoning from effect to cause, what we would properly call the abductive explanation. The utility of our method is twofold. On the one hand, it extends the source of hypothesis generation to situations not only instinctive but also allows its combination with inferences. On the other hand, our method provides a finer analysis of the cognitive mechanisms that it is supposed to be present in all creative processes and that the abductive Peircean mode of description seems to contemplate although not explicit. The method is tested on several examples.
Copyright © 2020 for this paper by its authors
Attribution 4.0 International (CC BY 4.0) the fields of epistemology and philosophy of the science and mind were added to this type of psychological reasoning, forming a set of ideas that, among other things, favors the role of reason, pushing the notion of associativity into the background.
These theories, extrapolating broadly, postulate a priority of the complex human capacities of reasoning, as opposed to associative mechanisms considered much simpler, thus constituting two different systems of reasoning: system S1, linked to the associative factors of thought, is generally characterized as an implicit, automatic, fast, innate, and evolutionarily older system, shared with other animal species. On the other hand, system S2, linked to reflexive reasoning, is characterized as a slow system, which allows hypothetical thinking, and it is the product of more recent evolutionary developments. These two systems have been linked to the dichotomy between a conscious, deliberate, and controlled reflection (S2), on the one hand, and an associative intuition, also considered by some authors as non-conscious (S1), on the other.
Notwithstanding the profound implications of this classification, which gave rise to
harsh questioning, among which this article constitutes a case to be added to such a
list ([
Compared to Peircean abduction, transduction only ensures the first inferential,
epistemic, and cognitive instances in scientific inquiry, instances of hypothesis
generation, while abduction also includes the selection of hypotheses. In the various
approaches that Peirce carried out throughout his life regarding the concept of
abduction, which was not only changing its name (hypothesis, hypothetical inference,
retroduction, presumption, among others) but of referential content, we highlight the
skillful Peircean decision to combine both instinct and inferential processes, when
describing how to produce novelty in creative acts of various types. But as Paul
Thagard [
Peirce claimed that abduction could generate new ideas, but he did not specify how this could occur. If abduction is analyzed as a logical schema, then it is utterly mysterious how any new ideas could arise. The schema might be something like: “q is puzzling, p explains q, so maybe p”. But this schema already includes the proposition p, so nothing new is generated. Hence logic-based approaches to abduction seem impotent to address what Peirce took to be a major feature of this kind of inference2. [38: 453]
In the same spirit of Thagard’s critique -which this author solves by means of a neurological subpersonal response-, in this article we opt for a non-instinctive explanation of these genetic processes, and although surely some neurocognitive aspect can be recognized here, we will limit ourselves to developing the notion of transduction, which is managed on a personal level, both conscious and nonconscious.
Such transductive activity is characterized, as a variety of one of Peirce’s conceptions of abduction, in terms of a blend between intuitive acts and abductive inferences. Only, unlike this author, (1) instead of instinct mastering intuition, we speak of expert knowledge and, (2) we distinguish two stages during abduction: a first that makes the formation of hypotheses, and a second, relative to the selection and adoption of one of such hypotheses. While Peirce considers that the formation of hypotheses (stage one) is merely instinctive, we will assume that this task -which we will call “transductive” as a particular and primitive case of the abductive process- is much more sophisticated and complex, and is dominated by a cluster of skills, activities and non-deductive abilities such as: iconic visual inferences, analogies, metaphors, diagrams, among others, all contributing to the construction of one or more hypotheses that explain the emergence of some creative insight, in response to a problem that motivates and drives the creative process. Thagard refers to a type of instinct which Peirce highlights, the instinct to guess, which, according to the Canadian philosopher, is not convincing currently, due to recent neuropsychological research:
Peirce’s suggestion that abduction requires a special instinct for guessing right is not well supported by current neuropsychological findings (…) I prefer the suggestion of Quartz and Sejnowski that what the brain is adapted for is adaptability, through powerful learning mechanisms that humans can apply in many contexts. One of these learning mechanisms is abductive inference, which leads people to respond to surprising observations with a search for hypotheses that can explain them. Like all cognitive processes, this search must be constrained by contextual factors3 such as triggering conditions that cut down the number of new conceptual combinations that are performed. [38: 458] Although Thagard acknowledges that such instinctive activity does not seem to reveal the prevailing mechanisms in the processes of creative change, and attributes this task only to subpersonal elements, the author considers as “merely contextual” that the intuitive quick decisions of non-instinctive type usually have to do with the 2 The italics are ours. 3 The italics are ours. deep knowledge and familiarization that the innovative agent has in relation to her problem solving. In our proposal, the creative agent’s expertise which allows for the shortening of paths and making creative shortcuts, is not something accidental but it is the key element in cases where a thorough and deductive reasoner -such as a computer programmed to solve specific problems- would lack the competence to act, at least, to date, not yet in a completely generalized manner, with isolated cases being successful.
