Jonas - Abductivism As A New Epistemology
Jonas - Abductivism As A New Epistemology
Jonas - Abductivism As A New Epistemology
https://doi.org/10.1007/s10670-022-00634-7
ORIGINAL RESEARCH
Abstract
Logical abductivism features in the recent literature as a new epistemology for logic,
in connection with logical anti-exceptionalism. According to this account, (i) logical
knowledge is obtained by the justified choice of a logical theory using abduction,
which (ii) replaces problematic approaches in traditional logical epistemology. We
argue that such claims are not properly warranted; they conflate justification and the-
ory choice methods. Abduction requires that one starts with justified data, and justi-
fication, in such cases, needs to appeal to sources that are not granted by abduction.
In this sense, abduction cannot be all there is to the epistemology of logic. Some
abductivists, we shall see, even appeal to methods of the traditional epistemology
to obtain the data. Keeping justification and theory choice separated may contribute
to avoid obscurities in discussion of the nature of data in logic and in understanding
more clearly how to frame the anti-exceptional nature of logic.
1 Introduction
Logical abductivism is the view that logical theories are rationally chosen by use of
the abductive method (see Priest, 2006, 2014, 2016; Hjortland, 2017, 2019; Martin,
2021. Roughly, the method prescribes that one choose the theory that scores best on
a plurality of criteria, such as accounting for the data, simplicity, explanatory power,
fruitfulness, generality, among others (which criteria to include in the computation
is a matter of dispute, of course). Along with logical abductivism, anti-exception-
alism about logic (see Hjortland, 2017, 2019)—the view that logic is not special,
not analytic, not a priori, subject to a posteriori evidence —, also recommends the
abductive method to account for theory choice, so that abductivism and anti-excep-
tionalism are frequently found connected. In a nutshell, anti-exceptionalism is a kind
of general position recommending in broad terms that logic should be more like
13
Vol.:(0123456789)
J. R. Becker Arenhart
empirical science, and as far as part of the current literature goes, it seems that the
use of abductive method is considered by most anti-exceptionalists to be precisely
the link establishing that similarity (again, see Hjortland, 2017, 2019). Certainly,
such approximation between logic and science could be framed in different terms,
but abduction is the main source of connection in current literature.
The major reason for adopting such a combination of views is related to the claim
that abductivism is able to provide us with a new epistemology for logic. In particu-
lar, it is said to allow us to overcome traditional approaches to the epistemology of
logic requiring a priori methods, such as those describing us as creatures endowed
with a kind of rational intuition giving us access to true logical propositions, or
those related to semantic methods, according to which a logical proposition is true in
virtue of the meanings of logical terminology, as it was the case for logical positiv-
ists (see Priest, 2016 and specially Martin, 2021, p. 9072, and Martin and Hjortland
(forthcoming), where logical rationalism and logical semanticism are presented and
criticized). Given that, the claim is that abduction is:
1
And it is this which closely connects abductivism with some formulations of anti-exceptionalism,
which see logic as not being a priori (see Hjortland, 2017, p. 632; 2019, pp. 250–251). For the pur-
poses of this paper, we shall be concerned with these versions of anti-exceptionalism that frame logic
as being not a priori, and which see abduction as a new epistemology for logic, substituting traditional
a priori methodology. Our purpose in the paper is to show the inadequacy of this claim, and to open the
way for future more science-friendly formulations of anti-exceptionalism that are not exclusively focused
on theory choice methods. Certainly, not every abductivist is an anti-exceptionalist, but going into the
details of such labeling would not be productive now, given that such terms are not precisely fixed in the
literature (see Hlobil, 2021; Rossberg and Shapiro, 2021). So, we focus on this specific incarnation of
anti-exceptionalism associated with abductivism, for the purposes of this paper.
13
Abductivism as a New Epistemology for Logic?
Notice how ‘knowledge’ and ‘theory choice’ are put on the same level, as if the
mechanisms for justification of belief in particular propositions and mechanisms for
theory choice were playing the same role. So, in a sense, the proposal indicates that
the abductive method for theory choice is a rival to traditional epistemology, and the
method of theory choice plays both roles of granting justification and addressing the
issue of which logical theory to choose. Ben Martin is also very explicit in connect-
ing abduction and epistemology (the ‘account of logical epistemology’ he refers to
in this quote is precisely logical abductivism, which is opposed to logical rational-
ism and semanticism):
According to this account of logical epistemology, logical propositions are not
directly justified by intuitions or definitions, but rather logical theories are jus-
tified by their ability to best accommodate relevant data. In other words, logi-
cal theories are justified by abductive means. (Martin, 2021, p. 9071)
Abduction, then, is not only a method for theory choice, but also an account of
justification of logical propositions. And we find the same theme again later, in a
clear statement of the rivalry of abductivism on the one side, and rationalism and
semantic approaches connected to analyticity on the other:
According to the theory, we come to be justified in believing a logical proposi-
tion p not by either directly intuiting its truth, or understanding its constituent
parts, but rather by being justified in believing a logical theory L containing
p. Further, we come to be justified in believing a particular logical theory L
because it better accommodates the relevant data, and possesses more relevant
theoretical virtues, than other available theories (Martin, 2021, p. 9074)
So, let us repeat: the plan, as far as these versions of abductivism are concerned,2
is that abduction is not only a method for theory choice, but it also replaces tra-
ditional accounts of logical epistemology, as based on semantics and intuition
(the a priori methods; again, see also the discussion in Martin and Hjortland
(forthcoming)).
