The
Dynamics
of
Duality:
A Fresh Look
atthe
Philosophy
of
Duality
Yoshihiro
Maruyama
The
Hakubi Centre and Graduate School of
Letters, Kyoto University
Mathematical,
Physical,
and Life
Sciences
Division, University
of
Oxford
[\mathrm{T}]\mathrm{h}\mathrm{e}
development
ofphilosophy
since the Renaissance hasby
andlarge
gone fromright
to leftParticularly
inphysics,
thisdevelopment
has reached apeak
in our owntime,
inthat,
to alarge
extent,
thepossibility
ofknowledge
of the
objectivisable
statesof affairs isdenied,
and it is asserted that we mustbe content to
predict
results of observations. This isreally
the end of all theo‐ retical science inthe usual sense(Kurt
Gödel,
The moderndevelopment
of
thefoundations of
mathematics in thelight
of philosophy,
lecture neverdelivered)1
1
Unity,
Disunity,
and Pluralistic
Unity
Since the modernist
killing
of naturalphilosophy seeking
auniversalconception
of the cosmos as aunitedwhole,
oursystem
ofknowledge
hasbeenoptimised
for the sake of eachparticular domain,
and hasaccordingly
beenmassively
fragmented
and disenchanted(in
termsof Webers
theory
ofmodernity).
Today
welack aunifiedviewof theworld, living
inthe age of
disunity
surroundedby myriads
of uncertainties andcontingencies,
which invade both science andsociety
asexemplified by
different sciences ofchance,
including
quantum theory
and artificialintelligence
on the basis of statistical machinelearning,
and
by
thedistinctively
after‐modern features of risksociety
(in
terms of the theories oflate/reflexive/liquid modermity),
respectively
(this
is notnecessarily
negative
likequite
some
people
respect
diversity
morethanuniformity;
contemporary art,
forexample,
gets
alotofinspirations
from uncertainties andcontingencies
outthere innatureandsociety).
This modern and late‐modern process of disunification and contextualisation in favour of scientific and othersorts ofpluralism
rather than monism has gone hand‐in‐hand with:lAccordingto Gòdel, scepticism, materialism andpositivism standontheleft side,andmetaphysics, idealismand
theologyontheright. Gödelsphilosophyofphysicsis rarely addressed,the onlyexception beinghis workongeneral
relativity,andyetthismanuscript suggestshispositionisnevernaiverealism/platonism. Heindeedpaysprudentattention towhat he calls the leftspiritof thetime,somesortof antirealismoragnosticism,andseriouslyattemptstomake foundations of mathematics coherewith itthroughKantianepistemologyand Husserlianphenomenology. Itisanendeavourtoreconcile therightand theleft,andinthepresentarticleweshall addressexactlythesameissuewith the idea ofduality (between therightand theleft). Apartfromthat,thisquotealso sounds likeChomskyscriticismof statistical datascience,and the leftwardpropensityGödelwasanxious aboutseemstohave survived until thepresent,whetherunfortunatelyorfortunately.
\bullet the late‐modern shift of
analytic philosophy
from the Vienna circle to the Stanfordschool,
the formeradvocating
theunity
ofscience,
as seen in their series ofpub‐
lications entitled International
Encyclopedia of Unified
Science,
and the latter thedisunity
ofscience,
as seen intheirpublications
such as ThePlurality
of
Scienceby
Suppes
[52]
and TheDisunity
of
Scienceby
Galison et al.[23];
\bullet and atthe sametimewith the
post‐modern
shift of continentalphilosophy
asrepre‐sented
by
the so‐called end ofgrand
narratives thesisby
(way
notorious)
Lyotard
[35],
which was,arguably,
aglobal phenomenon
asmaybeobservedevenintheana‐lytic philosophy
of science movefrom theglobal
general
philosophy
of science tothe localphilosophy
ofspecial
sciences(from
singular
science toplural
sciences).
\bullet Note that this
change
furtherwenthand‐in‐hand with the demise offoundationalism,
the
unity
of sciences nearest kin and aparticular
kind ofgrand
narrative as had been endorsedby
quite
amajority
ofpast
philosophers longing
for the so‐calledArchimedean
vantage
point
according
toRorty
in hisPhilosophy
and the Mirrorof
Nature[49],
who alsoemphasised
thepriority
of thecontingent
overtheabsolute,
andthereby
tookastep
towardspost‐analytic philosophy
(note
further that the mirror ofnaturewasthrown awayinartas
well;
Kleesays artdoesnotreproduce
thevisible).