The presence of expert knowledge in scientific problem solving will allow, in successful circumstances, to solve complex situations in an accelerated manner, without requiring that the obtained solution constitutes a complete and linguistically formulated representation, but rather a plausible explanation of the reason for such problematization. That is why, in this paper, we will focus on explanations based on similarities, in processes of scientific problem solving. This kind of explanation will turn into patterns of transductive argumentation, the focus of our discussion. 2
A common theme in current philosophy of science is the analysis of the
relationship between constituted theories and the models used to obtain adequate
representations of such theories. This model-theoretical link, widely discussed, has
followers who accept the notion of similarity as the bridge established to settle the
conceptual distance between these two extremes, theories, and models. We can
especially mention Ronald Giere [
Such representational vehicles are not only limited to models [
Taking into account that we will not necessarily adhere to already systematized theories but to hypothetical applications seeking consolidation, this article is oriented toward discovery processes and non-standard representational vehicles, as in the case of icons (images, diagrams and metaphors) that Charles Sanders Peirce characterized in his semiotic writings. Within this thematic framework, we will concentrate on certain patterns of argumentation that we will describe under the label of “transductions”. We attempt to put forth a characterization of such transductive arguments having the proposition of an explanatory perspective of the role of the means of theoretical characterization as our main goal, relegating representative functions to the background.
The notion of resemblance (in art) or similarity (philosophy and linguistics) has played a central role in philosophical discussions on the representative potential of a theory or a model. Just as there were important supporters of the acceptance of similarity as a basic characteristic of various representational theories of science, there have also been, and still are, a large number of detractors. In this paper, we posit that the dominant critique of the notion of similarity in the characterization of scientific theories or models is based on the emphasis placed on the notion of representation. We consider that, if, instead, the accent is placed on the notion of explanation, the risk of applying similarities decreases and thus avoiding the elimination of a source that we consider to be constitutive of the processes of formation of ideas, concepts, arguments and/or scientific theories.
Nevertheless, it is also necessary to explain to what notion of explanation we refer.
In this sense, in the text Lessons on pragmatism [
The surprising fact, C, is observed.
But if A were true, C would be a matter of course.
Hence, there is reason to suspect that A is true. [30: CP 5.189]4
Here, hypothesis A would be explained from its unintended relationship with the fact C: “it is the idea of putting together what we had never before dreamed of putting together which flashes the new suggestion before our contemplation” [30: CP 5.181]. Such hypothesis A is the conclusion that explains facts or evidence (C) mentioned in the premises, where its assumption explains the “surprise” [30: CP 5.189] or the “curious circumstance” [30: CP 2.624]: “Hypothesis is where we find some very curious circumstance, which would be explained by the supposition that it was a case of a certain general rule, and thereupon adopt that supposition” [30: CP 2.624].
Since abduction is used as an explanation of the causes that lead to certain effects,
it should be clarified that this results in a regressive progression or a retrojection,
while, instead, deduction according to Peirce, is a progressive projection, which leads
to represent a prediction of causes to effects. The latter seems to contradict the
standard inherited conception of explanation in the philosophy of science, which is
usually understood as deductive: given a theory T, if we indicate its causes as C and
its effects as the set of statements E, the latter requires of an explanation from TL,
with L the logical consequences. But as Wesley Salmon [
We are using the standard practice of referring to Peirce’s Collected Papers by ‘CP’, followed by the corresponding volume and paragraph number. Peirce, who took, in his last writings, deduction as the second step in scientific inquiry, after the application of abduction, before final inductive testing. Just as Peirce distinguishes abduction -as the first step in the scientific process- from induction -as its last step-, transduction also differentiates them. Indeed, although both types of reasoning are ampliative, they are so for different reasons. Induction consists in the generalization of a particular case (problem) A, to a given class Ω, of members with solutions on the set CΩ. Thus, individuals belonging to a class already known are amplified. In contrast, in transduction, a new property is conjecturally created for a particular case (problem) A, creatively inferred from the capture of a similarity of A with another problem B. Therefore, the properties predicable by A are extended, as for example a solution CA to problem A, from an analogy with the known solution CB of B. Furthermore, in the case of transduction, it could eventually also be concluded that a class Ω is created such that A, B ε Ω; and also that CΩ is created, a kind of problem solving methodology or heuristic of Ω members, such that CA, CB ε CΩ. Thus, induction generalizes, while transduction conjectures.