Our claim in this paper is that the project of advancing abduction as a new epis-
temology replacing ‘old’ epistemology faces considerable challenges, basically
because of its conflation of theory choice with other tasks of epistemology related
to the justification of particular propositions. Abduction is not a rival to traditional
approaches to the epistemology of logic, a priori or not, but at best a complement
to them. We shall argue for that claim based on the abductivists’ considerations con-
cerning the workings of the abductive method and the nature of logic. In a nutshell,
the abductivists we are focusing on here claim that the evidence a logical theory
must account for comes in two flavors: either it is semantic evidence, about mean-
ing (the ‘metalinguistic account of logical consequence’), or else it is non-semantic
evidence, in which case logic is taken to deal with general features of the world (the
‘non-metalinguistic account of logical consequence’). We claim that in each case,
2
And we shall only be concerned with these versions of abductivism here.
13
J. R. Becker Arenhart
abduction does not replace the appeal to at least some traditional approach to epis-
temology. We discuss, in particular, the account advanced by Priest (2006, 2014,
2016) when it comes to a semantic approach to logical consequence, and, more
briefly, Williamson (2017) when it comes to non-semantic approaches. One can
claim that a requirement that our theory accounts for the data is present in both an
epistemology of logic and in the process of theory choice, but still, providing for an
epistemology and providing for an account of theory choice are two distinct kinds of
tasks. This makes abductivism and the versions of anti-exceptionalism associated to
it less revolutionary, but also less confusing. Once such topics are in place, the real
epistemological problems for such views may be more properly addressed.
The paper proceeds as follows. In Sect. 2 we introduce the main ingredients
required for abductivism and the specific version of anti-exceptionalism we shall
be concerned with. Of special interest is the distinction between metalinguistic and
non-metalinguistic accounts of logical consequence. In Sect. 3, we examine Graham
Priest’s view on abductivism and his appeal to the tools of the old epistemology of
logic in order to have abduction running. We also discuss the approach and Timothy
Williamson, which is typically listed as anti-exceptionalist, and we argue that in this
case abduction does not play the expected role of a new epistemology as advertised
by abductivists (i.e., it is not a substitute for other forms of epistemology). In Sect. 4
we advance a proposal in which abductivism can live in peace with whatever one
wishes to call ‘the old epistemology’, in case one still opts for abduction. We con-
clude in Sect. 5.
As advanced by some abductivists (v.g. the quotes in the previous section), abduc-
tion involves a comparison of a family of rival theories over a set of criteria that
such logical theories should exemplify, such as adequacy to the data, simplicity,
generality, fertility, and so on. We are rationally justified in adopting the theory/the-
ories scoring best in such evaluation process. It is open to discussion whether one
single theory will have to score best, or, perhaps, whether two or more theories will
be equally rationally justified due to equal scoring. This point is not important for us
now (the method is neutral on the ‘monism versus pluralism’ issue).
Now, although the abductive process involves considering how theories fare
according to such criteria, adequacy to the evidence or data is certainly the most
important one; a theory must account for the data if it is to be a legitimate candidate
for rational choice (Hjortland, 2019, p. 257). What are the data for logic, however, is
an intricate question, depending very much on what one takes logical theories to be
about. Given the ambitions of using abduction as a substitute to traditional accounts
of epistemology, the following questions are of utmost importance:
13
Abductivism as a New Epistemology for Logic?
As we shall see, answers to these two questions connect in such a way that
the attempted substitution of traditional epistemology by abduction is not really
achieved. In short, the question for what counts as evidence for a logical theory
depends on what one takes a logical theory to be about (and what one takes the aims
of a logical theory to be; see the critical discussion in Hlobil (2021) see also Aren-
hart and Molick (2020). For the abductivists we are concerned with here, logical the-
ories are about logical consequence. Now, facts about what follows from what come
in two flavors: it may be related to the behavior of language or else with general
features about the world. Evidence for the adequacy of a logical theory, as a result,
comes from either linguistic facts or from facts about the world. We shall argue that
the bridges between us and the behavior of language or the world (depending on
how one answers question (i)), alas, is mediated by none other than some of the
traditional tools of the ‘old’ epistemology (i.e., it cannot be performed by abduction
alone). So, in a sense, abduction cannot be a new epistemology; it may at best work
in conjunction with some form of the old epistemology. This is a claim we shall
substantiate in the remainder of this paper, of course. To get things clearer, we now
check more carefully how answering questions (i) and (ii) leads to it.
Let us begin with the first question. Clearly, for the versions of abductivism
being discussed here, logical theories are about what follows from what, i.e., logi-
cal consequence. Here we follow Priest in assuming that advancing a logical theory
requires advancing answers to many different questions:
The central notion of logic is validity, and its behaviour is the main concern
of logical theories. Giving an account of validity requires giving accounts of
other notions, such as negation and conditionals. Moreover, a decent logical
theory is no mere laundry list of which inferences are valid/invalid, but also
provides an explanation of these facts. An explanation is liable to bring in
other concepts, such as truth and meaning. A fully-fledged logical theory is
therefore an ambitious project. (Priest, 2016, p. 39)
Furthermore, the facts about validity are the target of such theories, not to be
identified with the different notions of validity according to a theory; so, we are
talking about what is typically called ‘genuine validity’, or validity ‘tout court’, not
‘validity according to system S’. Concerning this distinction, there is an ambiguity
between logic as a theory on the one hand, and what it is a theory about on the other.