From anotherangle,
the rise ofanalytic philosophy
itself was, in a way, asymptom
ofdisunity
orthe end ofgrand
narratives inphilosophy
as awhole, mostly
annihilating
thetradition of
systematic
philosophy
in theAnglophone
world that had still remained even after the death of naturalphilosophy. Today
science(including
humanities)
isfundamen‐tally pluralist
in variousrespects,
and theunity
ofscience isalmostempirically
refuted asit were, not
just
innatural sciences but in other sciences aswell. To name but afew,
theatmosphere
oflogic
has shifted from the absolutelogic
of thesingle
worldtorelativelogics
ofdifferentdomains,
orfrom theRussell‐Wittgensteins logic
asrepresenting
thestructureof the worldtothe
Carnaps logic
asendorsing
theso‐calledprinciple
oftolerance,
andnowlogical pluralism
(see,
e.g.,[2])
ismostly
consideredsomething
obvious inpractice,
ifnotin
theory.
Pure mathematics has shifted from theBourbaki‐style conceptual
modernist mathematicsjuggling
withgeneric
structures to the more down‐to‐earthcomputational
post‐modernist
mathematics with nuancedperspectives
onparticular
structures as exem‐plified by,
say,quantum
groups and low‐dimensionaltopology
(which
also utilisecategory
theory
in substantialmanners).
Thepost‐modern
mathematics is similar to the 19thcentury
mathematics,
andyet
it still builds upon theconceptual
methods of modernmathematics,
whichwere neveravailable atthat time and allowed for solutionsto severallong‐standing problems
as even thegeneral public
have heard of inpopular
scienceread‐Berman
[7]
andothers, though
it isbasically
asociological
concept
andneverapplied
tomathematics and other formal sciences before. Therecent
propensity
for \mathrm{t}retrieving
real‐ ism[17]
inbothanalytic
and continentaltraditions,
asexemplified
by
structural realism[30]
andspeculative
realism[41],
wouldcount as a case of re‐enchantment inphilosophy
aswell;
note thatphilosophy
before the realist turn had beendisenchanted,
fromthings
to ideas
(the
Copernican
turn),
and to words(the
linguistic
turn),
according
toRorty
[50].
The moveinphilosophy
fromthings,
toideas,
towords,
and backtothings
may becompared
with themove inmathematics fromthings,
toinfinities,
tosymbols,
and backto
things.
This is afineranalysis
of disenchantment and re‐enchantment inphilosophy
and
mathematics;
the same, more nuancedanalysis
could be elaborated for other fields aswell.All this tells us that the notorious idea of the end of
grand
narratives was indeedgenerally
truein science as well. Yet if thepost‐modernist terminology
isinappropriate
I would instead
rely
upon theconcept
of unitism asopposed
to that ofdisunitism,
whichactually
better illustrates theplot
of thestory here,
thatis,
the transition fromunity,
todisunity,
and tosomething
doing
justice
to both. Here I would characterise unitism and disunitismby
the presence ofglobal
meaning
and absence ofit,
respectively;
remember that
disunitism,
or course, may stillkeep
some localmeanings.
The end ofgrand
narrative,
then,
maybeunderstood asthemove from unitismto disunitism aboutknowledge:
theglobal
unity
of thekingdom
ofknowledge collapses
(through
the unexit affairperhaps)
whilstglobally meaningful knowledge
(resp. Truth)
cutsdowntoseparate
pieces
ofmore localknowledge
(resp.
truths,
which are notyet
post‐truth),
and themultiply
dividedkingdom
of disunifiedknowledge
starts tothrivethereafter,
followedby
therepeated,
unending
formation ofmore and morelocalisedknowledge
communities inwhich
agents
areloosely
interlaced with each otherby
theirepistemic
family
resemblanceand in which
knowledge production
gets
moreandmorelocalised,
andonly
makessense moreandmorelocally.