Given the mentioned separation between deduction and abduction, it can be clarified, as Thagard says “what constitutes the causal relation between what is explained and what gets explained” [38: 449], but, in contrast with his neuronal strategy, we posit as constituents of such an explanatory relationship, to a resemblance, obtained due to latent expert knowledge in the detector agent thereof.
Therefore, the proposal of the notion of transductive argumentation developed herein, seeks to question the importance given to the representational role of the models, contributing with the notion of transductive icons as non-standard representational vehicles oriented toward presenting explanations of scientific practices, instead of constituting types of modelistic idealizations and/or abstractions, which often lack an effective description of the processes of creative change.
Semantic approaches, in general, propose to account for the relationship between theory and model in terms of isomorphisms, thus connecting, on the one hand, a mathematical space of representation, with a model of physical system, on the other. One of its supporters, Ronald Giere, proposes, on the other hand, to privilege the notion of similarity in detriment of the concept of isomorphism as the type of link between theory and model. This “weak” strategy [12: 81] fulfills the function of better describing scientific practices while models remain similar with certain respects that are relevant to the theory, as well as the degree of similarity with which idealizations occur. Although this notion of similarity as a substitute for isomorphism was highly criticized, it gained value in relation to the not-so-considered analysis of contextual epistemic factors, aspects that are also taken into account in our transductive proposal.
However, instead of talking about a “weak” relationship between theory and model through theoretical hypotheses obtained by similarity as linguistic devices that are not necessarily isomorphic, we focus on an also weak (or more comprehensive and general) relationship between a concrete and particular problem, and the finding of a specific and particular “transductive icon”, which is linked to the elements of the original problem in a way that is not necessarily linguistic (although it is semiotic, in the Peircean sense), through certain processes of similarity, which do not strictly require a connection of isomorphism, but a weaker one of analogy. Although it should be noted that analogies are usually understood in linguistic terms (not always though), this does not imply falling back into the same difficulties of the correspondence rules applicable in a structuralist style typical of the received view. Also, the not necessarily linguistic mode of expression should not obscure the understanding and explanation of the obtained plausible solution. Let us then make explicit the notion of transduction. 3
In a 2014 article, considering the work of Peirce, Ahti-Veikko Pietarinen and Francesco Bellucci maintain a dichotomous distinction between: (a) instinctive processes, “by which we pass directly from a proposition to another without any mediation”, and (b) reasoning, “a process by which we pass from a proposition to another mediately, that is, upon some reason”. But what makes the dichotomy much stricter consists of the following division: while “in reasoning there must be some voluntary and controlled act that can evaluate the conclusion with respect to a range of other alternative acts”, other kinds of acts, such as instinct, are involuntary ones. From there they conclude that retroduction (abduction) “is subject to logic insofar as is performed according to some5 principles governing voluntary and controlled acts”6.
Therefore, the reason versus instinct dichotomy, according to these authors, (1) lies at the base of the Peircean notion of abduction, (2) is characterized by attributing voluntary and controlled acts to reasoning while it is not possible to do so when referring to instinctive processes, (3) makes retroduction be a matter of logic as it operates as a reasoning. Consequently, the instinct that Peirce attributes to abduction to generate hypotheses, according to these authors, is an unmediated, involuntary, and uncontrolled task, and thus it is not a matter of logic.
In this way, logic would deal only with “reasoned” aspects of abduction, which would result in the generation of hypotheses (a purely instinctive task) being excluded from its area of influence. In this train of thought, abduction would only intervene in the choice and adoption of hypotheses but not in their formation, which, according to this, is an instinctive task.