As Priest explains the point:
‘Logic’ is ambiguous. It can mean both the theory of an investigation and the
subject of the investigation. In the same way, the word ‘dynamics’ is ambigu-
ous. It can mean a theory, as in ‘Newtonian dynamics’, and it can mean the
way that a body actually moves, as in ‘the dynamics of the Earth’. It is logic
in the first of these senses that I am talking about in this essay. Theories come
and theories go, and a dominant theory can be replaced by another. Logic, in
this sense can clearly change. Logic in the latter sense is a different matter. It
is constituted by the norms of correct reasoning, that is, the norms of what fol-
lows from what, and it is the theorising of these that logic in the first sense is
aimed at. (Priest, 2016, p. 40, see also Priest, 2014, p. 216)
13
J. R. Becker Arenhart
Hjortland (2019, pp. 252–253) and Blake-Turner and Russell (2021) make a
very similar point. The plan is that a logical theory is an attempt at successfully
describing facts about logical validity, just as theories about dynamics attempt to
successfully describe moving bodies. The facts about moving bodies do not change
as we change our theories about them. So, in this analogy, assuming that a logical
theory is about logical consequence and validity, what, more exactly, are the facts
such theories are willing to describe? Well, it all depends on how one takes logical
theories to be applied to their target, constituted by inferences in natural language.
Here, an important descriptivist component enters the stage, given that even in logic
it is assumed that there is something out there to be described by our theories.3
The descriptivist component was implicit in the distinction between two senses of
‘logic’, but we are able to make a more precise discussion of what logical theories
are about if we bring it to the center of the stage. Clearly, it is being said that logi-
cal theories are applied to describe correct inference in natural language; but logi-
cal consequence can take two different roads from there, depending on whether one
sees it as concerning meaning or as concerning general features about the world. As
Hjortland puts it:
[...] logical theories, in the present sense, are not normative. Rather, they
describe logical properties by, for example, attributing validity to arguments,
or properties such as truth-preservation and variable-sharing to validity.
Exactly what is being described—language or the world—will depend on the
theory in question. (2019, p. 255)
Basically, what Hjortland advances is the claim that the facts about what follows
from what are facts about the workings of language, in what he calls a ‘metalin-
guistic approach’, or concerning general features about the world, in a ‘non-metalin-
guistic approach’. Examples of metalinguistic approaches are Carnap’s well-known
views on logical consequence, and Shapiro’s ‘logic as modeling’ proposal in (2014).
The general plan is that what follows from what is determined once we have the
meaning of logical vocabulary, along with some further properties of logical con-
sequence. For instance, one may attempt to argue that proposition B follows from
propositions ‘A or B’ and ‘not-A’, given what ‘or’ and ‘not’ mean.4 Examples of
non-metalinguistic approaches are advanced by Maddy (2002, 2014) and William-
son (2017), who, in different ways, consider logical theories as describing general
features of the world. As an example, one may argue that given that no object o can
both instantiate and fail to instantiate a property P, then, the statement ‘it is not the
case that o has P and o does not have P’ must hold true for any such object and prop-
erty.5 In sum, then, logical theories are theories about validity, but the facts about
3
That logic is descriptive is an assumption that is problematic in the wider context of the philosophy of
logic. However, given that it is more or less a standard assumption in the context of the particular debate
we are concerned with, we shall assume it for the sake of argument.
4
Or, alternatively, of course, deny that B follows from such premises, given what one deems to be
appropriate meanings to such logical vocabulary.
5
Again, this is a controversial example (see also Tahko, 2009 for a formulation of this specific case). It
serves only to illustrate how the non-metalinguistic approach would proceed, in its first steps, to justify
logical validity in terms of features of the world.
13
Abductivism as a New Epistemology for Logic?
validity may be facts about language and meaning (metalinguistic approach) or facts
about the general structure of reality (non-metalinguistic approach).
Given these two very broad classes of understanding what logical consequence
is about, two different accounts of data will arise, addressing then the second ques-
tion we presented before. The point is that if logic is about language, then, semantic
considerations will have to be considered, and we shall have to check how mean-
ing engenders validity. On the other hand, if logic is about the world, other kinds
of considerations will enter the stage, relating logical consequence with facts about
the general structure of reality. The point we shall argue next is that such consid-
erations require epistemological resources that are not provided by abductivism, so
that abduction cannot be considered a replacement for the more traditional tools of
epistemology. In other words, abductivism assumes that there is some way one can
independently access the data given in terms of language or reality; only after that
logical theory choice can be performed.
Before we move on, a remark is in order.6 Clearly, to stipulate that logic is about
validity and logical consequence does seem to settle by fiat, as it were, a very contro-
versial question about the subject matter of logic. One could complain that by doing
so other approaches to logic get excluded which could evade the kinds of difficulties
we are discussing here, such as (a) versions of psychologism, seeing logic as related
to the way people think or ought to think (descriptive and normative psychologism,
respectively; see Haack, 1978, p. 238), or Platonic approaches, such as the one typi-
cally associated with Frege, which sees logic as describing an objective and inde-
pendent third realm. Certainly these would be interesting additions for a discussion
on a more fundamental controversy on the subject matter of logic. However, allow-
ing for such views on the nature of logical consequence would be problematic in the
current context of this paper, not only for a matter of scope. For a start, descriptive
versions of psychologism would certainly corroborate our major claim that use of
abductive method cannot avoid using traditional approaches to epistemology (such
as induction, for traditional psychologists); discussion of logic as being about a third
realm would bring in the kind of epistemological challenges concerning the access
to such a realm (typically achieved by intuition), which is precisely what abductiv-
ists claim to avoid. That is, introducing these option would bring the discussion to
an even more fundamental issue, one where in some cases the abductivist claim fails
(as in the case of psychologism), or one where it is openly making use of the episte-
mological methods it is thought to avoid (in the case of Platonism). Given our focus
on assessing the merits of some abductivists’ claims on their own grounds, we con-
centrate on their preferred accounts of logical consequence for the sake of definite-
ness of this paper.
6
We thank two anonymous referees for pressing on this point.