Put another way, there is afundamentaldiscrepancy
between thefollowing
twoviews onthe terminusofreason/science/civilisation,
i.e.,
thefinalendpoint
orutmost limit towhich thedynamics
ofreason/knowledge/society
eventually
leadsus:\bullet the modernist view that the world
gradually
converges into oneand the sameideallimit as in the
Hegelian
and Kantian or neo‐Kantianthoughts
[20,
21,
22] (for
in‐stance because the convergence or
patching
condition ofreason/knowledge/society
is
satisfied}
to thephysicist
this meansphysics finally
leads to the so‐calledtheory
ofeverything,
yet
it isstill unclear whether ithappens
atthe end of theday
ornot);
\bullet thepost‐modernist
viewthat the worlddiverges
into diverse local limits(for
instancelimit or
global
elementpatching
local elementstogether;
to beprecise
there isyet
another case,i.e.,
there is no limitwhatsoever;
atoy
example
of this case would bethat
philosophical
debates or real‐lifequarrels usually
lead to noconverging point,
people eternally
arguing against
eachother;
it ishopeless,
butmight
bereality);
\bullet the mathematician would
image
what each case would look like in terms of so‐called filter convergenceor sheaf condition
(the
latter canactually
modelquantum
contextuality
as doneby
Isham‐Butterfield andAbramsky
et alalthough
everyoption
ismathematically possible, nonetheless,
there isusually
aunique
limit,
i.e.,
if space involved is
compact
Hausdorff(see
anytextbookon ultrafilter convergence;note however that ultrafilters may not exist without indeterministic
principles;
toput
it inlogical
terms,
complete
ormaximally
consistenttheories may notexist).
The
kingdom
ofknowledge
is notmerely
afancy metaphor:
division insociety
anddivision in
knowledge
(and
also division inreason)
are notseparate
phenomena,
butthey
are consequencesof the samedisposition
inhumanhistory,
which I characteriseby
disunitism here. Whilst
Lyotard
focused his attention toknowledge
in his end ofgrand
narrativesthesis,
it ispart
of theevenwidertendency
ofmeaning
lossormeaning
bleachfrom the
unitism/disunitism
perspective.
And thiscompletes
the first half of the shift. Thereare ofcoursecountermovementsagainst
disunitism,
and evenproponents
of dis‐unity
are nottotally
content with thedisunity
of science in some sense. Theconcept
ofre‐enchantment
by
Berman[7]
isactually
an ideal for the future as well as an accountof the after‐modern intellectual
tendency. Although
we have come tonaturally
sepa‐ rateobjects
andsubjects,
fact andvalue,
andsofourth,
in themechanistically
dividedworld after the scientific revolution in the
Renaissance,
andalthough
holisticmeaning
hasthereby
been lost in ourconception
of theuniverse, nevertheless,
Berman[7]
argues thatthey ought
tobe reunited in ordertoovercometheproblems
ofcontemporary
society
andtorecoverits holistic
coherency
intheprocess of the re‐enchantment of the world. Yet Iwould ratherarguethat the mechanisticviewis
just
dualtotheholistic view of theworld,
and so
they
arestructurally equivalent
to each other inacertain sense. It is even math‐ematically
true that the Newtonian mechanistic and Leibnizian holistic views of spaceare indeed
dually equivalent
to each other. The same holds for thetheory
ofmeaning
inthe
analytic tradition,
and the truth‐conditional and verification‐conditional semanticsmaybeseen as
equivalent
as Ihaveargued
elsewhere. This instantiatesthe basic idea ofduality
discussedthrough
thefollowing
sections. What is at issuehere, however,
is theunity
versusdisunity
debate. And mymajor point
about it is thatduality
allows us toreinstate
unity
withoutbreaking
disunity.
Letusunpack
themeaning
of this in thecaseofthe holistic one, and what
duality
says, ofcourse, is not thatthey
areequal. Although
Hegelian
dialectics would urge us to findyet
another view to reconcile the two views inconflict,
however,
duality
refuses todo so, and proposes to see them asrepresenting
the samestructure.Duality keeps pluralism
inthat anyof thetwo views is notdenied,
andthatthere is no
global,
third viewencompassing
the twoviews,
andyet
it attainsunity
in that there is one and thesame structuresharedby
them. What thispluralistic
unity
meansfurther shall be articulated in the
following
sections.2
2
Dualism
The
major
concernof thepresent
article lies inelucidating
thedynamics
orlogic
ofduality
via bothphilosophical
and mathematicalanalyses. Duality
has been the funda‐ mentalwayof humanthinking
asbroadly
seeninhumanintellectualhistory;
forexample,
Cartesian dualism andHegelian
dialectics instantiate(of
coursedifferent)
forms ofduality
inabroadersense. Inthe
following
Ishall illustratephilosophy
asaconceptual
enterprise
pursing
the ideaofduality,
arguing
thatanumber of fundamentalproblems
inphilosophy,
including
thelong‐standing
realismversus antirealismissue,
mayberesolvedby
virtue ofduality.