From the above, we may ask: (1) are we sure that hypothesis formation can be relegated to instinctive only processes, “instinct for guessing correctly” [28, MS7 690-692]? If the answer to this question were negative, that is, if there was something else between instinct and reasoning, (2) what will that be? (3) what mechanisms would operate to produce new hypotheses in addition to and apart from Peircean’s instinct? If such mechanisms exist, (4) could they be inferential? (5) Should they necessarily be “voluntary and controlled acts” (as Peirce, Pietarinen and Bellucci
7 All references to Peirce’s Manuscripts, numbered in accordance with the R. Robin’s
Annotated Catalogue of the Papers of Charles S. Peirce [
When it comes to solving a problem, and the difficulties to find possible solutions
persist, the appeal to other similar problems allows, on occasions, to pose like
answers as a hypothetical framework, thus opening a new perspective that has been
blocked up to this moment. Clearly, the hypothesis formulated acquires a plausible
character. In principle, its inconclusiveness is waiting for a later verification. The
cognitive opening introduced can be more or less creative depending on the scope of
such resemblance. It is known that similarity reasoning is controversial, but, in
positive cases, it allows us to extend the analogy to multiple situations, where the
benefit of the risk comes from accepting momentarily, tentatively and provisionally
this type of argument. Therefore, taking these provisions into account, it is possible to
immerse oneself in the world of resemblance, taking it as a primitive, and bet on a
creative openness, beyond achieving the justifications of the case. It is possible to
observe that this type of reasoning -which goes from the assumption of a similarity
and its analogical projection to a greater set of properties in common between the two
domains in relation of comparison-, is sustained under the belief of certain principles
of permanence (or invariance) and continuity, which apply the similarity in a different
domain and extends its incidence to more properties. Another notable characteristic is
that it consists of a type of reasoning between particulars, thus escaping
generalizations in other contexts. Therefore, it does not seek to cover more cases, as
occurs with induction, nor does it pretend to seek generality of the type of a deductive
argument. In this sense, we can distinguish at least three types of argumentative
inferences that configure work methodologies: top-down (deductive reasoning that
goes from generic to generic, or applies generic to particular), bottom-up (reasoning
which infers from particular to generic), and what we called “transductive” reasoning
(which infers from particular to particular). This last term was coined by
Gammerman, Vovk, and Vapnik [
By adopting a transductive inference, an analogy is constructed, described through a function from a domain A (the problem to be solved) to an image B (another problem already solved previously). This analogy allows us to explore how the problem A might be by comparing it with problem B, and how it might work if it were like the analogical problem B. Since we know a solution of B, all we have to do is transfer it to A. Therefore, the analogy must reflect the invariant structure that A has in terms of some invariance present in problem B and similar to that of A.
Thusly, we can portray the characteristics of the starting set A in terms of some already formed and known set B with which we are familiar. In doing so, imagining some aspect or property of the domain A in terms of something else, we are able to think about the original problem from their transductive model. This allows solving the initial problem in terms of a solution that is already known in the analogous previously solved problem. The evocation of the familiar and known in advance problem B, leads to solving problem A, given the similarity between A and B. In this way, there is a connection –once unthinkable and surprising- between A and B, in making the terms from A to the analogical model B, and vice versa as the analogical terms fit back to problem A.
The advantage that problem A acquires when interacting with problem B consists in the creation of a new cognitive scheme to characterize A in terms of B, embedding A into B and redirecting A to the solution of B, which is based on its latent invariants, to finish capturing the invariants in A that allow its solution.
Transduction is a kind of reasoning whose function is to introduce original, novel and significant elements in an unfamiliar and problematic context A, resorting to other familiar elements belonging to a domain B of prior knowledge, of which we could glimpse a plausible solution of a problem within it, which was evoked when trying to solve the problematic issues in A. We will say that A is the goal or target domain, and B is the source or vehicle domain.
We use the transductive framework to explain how ideas arise, and what cognitive mechanisms are involved in inferential processes that lead to the formation of associations between structures of two intervening domains in the resolution of a problem: the source domain, from which the problem and the information on this domain stem from, and the objective domain or destination, which cooperates in the resolution. From the transductive referential framework, for the purpose of solving the problem present in the target domain, not only logical reasoning intervenes in the passage from one domain to the other, but also other epistemic aspects, such as emotions and intentional attitudes.
A transduction is not just a simple passage or transfer (or transport) of certain
elements from the source B to the goal domain A, which, by the way, does occur; but
also a process of empathic fusion between elements of both domains, creating
something new in such a process, asymmetrically directed from source B to target A.