13
J. R. Becker Arenhart
Once we join abductivists and determine what logic is about, and as a result,
make up our minds as to whether logical consequence is metalinguistic or non-
metalinguistic, there are some hints in the corresponding literature about how to
look for evidence on the proper behavior of such consequence and, correspond-
ingly, on the appropriate theoretical description of it; our claim is that it is pre-
cisely here that one may find one or another approach to traditional epistemol-
ogy playing a role, sometimes a priori, sometimes not. It is important to get this
point clear. ‘Traditional epistemology’ of logic, as conceived by abductivists,
denotes either an intuition-based account of logical truths and validities, or some
form of semanticism, i.e., the view that knowledge of logical truths and validi-
ties are determined by knowledge of the meaning of the semantic constituents of
propositions. Both, of course, are typically considered to work on a completely
a priori basis. However, for our purposes, there is also a posteriori accounts of
justification which is not identified with abduction and which is not, thus, to be
included in the abductivists’ solution to the problem of the epistemology of logic.
In this paper, ‘traditional epistemology’ covers epistemology that is not of the
abductive kind (i.e. associated with the claim that justification comes with theory
choice). It is perhaps a side effect of current association of anti-exceptionalism
and abductivism that use of the abductive method has been directly related to an
a posteriori epistemology, in direct opposition to a priori epistemologies (again,
see Hjortland, 2017 and Martin and Hjortland (forthcoming)). This confuses the
discussion, by not allowing any place in the dispute for a posteriori epistemol-
ogy that is not abductive (for instance, as it features in traditional psychologist
approaches to logic). Our discussion shall point to this issue soon.
Given this separation between abduction as a source of justification and knowl-
edge on the one side, and more traditional epistemology on the other, we hold
that contrarily to what has been claimed by abductivists, the abductive method
will play no important role until the former has played a role. That is, given a
choice on whether logical consequence is to be understood as metalinguistic or
non-metalinguistic, different kinds of evidence or data will be required, and they
will be accounted for, typically, in terms of some account of the old epistemol-
ogy, which was claimed by abductivists to be replaced by use of abduction. What
is central to our point is that some form of connection with the evidence will have
to be provided, a connection not established by abductivism itself, but rather by
some account of the old epistemology, a priori or otherwise, depending on the
case. Anyway, the claim is that the working of abductivism requires something
like the old epistemology playing a role beforehand. We shall now bring to light
how that interplay introduces itself.
13
Abductivism as a New Epistemology for Logic?
13
J. R. Becker Arenhart
pure logic is the correct semantics for vernacular negation. Seen in this way,
a dispute between rival logics is, then, exactly a dispute over meanings. This
may surface, for example, as a dispute over the truth conditions of vernacular
negation. It is therefore entirely open to someone who holds that our theory of
logic is revisable to hold that the correct logical principles are analytic, that is,
true solely in virtue of the meanings of the logical connectives employed; it is
just that which principles are analytic is a corrigible and theory-laden issue.
(Priest, 2006, p. 171)
So, semanticism is definitively not out of the game (and semanticism, recall, is
one of the old epistemologies, as spotted by Martin (2021) and Martin and Hjortland
(forthcoming)). Given that our access to the meaning of connectives is mediated by
our theories, we appeal to them in order to establish, fallibly, what is valid. Still,
however, this is a form of semanticism.
But one could complain here: it could be argued that while it is true that a given
logical theory determines the validities through meaning, there is still an issue about
which such theory concerning the meaning of logical particles to choose, and then,
validity granted by meaning per se does not provide justification. There is still the
need of justifying the adoption of a given specific theory endowing the connectives
with a given meaning.
What the objection seems to emphasize is the fallibility of our theoretical choices.
While it is true that we can determine some validities of a theory after the theory is
chosen, we are always open for theoretical revision, given that we may be mistaken
in choosing the theory we have actually adopted. But this does no harm to our point:
still, the method for determining validities, once a theory is adopted is provided by
semanticism. That is, there is a place, and not a minor one, to be occupied by the old
epistemology. The role of abduction is much more restricted than abductivists would
recognize. We are not permanently choosing theories. Under the assumption that we
do use a logic (and Priest and abductivists grant that), then, the logic we have (to
use Priest’s expression) gives us the meanings of the connectives for the time being.
From that, we are justified in deriving inferences as valid in terms of meaning. So,
semanticism operates freely when a system is in charge to fix meanings for logical
operators. Of course, we may be wrong in believing that the logic we have is the best
one available, but that is another issue. Assuming that theory choice is fallible, the
possibility that our current logic is not the best is always open. The point, to empha-
size again, is that semantic considerations do guide our epistemology of logic, once
a theory is in charge, and that, according to Priest, relies on the semanticism.
Furthermore, semanticism also plays an important role in providing the evidence
before we choose a theory; we choose a theory that accounts for some validities we
consider as initial data to be explained. That is connected with the introduction of
one further ingredient in Priest’s picture, which also brings it closer to the traditional
epistemology front: the appeal to intuitions. According to Priest, we recognize that
some inferences are valid, and that others are not, using our direct intuition of their
validity/invalidity (Priest, 2016, p. 41); some inferences strike us as valid, some
strike us as invalid. This furnishes us the data to be accounted for in logical theory
choice, even if it is recognized that such intuitions may be mistaken.
13
Abductivism as a New Epistemology for Logic?