Theunity
of realism and antirealism has been discussed in mypreceding
works[36, 38]
aswell. Thissortof idea constitutes thephilosophical
strand of thepresent
article,
and
yet
there isanother,
mathematical strand to shedlight
on the mechanism of dual‐ity.
Inlight
ofduality
theoriesdeveloped
sofarthrough
theduality
theorists dedicated collaborativeendeavour,
I wouldsay, we are nowready
tounpack
thegeneral
mechanismof the way how
duality
emerges,changes,
andbreaks; indeed,
a succinct answer tothequestion
shall begiven
later. In thefollowing
Iamgoing
togive
aconceptual
account ofdualism, duality,
anddisduality
in theabstract; disduality
isbasically
meant to be the absence ofduality,
orduality‐breaking.
Dualism andduality
are related to eachother,
andyet
surely
different. Howthey
are different isnot thatobvious, though.
Developments
ofphilosophy
have centred around a tension between realism and an‐ tirealism(or
in Gödels tems between theright
and theleft;
see thequotation
andfootnote
above).
And the dualistic tension may be illustratedby asking
the nature ofa
variety
of fundamentalconcepts.
What isMeaning?
The realist asserts it consists inthe
correspondence
oflanguage
toreality,
whilst the antirealist contends it lies in theautonomous
system
or internalstructure oflanguage
orlinguistic
practice
(cf.
theearly
Wittgenstein
[57]
vs. the laterWittgenstein
[58];
Davidson[14]
vs. Dummett[18]).
Whatis Truth? To the
realist,
it is thecorrespondence
of assertions tofactsor statesofaffairs;
2Inmy 2016SynthesepaperI haveactuallyformulated the idea ofpluralisticunityinahttledifferent,moreabstract manner. Theconceptionofpluralisticunityheremaybeseen as aninstance ofthat.
to the
antirealist,
it has no outwardreference,
constitutedby
the internalcoherency
ofassertions or
by
somesort of instrumentalpragmatics
(cf.
Russellvs.Bradley
[8]).
Whatis
Being?
To therealist,
itispersistent
substance;
totheantirealist,
it emerges withinanevolving
process,cognition, structure,
network,
environment,
orcontext(cf.
Aristotle[1]
vs.
Cassirer/Heidegger/Whitehead [10,
26,
56 What isIntelligence?
To therealist,
it ismore than behaviouralsimulation,
characterisedby
theintentionality
ofmind;
totheantirealist,
it isfully
conferredby
copycatting
as intheTuring
testor the ChineseRoom(cf.
Searle[51]
vs.Turing
[53]).
WhatisSpace?
To therealist,
it is acollection ofpoints
with noextension;
to theantirealist,
it is astructure ofregions, relations, properties,
orinformation
(cf.
Newtonvs. Leibniz[54]; Cantor/Russell
vs.Husserl/Whitehead
[3]).
Tocast these instances of dualism in more
general
terms,
dualism may be conceived of asarising
between theepistemic
and theontic,
orbetween the formal and theconceptual
in Lawveres terms[32],
as inthefigure
below:The best known
dualism, presumably,
would be the Cartesian dualism between mind andmatter
(or
mind andbody),
inwhich the ontic realm ofmatter and theepistemic
realm of mind areseparated.
The Kantian dualism betweenthing‐in‐itself
and appearancecanreadily
beseen as a caseof theontic‐epistemic
dualism.Cassirer,
thelogical
Neo‐Kantianof the
Marburg School,
asserted thepriority
of thefunctional overthe substantival[10],
having
builtapurely functional,
genetic
view ofknowledge,
whichwasmainly
concernedwith modern science at an
early
stage
of histhought
asin Substance and Function[10],
and
yet
eventually
evolved to encompasseverything including
humanities in his maturePhilosophy of Symbolic
Form[11].
It is anall‐encompassing
magnificent Philosophy
ofCulture
[34],
indeedsubsuming myth,
art,
language,
humanities,
and bothempirical
andexact sciences. Cassirernow counts as aprecursor of what is called Structural Realism
[19,
30,
31].