It should be noted that the inverse fusion of A to B, if it occurs, would cause another
transductive process, which is not necessarily the same as the previous one:
comparing ‘Achilles’ to ‘lion’ is not the same (as Aristotle [
The asymmetric fusional passage from B to A creates something new in A, from the experience of making it analogous to B. For example, the trait of audacity in Achilles is highlighted, which, perhaps otherwise would not have been noticed, but when comparing Achilles to a lion, it emerges evocatively immediately, even if it could have gone unnoticed before. Eventually, new ideas emerge from the experiential formation of the association of B with A, on the basis of the familiar and prior knowledge that is possessed of B and that, therefore, is evoked almost immediately, in order to transfer it transductively to A.
The transduction between A and B does not consist of a single application between A and B that makes the transfer of elements from B to A -as it occurs in any analogical transposition-, but in a series of transfer procedures and articulated steps that make up the way of solving the problem that domain A departs from. This set of steps is characterized by the initial search (and eventual finding) of a resemblance M, which, in subsequent successive stages is creatively elucidated and, in the best-case scenario, it is taking shape to form a plausible solution of the problematic situation (of or in) A, from familiar and known elements (of or in) B.
The main reason for the importance of similarities M is that they have a “beginning
of the ball of yarn effect”, i.e. they allow us to kick things off when we had nothing
but A at hand, or we believed so at least. Similarities offer new elements when there
seemed to be no possibilities for future action, when we appeared to be facing aporia,
i.e. dead ends. And this is something that, for example, Greek mathematics used when
conceiving the notion of “analysis” as a discovery, which consisted in assuming an
existing solution to a problem to be solved, even when we do not know if there is one,
and then proceed to operate with this supposedly existing putative entity. This
situation of “analysis” also occurred -not with that name- in the cultures of Ancient
Egypt and Paleo-Babylonian mathematics, when they operated by the method we call
today “false position” [
By adopting a transductive inference, an iconic sign is evoked, an analogy and an
abduction are constructed, described through a function from the domain A (the
8 Aristotle says: “A simile is also a metaphor; for there is little difference: when the poet says
[of Achilles], ‘He rushed as a lion’, it is a simile, but ‘The lion rushed’ [with ‘lion’ referring
to a man] would be a metaphor; for since both are brave, he used a metaphor [i.e., a simile]
and spoke of Achilles as a lion” [1: 205].
problem to be solved) to the icon B (another problem already solved previously). The
analogy allows us to explore how the problem A might be by comparing it with
problem B, and how it might work if it were like the analogical problem B. Since we
know a solution to B, even though we cannot explain or make it explicit from the
beginning, all we have to do is to transfer it to A. Therefore, the analogy must reflect
the functional invariant structure that A has in relation to what caused the problem
there, in terms of some invariance present in problem B and similar to that of A. To
capture the analogy between the two domains is to put in evidence, to explain a
certain transplant function that connects them and then to backproject it from B to A
and to analyze what happens when “implanting” the found and highlighted properties
of B now in A [
Our account deals with linguistic representations, the tentative and plausible hypothesis. However, that does not always happen in the first steps of the acquisition of icon B, being B, by its nature, not necessarily linguistic. Icons, as semiotic objects, thus constitute the starting point of the creative process of solving the problematic situation A. From there on, there is a whole process in the search and refinement of solutions to A, more in line with systematic linguistic styles, typical of scientific research. But the relation of similarity obtained, in itself, does not require, in a first stage, linguistic specifications. This makes the proposed perspective to exhibit subjective nuances.
Figure 1 shows how a complete transductive process develops with all the associative steps involved, which manifests a seemingly linear sequence, i.e. a sequence which although linear in form, is not necessarily linear in act.
Before going on to describe such a sequence of associative phases, we find it necessary to comment the following: that a whole transductive argument is posed sequentially does not mean that every creative subject passes through all instances, consciously or in a non-conscious manner. One can skip stages, start directly with a metaphor (and not with an image or a diagram), or with an analogy (not with a more rudimentary icon), or even with the solution already fully delineated in the mind; or it may even happen that what one finds is all failed, and none of the steps taken pay off; or also, one can be stuck going back and forth repeatedly through previous instances. One can alter the order of the sequence depending on how it raises awareness and realize how to build such a process.