Notice that there is an ambiguity here (not only here, but in the whole discussion
concerning semanticism and rationalism in the abductivists discussion). Intuition
seems to play two roles: (a) the role of data, as the inferences we immediately recog-
nize as valid (as in Priest, 2014, 2016; Hjortland, 2019) and (b) the role of a faculty
or capacity that enables one to recognize some immediate truths (as indicated in
Martin, 2021). It is not completely clear to us how Priest himself is understanding
the role of intuition, perhaps as a kind of mixture of both notions. What he seems
to have in mind is that through intuition we have access to some logical validities;
and given this immediate recognition of the validity of some inferences, we consider
them, even if fallibly, to be the data a theory has to accommodate, granting them a
kind epistemic privileged status. Furthermore, the data of logic (on a metalinguistic
account, of course) are given by meaning. Given the meaning of the connectives,
we perceive, as it were, intuitively, that some inferences are valid. Given our under-
standing of the meaning of some logical components, such as negation and the con-
ditional, some inferences, such as modus ponens, or double negation elimination,
may strike us as valid, and that prima facie validity is something to be accounted
by our theories (the adequacy to the data requirement). This kind of argument is
recurring in Priest’s defense of dialetheism, the view that there are some true con-
tradictions: due to some intuitive facts about how language behaves, negation should
mean something different than classical logic allows for; some sentences must be
true, while also having true negations. Priest et al. (2018) Sect. 3.2, put it as follows:
Overall, such paradoxes as the Liar provide some evidence for the dialethe-
ist’s claim that some contradictions are provably true, in the sense that they are
entailed by plain facts concerning natural language and our thought processes.
Extended Liar paradoxes like ‘This sentence is not true’ are spelt in ordinary
English. Their paradoxical characteristics, dialetheists argue, are due exactly
to the intuitive features of ordinary language: unavoidable self-reference; the
failure of metalinguistic hierarchies, which only produce languages that are
expressively weaker than English; and the obvious presence of a truth pred-
icate for English, ‘is true’, which is characterized, at least extensionally, by
either the Tarskian T-schema or rules amounting to the transparency of truth.
So, our access to the evidence is provided by the traditional means, through an
intuitive recognition of the correct semantic behavior of logical terminology, pre-
cisely those that were thought to be avoided by adoption of abductivism! If we are
justified in believing a proposition precisely when we have evidence for it, then, at
least some part of the justification needs to be performed by the old epistemological
means, those that are in part a priori, and that are deemed too mysterious and prob-
lematic by the abductivists (Martin and Priest included). Rather than substituting the
old methods by abduction, there is a place for them when we work inside a theory
(the adopted theory) and also for determining the evidence that our theory (or any
candidate for the correct theory) must account for (i.e. in establishing the data).
This may be put another way: either the use of intuition or validity character-
ized by meaning by some abductivists is justified, or else it is not. In the first case,
it seems that the attack on rationalists and semanticists is not fully justified, and the
urge for a new epistemology for logic cannot be based on the failure and inadequacy
13
J. R. Becker Arenhart
of the traditional methods (given that now they are allowed). In case such use is not
justified, then, of course, the abductive method as a new epistemology must be refor-
mulated, so that it is not based on such uses of intuition and/or semantical methods.
In any case, the metalinguistic approach to logical consequence, as connected with
abductivism, is still not properly formulated.
As we shall discuss later, what needs to be distinguished here are two different
stages of theoretical work: investigation inside a theory, where semanticism applies,
and investigation of situations involving theory change, concerning the dynamics of
science, where abduction may be used, but still relying on intuition to acknowledge
the data (as in Priest’s account). Abduction is not playing the role of a new episte-
mology, but elaborating on the available workings of a priori epistemologies (in this
case, at least).
As we have seen, one can appeal to abduction, and still, one will need some kind of
appeal to both semanticism and rationalism, in case one is adopting a metalinguistic
approach to logical consequence (needless to say, one can be a semanticist and not
adopt abductivism). Use of the new epistemology does not avoid the traditional epis-
temology, and, as we have suggested (to be further discussed in the next section), the
idea that abduction could be a new epistemology is based on a confusion between
investigation of a theory and investigation of scenarios where theory change is
involved. But what happens when one adopts abductivism and non-metalinguistic
approaches to logical consequence? Couldn’t it be the case that when one adopts this
option, then, abductivism could play the proper role of a new epistemology substi-
tuting traditional approaches? It seems that even here there is no revolution of the
kind expected, but rather, reliance on some form of traditional epistemology. Let us
briefly check.
The most well-known approach to non-metalinguistic evidence comes from what
Hjortland calls the Quine-Williamson account (see Williamson, 2017; Hjortland,
2019, Sect. 5). In a nutshell, the account suggests that one should test different logi-
cal consequences in terms of a set Γ of principles of some independently well-estab-
lished theory (a physical theory, a theory of truth, of mathematics, and so on). The
plan is that by applying the competing logical systems L 1 and L
2 to derive the set of
consequences Cn1(Γ) with one of the systems, and obtaining Cn2(Γ) with the second
(assuming only two systems, for the sake of simplicity), these sets may be compared
on how well they fare when it comes to check their coherence with empirical facts
that are independently verifiable:
The sentences in Γ are not specifically about validity or other logical prop-
erties. They are run of the mill claims about the world. When we compare
the two output theories, then, we are comparing them with respect to evidence
about non-metalinguistic states of affairs. The result is indirect confirmation
or disconfirmation of logical theories, without logic-specific evidence. (Hjort-
land, 2019, p. 264)
13
Abductivism as a New Epistemology for Logic?
Notice that in this case, specifically logical evidence is not used, but rather, one
applies knowledge of what is the case in other domains, such as physics. After that
is done, one logical theory may be more successful than the rival ones to account for
the data of the target physical theory, in terms of what can be properly derived from
set Γ, and the process of theory choice may begin. In other words, one may consider,
for instance, Γ as incorporating a formulation of principles of a well established
physical theory, such as non-relativistic quantum mechanics. When we use distinct
logics L1 and L2 to derive the logical consequences from such principles, we may be
able to compare, directly with well established facts about quantum theory, whether
some consequence set is better suited than the other to account for such empirical
and theoretical facts as already physically established in quantum mechanics. The
logic that suits better the facts already known in quantum mechanics is to be chosen
as the more appropriate.