The linearity of a creative model was a characteristic sign of the model proposed
by Graham Wallas [
Figure 1 shows the five elements at play in a transductive process: the interpretant S, the target object A, the iconic source sign B, the ground M, and the consequences CA, CB, and CM from A, B, and M respectively. Let CA be the proposed hypothesis, which eventually solves the problematic situation (in or of) A. In addition, Figure 1 presents six moments that make up the stages of the transductive process. These are: [i] Evocation of B: given A and the problem to solve within it, icon B emerges. [ii] Emergence of an asymmetric similarity from B to A.
[iii] Emergence of an aspect M in B that is also owned by A, and that operates in A as it operates in B, based on the proposal of a continuous proportion: M M:B::B:A. [iv] Familiar and expertly known solution CB to B.
[v] Extractive analogy of plausible solution CA for A, from known solution CB for
[vi] Plausible explanation of such an analogy: the relationship of M with A; how M would operate in A.
The formation of the hypothesis CA is carried out by means of a transductive process of three applications, three types of associations that follow one after the other, in an orderly manner forming a commutative diagram, described in the following three sections.
As previously mentioned, from a gestalt flash, the evocative insight of icon B is produced. A subsequent immediate process consists in the search for explanation of such phenomenon, which is assumed as an inferential process of association by similarity. Why is B associated with A? What kind of association is this? In what sense is similarity discussed here? What elements make up B that cause them to emerge during the elucidation of problem (of or in) A? What causes the emergence of B?
Similarity is a relation (between a target situation A and a vehicle B) that may be used to explain the problem within A, in virtue of the fact that an agent S interprets creatively and tentatively- that the vehicle B is similar to the target A in relevant aspects and to a certain degree of similarity. Then we can use B to explain A. Michael Poznic states:
Similarity is more fundamental than the condition of inferential capacity. Similarity can be used in order to explain why representational models could have an inferential capacity in the first place. In a nutshell, only because there is a similarity9 between vehicle and target, it is justified to derive certain surrogative relation, the inferences from models to target system can hardly be justified. [34:346]
The explanation process allows to regard B as if it conforms a specific fictional
scenario which will be the vehicle of our future reasoning [
aspect M fulfills two different roles: on the one hand it indicates a facet of B that wants to be highlighted. On the other hand, it is expected to describe another facet of A. Both the interpretation of B in terms of M first, and then the interpretation of A in terms of M, are what eventually lead to the solution of problem (in or of) A from the known solution of B. Consequently, the plausibility of the emergence of M can be observed. This situation is expressed through the following typically abductive reasoning:
B s A
→ M →( B→ A) plausible M
The detection process does not trigger properties at random, but it selects a salient one with a priority given to the relation with the target situation A, based on some criteria that dominates agent S, based on its expert knowledge. The resulting selection is a shortcut for the computational explosion, of a totality of properties that B has.
On the other hand, and almost simultaneously, it is sought to understand how the appearance of the middle term M that connects B with A allows to put forward a plausible solution CA for A. The answer to this question is to establish an analogical relationship with the objective of, in the course of making visible a solution CB of the familiar and known vicarious problem B, making it possible to similarly extract a solution CA for A. This reasoning leads to the second type of association.
Next step is to connect the icon B to the target system A and translate claims from the behaviors or functions of the icon B into claims about the target. Once the asymmetric similarity from B to A is pointed out, the agent begins to interpret the target A as something like the vehicle B. This means that whereas previously she took A to have no connection with B, she now understands A in a different way, not merely as a similarity that makes B stand in for A, but the projection of all the inferences from that similarity, i.e. an analogy. If M is an aspect that emerges from the similarity of B with A, then when explaining the relationship of M with B, the key emerges and will explain the relationship of M with A, which could tentatively give the expected solution for A.
Thus, the analogy between B and A generates activation of ideas CA related to A on the basis of (i) an activation of existing and constructed connections M between A and B, and (ii) a familiar solution CB of the problem on B, similar to the situation problem on A. This analogy identifies possibly new connections between A and B, and, above all, unexpected effects CA in A. There are similar patterns of co-occurrence with some plausible effects or consequences (CB and CA), co-occurrence across some (not necessarily all) effects from B and A.
In this second associative stage, a specific iconic diagram emerges, which seeks to be closed or commutative so as to indicate an ampliative reasoning. A commutative diagram is a diagram such that all directed path in the diagram with the same start and endpoints lead to the same result. In this case, the directed paths under such conditions are: eAs: B→A→CA and aeB: B→cB→cA where s is the relation of similarity, a is the analogical relation between B and A, eB is the relation of explanation and extraction of consequences from B, and eA is the relation of explanation from A. This will lead us to state that the existence of CA is plausible, such that eAs= aeB (cf. Figure 1). If B is similar to A, then the effects or resolutive consequences of problem (in or of) B might be similar to those corresponding (in or of) A. Or, also said in terms of decisions: “like cases should be decided alike” (the legal principle of stare decisis, the doctrine of precedent, cf. [4: 246].
Here we will talk about a contiguity-based interpretation of analogy, by claiming that A and CA, on the one hand, and B and CB, on the other, are contiguous elements in their respective domains. An association by contiguity “supposes that similar ideas have been conjoined in experience until they have become associated” [30, CP 7.499]. It constitutes a type of analogical process, in which, from observed similarities, we can infer a further similarity, when a phenomenon A calls to mind another phenomenon B because A and B have been contiguous, i.e. have been experienced together.
Echoing Bartha here [4:321-322], to express an analogical argument, three steps are needed: (1) the construction of a hypothetical model (or, instead, our iconic sign B), (2) a derivation of some dependence result in that model (our CB), and (3) an analogical inference (our function a) back to the phenomenon (our A). Thus, the general form of an analogical argument10 is:
B is similar to A in certain (known) respect M.
B has some further feature CB.
Therefore, A has some feature CA similar to CB.
Bartha proposes to describe analogical arguments from what he calls “prior associations”, a vertical relationship within the source domain (our iconic sign B) that one hope to extend to the target (our A): “any acceptable analogical argument must be based upon an explicit prior association in the source domain, and there must be potential for generalizing that association to the target domain” [4: 227].
Moreover, according to Bartha, there are at least four types of analogical arguments: predictive, explanatory, functional and correlative. In our case, it is an explanatory one, where we transfer CA. In an explanatory analogy, CB explains B, and this implies that it is plausible that something similar to CB is CA, that explains the similar phenomenon A. Here the association is asymmetric, with B being prior to CB. The prior association is the causal vertical relationship B→cB, where CB explains B, and the direction of the prior association is reversed, from effects to causes: “the analogical argument is meant to provide support for the idea that similar features (…) in the target domain [A] are explained by a similar hypothesis [CA]” [4: 25]. This abductive (explanatory) analogy involves a transfer of plausible causal knowledge, which has the following scheme:
CB→aCA CM→(CB→CA)
plausibleCM
Thus, CM is the raison d'être of the formation of CA from CB, all thanks to the detection or construction of the aspect M that causes the link between B and A. 10 Cf. [4:13].
In the previous associative step, a proportional type analogical transfer was carried out, which not only described the asymmetric similarity of B towards A, but allowed us to advance in some set of consequences CA of such a relationship of similarity, by comparison with the inferred consequences CB that had been obtained for B: something will be transferred from B to A transforming the initial problem A into a new situation, overshadowing the previous properties that it possessed, and that previously failed to lead it to its resolution. The salient property M allows now to reach to some eventually relevant information to solve the problem at hand.
The third and final associative stage is characterized in metaphorical terms, where a plausible answer is finally obtained for the original problem A: having made (i) an association s by similarity, which abducts feature M as a cause of similarity, and (ii) an association a by analogy, which abducts the effects CM as their cause, now it would be possible to abduct CA from A, as the analogy plausibly produces CA. Indeed, identifying these new connections can lead to this novel hypothesis concerning a potential solution for A, in the format of a retroductive inference e−A1, an association of CA effects to causes A (cf. Figure 1). Such inference e−A1 : CA → A constitutes a metaphorical process, which associates two different elements CA and CB, which, however, have some point in contact and connection11. As many as their differences may seem, the metaphorical similarity relationship between them shortens this distance, creating an unexpected cognitive shortcut, a surprise. Of the three associations presented to describe a complex transductive argument, the latter e−A1 is where abduction is most explicitly presented as a retroduction: “abduction is the inference of a cause from its effects” [29, W1:18012]: “I have on reflection decided to give this kind of reasoning the name of retroduction to imply that it turns back and leads from the consequent of an admitted consequence, to its antecedent” [28, MS 857: 5, n.d.].