Isn’t this use of abductive method? Certainly, this can be seen as an application
of the method: we are suggested to compare two logical theories according to their
capacities to deliver what is already established as physical facts, from the principles
of a physical theory (in the case of quantum mechanics example, anyway). However,
notice that in order to set the method of theory choice running, we still need the
appropriate epistemology in place for the corresponding physics. In this sense, if
we do indeed choose the system that scores best when such a kind of comparison
is provided for, theory choice is not a substitute for epistemology, but rather a step
that is provided for after the epistemology that is available beforehand for physics
has already done its work. That is, abductivism for logic relies on some epistemol-
ogy we already have, if any, for the relevant physics, the one that tells us about the
non-metalinguistic states of affairs that hold. Furthermore, different theories are to
be compared once the relevant physical facts are already independently available,
that is, when the evidence is clearly already obtained by other epistemic sources and
is equal for both logical theories. In this case, again, what results is that abductiv-
ism is not providing the complete epistemological story for logic. Certainly, in the
case of non-metalinguistic approach to consequence one may still argue that the a
priori epistemologies are being avoided, but that happens not for the reasons typi-
cally advanced by abductivists, i.e., that abduction is providing the complete episte-
mological story, but rather because we are relying on an empirical approach to the
relevant physics. One could, in some cases, even dispense with abduction and theory
choice: suppose only one system of logic (let us say, classical logic) accounts for all
the relevant physical facts that we expect to be derived. It would then result that we
are justified in choosing classical logic, without having to run an abductive selection
between rival systems.
The trouble for abductivists here may also be understood when we compare the
use of the abduction method in logic (as recommended by abductivists) and in phys-
ics (as typically understood in philosophy of science). In physics, we gather the
evidence by empirical methods or whatever method we have, and evaluate different
theories or hypotheses in terms of how well they explain such evidence. Abduction
is not offered as an alternative epistemology that could replace, for instance, obser-
vation and experimentation, as a means of justification, but has a place in theory
choice scenarios. In logic, however, one is being told that abduction is to substitute
13
J. R. Becker Arenhart
the old epistemology, the methods that provide the evidence available in the first
place. There is a clear difference in both scenarios, and it is not to be expected that
abduction could play such a role in logic. The use of traditional methods of episte-
mology, by non-metalinguistic approaches and metalinguistic approaches as well,
attests to that, despite what is advertised by abductivists in logic.
So, given that we shall need some epistemological account in the traditional fash-
ion (which may involve appeal to intuitions and/or semantics, at least in the case of
metalinguistic accounts) in getting to know what is valid, what role is reserved for
the abductive method in the approach being depicted? Well, certainly not the role
of a substitute for the old epistemology, as advertised by the abductivist approaches
we have been focusing on. Rather, we suggest, abuction may provide precisely for
theory choice when we already have a bunch of evidence in which we are justified to
believe (even if fallibly). Given our set of initially accepted inferences and validities,
those that strike us as correct in terms of the meaning of the connectives involved,
or those that are taken to be in tune with our best physics and other non-logical theo-
ries, we may proceed to choose a theory that best accounts for it, in case our current
logical theory does not do so. Also, when different theories all do account for the
same evidence, then, abductivism may provide for a method of theory choice that
takes into account other virtues, such as simplicity and explanatory power.
That abduction is not the full epistemological story should come as no surprise.
Usual scenarios where abduction is employed concern application of some search
method for the best explanation when both what needs to be explained and what
are the possible competing explanations are already available. That is, such method
is used when competing stories for the same data are available, and when we are
already justified in believing in the data by some independent means. As a textbook
formulation of abduction puts it (see Douven, 2021 for additional formulations and
additional discussion):
(ABD1) Given evidence E and candidate explanations H1, …, Hn of E, infer
the truth of that Hi which best explains E. (Douven, 2021, Sect. 2)
Similarly, Paul Boghossian frames the principle as:
(Inference to the best explanation) If S justifiably believes that p, and justifi-
ably believes that the best explanation for p is q, then S is justified in believing
q. (Boghossian, 2006, p. 68)
Clearly, both formulations indicate that one should choose the hypothesis that
best explains the data already available. That is very different from the high expec-
tations that logical abductivists have on abduction, according to which our logical
knowledge, even of the basic evidence, is supposed to be achieved after a proper
logical theory is chosen (again, see the quotes in the first section). In other words,
as abduction is typically understood, we use the data we have justified beforehand
13
Abductivism as a New Epistemology for Logic?
(by means of the old epistemology, perhaps) to ground our choice of a theory; we
do not, as per abductivism in logic, choose our theories to justify our data (as the
abductive epistemology we are focusing on would have it).
The morals for abductivism and the abductivist versions of anti-exceptionalism
we are concerned with here, then, should be the following: the abductive method
works to arbitrate disputes whenever two logicians query about which system to
choose; this, however, takes place in a setting where the data is settled, and there is
agreement that both theories do account for them, or most of them, at least. So, even
if we accept that the abductive method has a role to play, we also need to recognize
that there is a previous step that is required in providing for the evidence, and this is
mostly obtained by means of some form of the traditional epistemology. This comes
to the benefit of abductivists, despite their dislike of traditional epistemology: how
could we use a theory to account for the facts, when we do not have any facts to
guide our initial theory choice? Traditional epistemology fills this gap, and theory
choice, then, does not provide for a substitute for it, but rather, relies on some form
of it.