A
C A e→−A1 A plausibleCA
Once the hypothesis is put forth, in order to account for some surprising
phenomenon, a commutative diagram is obtained. When we have completed this
cycle of the commutative diagram, which begins in the problematic situation A and
ends in its eventual hypothetical solution CA, just then we are before what Peirce
called a “reasoning from surprise to inquiry (…) a reasoning from consequent [our
CA] to antecedent [our A]” [
Thusly, the cycle presented in Figure 1 is the hypothesis generating process, which is implicit in Peirce or replaced by the commitment to the presence of a conjectureforming instinct. We seek to replace such instinctive reduction by a whole process that triggers various associative mechanisms: first an association by similarity, secondly an association by contiguity and third, and finally, an association by cause and effect, which will allow the plausible hypothesis to be shaped, which would be the first abductive step of scientific inquiry according to Peirce. This would show a whole process hidden by an apparent instinct that would not require, according to Peirce, elucidation.
The above allows characterizing transductions as processes of origin of solutions to a problem, which are also carriers of the causes of such genesis. Therefore, the posed three stages of associative inferences lead to the formation of a hypothesis CA. Now we are ready to introduce Peircean’s abductive reasoning, bypassing its instinctive perspective, since, as Frankfurt says, “it is by such reasoning that we are led to adopt hypotheses for investigation (…) Hypotheses enter serious thinking14 through abduction” [9: 595]. It would seem that Frankfurt values abduction to the extent that, as he presumably understands it, in Peirce, it has a different linguistic format than what it might mean to represent the instinctive stage. On the other hand, in our approach, not everything is linguistically representable in the transductive stage, but when it is, it is the result of the indications that could emerge to consciousness from non-conscious reasoning. Why not also regard as “serious thinking” the previous transductive process that introduces new ideas? “Seriousness" does not mean to reduce everything to linguistic format, as Frankfurt seems to believe. Seriousness should mean making explicit the tacit but also the relevant to inquiry, recognizing that mental life can be elucidated sometimes and in fragments, and that every effort to do so is commendable.
The above transductive method is tested on several examples: Plato’s Meno’s first
mathematical problem [
The expert-agential similarity approach fulfills the following requirements: [i] not only does it happen that an icon B similar to the target problem A is activated by a primarily non-conscious incubation process, of which the solution is known by means of CB, suggesting an analogous tentative alternative of solution CA of A; [ii] not only is it emphasized that the interpretation of an agent S is involved, the one using imagination, creating a fictional scenario icon B which is the vehicle of further reasoning, so that it makes features of the icon B visible, that are usually overlooked, and act imagining situations according to certain implicit rules within the icon B; [iii] not only does agent S formulate plausible hypotheses to specify similarities between icon B and the target system A; [iv] not only does agent S change the interpretation of A while specifying the icon similarity to A via the dual cognitive mechanism; [v] not only is similarity the starting primitive point, the ‘beginning of the ball of yarn’, that indicates that there is no ex nihilo creation, even if it is distant in linguistic terms but not in semiotic ones; but, in addition, the process departing from the evocative choice 14 The italics are ours. of an icon B by similarity with target problem A captures what is happening when the agent S uses this icon to explain the problem in A because of the expertise that S possesses to creatively choose B, and thus initiating the transductive process (evocation of B, asymmetric similarity from B to A, proportional choice of aspects M shared between B and A, analogy that extracts a plausible solution CA for A, from a known solution CB for B), which eventually offers a plausible solution CA of A.
The consequences of what has been proposed here are that, in order to give a reliable characterization of a creative process that eventually allows us to present a conjecture or plausible hypothesis, it is necessary to accept that not everything can be stated objectively and that, in any case, when such a process is made explicit, it involves the presence of associations governed primarily by similarity, as debatable as this may be, given its high sensitivity to context. Consequently, everything is nuanced by subjective aspects, cognitive mechanisms that partially emerge to the consciousness of the resolver and that, by means of their expertise, acquire a degree of reliability despite their fallibility.
Despite having studied an important number of examples where the transductive
methodology is properly applied, more cases can be analysed to gain in-depth
understanding of other creative insights. This will lead us to include in future works,
among other things, the following central question: how to reconstruct from a formal
point of view the logical argument described in the commutative diagram, which
schematically and iconically represents the three types of associations that make up a
transduction. Carrying out this task implies, among other things, describing an
analogical transfer or paradeigma based on the assumption of resemblance as a
primitive, in terms of proportion-based reasoning [