Certainly, one finds in the literature the hope that the abductive method will sub-
stitute the old epistemology because it is supposed to settle disputes even in cases
where there is disagreement over what are the relevant data (see also Martin, 2021;
Martin and Hjortland (forthcoming) for the claim that contributing to logical theory
choice is a way to surpass actual epistemic obstacles in logic). But that can hardly
be the case. Given the nature of evidence in logic, which is heavily theory-laden (see
Arenhart & Molick, 2020; Hlobil, 2021), the application of the abductive method
can be of little help in dealing with disputes among people defending different
systems, because the friends of different systems will be employing their respec-
tive favorite theory to account for different data. One cannot them hope to select
the theory that best account for the data, given that different parts will be consid-
ering different inferences as data.7 In these cases, again, the difference in the data
can be explained by the old epistemology; typically, different intuitions concerning
what should be valid, or different meanings attributed to the connectives (or a mix
of both), or different understandings of the behavior of the physical bodies when
certain physical theories are adopted. That is, even in such cases a version of what
is deemed old epistemology may have a much more relevant role than abductivists
contemplate.
So, let us repeat it: given this active role for at least some forms of the ‘old
epistemology’, it results that abductivism is not a substitute or a rival for the old
epistemology, and the hopes that abductivism could help us with the issue of ‘the
right logic’ are ill grounded. The role of the old epistemology is still safeguarded in
some of the proposals of abductivists we have been examining: we need to justifi-
ably believe (even if fallibly) in some validities to constitute our initial evidence.
This is achieved by traditional epistemology (intuition or semantic methods, or some
7
For instance, while a classical logician will certainly consider the law of excluded middle as a proposi-
tion that needs to be accommodated, intuitionist logicians will not, and will deem a theory allowing for
irrestricted validity of LEM as wrong.
13
J. R. Becker Arenhart
form of empiricism). In some cases, where dispute sets in, we may need to choose
a theory that accounts for such data, and this is what theory choice delivers. The
abductivists’ hope was that the latter process could be the single means for justifica-
tion of logical inferences to begin with. This was not correct, as we saw, given that
the process of theory choice already requires some epistemic mechanisms to recog-
nize validities (and this, as advanced by Priest, was performed by the two of the old
epistemology approaches).
Given that this point is of crucial importance for our claim, let us frame the same
point in different terms. What the abductivists are doing is to put at the same level
two different tasks for epistemology. Rationalism, semanticism, and other forms of
traditional epistemology all account for justification of particular inferences or for-
mulas. The abduction method account for the justification of adoption of a whole
theory. These activities appear at different stages of the investigation. One thing is
to choose one among many competing theories accounting for the same data. Quite
another thing is to use one such theory to account for the validity of inferences and
statements. The identification of the two levels, with traditional epistemology and
abduction as rivals, is a consequence of not keeping the two levels properly sepa-
rated. As we have seen in the first section, Martin (2021) already induces such a
conflation of levels, and Hjortland also goes from justification of validities to justifi-
cation of theories:
An anti-exceptionalist can—and should—accept that the content of a logical
theory is in part linguistic or conceptual. What the antiexceptionalist denies
is that the linguistic or conceptual content provides a priori access to logical
knowledge, for instance because the claims are analytic. But it doesn’t fol-
low from the content of a logical theory being linguistic or conceptual that we
come to know it a priori. A logical theory [...] must ultimately be justified by
inference to best explanation, regardless of the metalinguistic content. (Hjort-
land, 2017, p. 643)
Notice how one goes from a theory concerning the justification of inferences and
propositions in the context of a selected theory to justification of a theory, and how
the latter is said to be the appropriate step to deliver us knowledge. The point we
want to emphasize here is that one need not make such a rivalry if the appropriate
roles are attributed to distinct stages of theory evaluation. A similar confusion is
found in Priest (2016) pp. 45–47, when he compares his epistemology as given by
use of the abductive method with Quine’s web of belief. Among the many criticisms
against the web of belief, Priest advances the claim that Quine is silent about how
belief revision works in the web of belief, while according to the method he (Priest)
advances, belief revision proceeds by theory choice by abduction. But notice that
this comparison is inapropriate. The web of belief is not Quine’s account of theory
choice! According to Quine, theory choice proceeds as based on the search for the
simplest theory accommodating science. The web of belief, on the other hand, is an
account of how justification of specific propositions is distributed along the whole
web, in a coherentist account of justification. Again: it is just not an account of the-
ory choice. So, Quine has both a holist account of justification (which is certainly
non-foundationalist), and an account of theory choice mostly based on the choice of
13
Abductivism as a New Epistemology for Logic?
the simplest theory accounting for experience and science broadly understood. The
latter has clear overlaps with abduction in theory choice, but is not to be identified
with the web of belief. To argue that the web of belief is not as good an epistemol-
ogy as abductivism is to invite confusion on the distinct stages of investigation.
Theory choice may have a role in justifying belief in logical propositions in some
very minor cases, for some extraneous conclusions of a logical theory, those that
are so complicated that no one would think of examining them in terms of meaning
or intuition in the first place, or some that are at odds with our prima facie intui-
tion; our belief in them is commanded by the fact that they are authorized by an
accepted theory, but this does not automatically lead us to hold that abduction is an
epistemology. Rather, abduction still labors over some kind of epistemology avail-
able beforehand, that is, some form of the old epistemology (and recall that this term
is understood broadly in the context of this paper). In sum, despite what is adver-
tised by abductivists, some version of the old epistemology has an active role to play
in abductivism too. All things considered, it seems that abductivists have gone too
fast in thinking that the theory choice procedure is a substitute for more traditional
accounts of epistemology.
5 Conclusion
We hope that the previous discussion can be of some help to disentangle some epis-
temic notions that the abductivists, and with them, some anti-exceptionalists, have
entangled in the characterization of the view (for criticism on the vagueness of the
formulation of the view, see Rossberg and Shapiro (2021). This is only a first step
in such a task. Notions such as theory choice, epistemology of logic, the role of the
a priori and the revisability of logic are typically not properly distinguished, and
it then results difficult to frame new approaches to the epistemology of logic as a
sound proposal, or as a proposal where anti-exceptionalism could be substantially
differing from mere abductivism. Our discussion in this paper has been a first step
in separating some of the important concepts that enter into the characterization of
anti-exceptionalism and avoiding some of its traps. The view, as a result, may sound
less spectacular, but it also gets more credible.
Let us recapitulate the main line of the argument here. We have followed the
abductivist/anti-exceptionalist basic background in distinguishing the metalinguis-
tic and non-metalinguistic approaches to logical consequence. In this paper, when
it comes to discuss the metalinguistic approach, we concentrated on Priest’s meta-
linguistic account of logical consequence, and the need it engendered of both ver-
sions of semanticism and logical rationalism. The facts of logic being facts about
meaning, some logical validities are used as evidence or data precisely because we
have an intuitive access to such validities. It is supposed that meaning of the logical
connectives intuitively lead us to believe in some validites. Here we found a place
for the old epistemologies. When it comes to non-metalinguistic approaches, on the
other hand, we have briefly discussed the view associated with Williamson, who is
typically cited as anti-exceptionalist. By siding with Quine in making logic closer
to empirical sciences, Williamson also requires that one has an approach to how we
13
J. R. Becker Arenhart
come to know facts about empirical sciences, independently of abduction. That is,
although abduction is part of the process of theory choice after some consequences
of physical theories are compared, there is a previous need of epistemological
account for the physics (and, indirectly, of the underlying logic of physics). Abduc-
tion, then, in both metalinguistic and non-metalinguistic cases comes at a later stage,
when we already have the evidence available and, perhaps, more than one theory
accounting for it.
The separation of abduction in the process of theory selection (the dynamics of
theories) and epistemology (in the statics of theories) may bring new light to the
very understanding of anti-exceptionalism, as a means to attempt to enlighten the
epistemology of logic. In fact, if anti-exceptionalism is the captured by the claim
that logic is not special, that its methods are similar to the scientific method, then,
focusing on the abductive method to account for logical theory choice should not
extrapolate the role of abduction in science itself. Attributing to abduction the role
of a new epistemology leads one to develop some hopes on the epistemology of
logic that just cannot be fulfilled. Better then to recognize that some things must be
kept separated, and that a closer look at science may be in order, fulfilling the prom-
ises of the definition of anti-exceptionalism. This, of course, is something to be done
in another work.
References
Arenhart, J. R. B., & Molick, S. (2020). On the very idea of choosing a logic: the role of the background
logic. In A. Costa-Leite (Ed.), Abstract Consequence and Logics – Essays in Honor of Edélcio G. de
Souza, (pp. 301–320). Rickmansworth: College Publications.
Blake-Turner, C., & Russell, G. (2020). Logical pluralism without the normativity. Synthese, 198,
4859–4877.
Boghossian, P. (2006). Fear of knowledge. Against relativism and constructivism. Oxford University
Press.
Douven, I. (2021). Abduction. In: Edward N. Zalta (ed.) The Stanford encyclopedia of philosophy (sum-
mer 2021 edition), <https://plato.stanford.edu/archives/sum2021/entries/abduction/>
Haack, S. (1978). Philosophy of logics. Cambridge University Press.
Hjortland, O. T. (2017). Anti-exceptionalism about logic. Philosophical Studies, 174, 631–658.
Hjortland, O. T. (2019). What counts as evidence for a logical theory? Australasian Journal of Logic,
16(7), 250–282.
Hlobil, U. (2021). Limits of abductivism about logic. Philosophy and Phenomenological Research,
103(2), 320–340.https://doi.org/10.1111/phpr.12707
Maddy, P. (2002). A naturalistic look at logic. Proceedings and Addresses of the American Philosophical
Association, 76(2), 61–90.
Maddy, P. (2014). A second philosophy of logic. In P. Rush (Ed.), The metaphysics of logic. (pp. 93-108)
Cambridge University Press, 93–108.
Martin, B. (2021). Identifying logical evidence. Synthese, 198(10), 9069–9095.
Martin, B., Hjortland, O. T., Maria, Lasonen-Aarnio, & Littlejohn, C. M. (Eds.). Routledge Handbook
of the philosophy of evidence. Routledge. PhilArchive copy v1. https://philarchive.org/archive/
MAREIL-4v1
Priest, G. (2006). Doubt truth to be a liar. Oxford University Press.
Priest, G. (2014). Revising logic. In P. Rush (Ed.), The metaphysics of logic (pp. 211–223). Cambridge
University Press.
Priest, G. (2016). Logical disputes and the a priori. Princípios v, 23(40), 29–57.
13
Abductivism as a New Epistemology for Logic?
Priest, G., Berto, F., & Weber, Z. (2018). Dialetheism. In: Edward N. Zalta (ed.) The Stanford Ency-
clopedia of Philosophy (Fall 2018 Edition). https://plato.stanford.edu/archives/fall2018/entries/diale
theism/
Rossberg, M., & Shapiro, S. (2021). Logic and science: Science and logic. Synthese, 199, 6429–6454.
Shapiro, S. (2014). Varieties of logic. Oxford University Press.
Tahko, T. (2009). The law of non-contradiction as a metaphysical principle. Australasian Journal of
Logic, 7, 32–47.
Williamson, T. (2017). Semantic paradoxes and abductive methodology. In B. Armour-Garb (Ed.),
Reflections on the liar. (pp. 325-346) Oxford University Press, 325–346.
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under
a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted
manuscript version of this article is solely governed by the terms of such publishing agreement and
applicable law.
13