Yet hisfunctionalistphilosophy
would better be characterisedasHigher‐Order
acertain
sense),
just
as incategory
theory. Cassirer, however,
doesnotsupport
ordinary
abstractionism from the concrete at
all,
and that he rather putsstrong
emphasis
onthegeneration
of the abstract viaconceptual symbolic
formation withnoprior
reference toreality,
whichitselfcomes out of thesymbolic
construction. Thesymbolic
generation
ofreality
had beenacentral theme in hisphilosophy
asawhole. Hisgenetic
viewevenpayed
due attention to the process of howstructures are
generated,
just
as intype theory.
Inlight
ofthis,
Cassirersphilosophy
may beregarded
as aconceptual underpinning
of theenterprise
ofcategory
theory,
and hisdichotomy
between substance and function asthatof
categorical duality.
Cassireractually
started his career with workon Leibniz and hisrelationalism
[9],
at which weshall have aglance
inthefollowing.
Mathematially,
theontic‐epistemic
duality
is bestrecognised
in the nature of space aforementioned. There were twoconceptions
of space at the dawn of mathematical sci‐ence: the Newtonian realist
conception
of absolute space and the Leibnizian antirealistconception
of relational space[54].
Hundreds of yearslater,
Whitehead[56]
recognised
asimilar tensionbetween
point‐free
space andpoint‐set
space(as
part
of hisinquiry
into Process andReality),
advocating
the latterpoint‐free
conception,
which may also be found inthephenomenology
of Brentano and Husserlas well(see
[3]).
Onthe onehand,
points
arerecognition‐transcendent
entities(just
likeprime
ideals/filters,
maximally
consistenttheories,
or what Hilbert called idealelements;
recall that thealgebraic
geometer
indeedidentify
points
withprime
ideals, which,
ingeneral, only
exist with thehelp
of the axiom of choiceor someindeterministicprinciple
likethat).
Ontheother,
regions
(or
any other aforementionedentities)
are morerecognition‐friendly
andepistemically
bettergrounded
(just
aspoint‐free topology
can bedeveloped constructively
or evenpredicatively
in theformal
topology
style).
Ingeneral,
realism and antirealism aregrounded
upon the onticand the
epistemic,
respectively.
As shown in the
figure,
quite
somemajor
philosophers,
whether in theanalytic
orcontinental
tradition,
have wondered about versions of theontic‐epistemic
dualism. The fundamentalproblem
of suchadualism isto accountfor how thetwodifferent realmscaninteract with each
other;
in theparticular
case of the Cartesiandualism,
it boils downto
explicating
how the mind can know about the material world whenthey
aretotally
(and
socausally)
separated.
Howcanthey causally
interact at allwhenthey
arecausally
separated?
It appearsimpossible;
this is thetypical
waythephilosopher
gets
troubledby
dualism inaccounting
for theontic‐epistemic
interaction.Philosophy
of mathematics faces an instance of the interactionproblem
as well. If therealm of mathematical
objects
and the realm of human existencearetotally separated
(in
particular causally
separated),
howcanhumanbeings
get
epistemic
access tomathemat‐ ical entities? If mathematicalobjects
exist in a Platonicuniverse,
as in Gödels realistphilosophy
forexample,
it seemsquite
hard to account for how it ispossible
to have acausal connection between humans in the
ordinary
universe and mathematical entities in the Platonic universe when the two universes arecausally
disconnected(this
sort ofproblem
isknown asthe Benacerrafs dilemma[5]).
Yet if mathematicalobjects
exist inthe
mind,
as in Brouwers antirealistphilosophy
(he
counts as an antirealist at least inthe sense of Dummett
[18]),
theaccount of interaction between the ontic and theepis‐
temic is much
easier,
since the onticis, just
by
assumption,
reduced to theepistemic
in this case;by
contrast,
then,
itgets
harder to account for the existence andobjectiv‐
ity
of mathematicalentities,
especially
transfinite ones. Forinstance,
how can humansmentally
constructfar‐reaching
transfinite entities and doeseveryones
mentalconstruc‐tion
really yield
the same results for sure?(To
remedy
the existenceproblem,
Brouweractually
endorsedarguably
non‐constructiveprinciples.
Otherwise a continuum can becountable as in recursive mathematics in the Russian
tradition;
recall that the numberof
computable
reals areonly countably
many.)
Summing
up, realistontology
makes itdifficult to account for the
possibility
ofepistemic
access to mathematicalobjects;
con‐versely,
antirealistepistemology
yields ontological
difficulties inconstructively
justifying
their existence andobjectivity.
A moral drawn from the above discussion is that there is a trade‐off between real‐
ismand antirealism:
straightforward
realistontology
leadstoinvolvedepistemology
with theurgent
issue ofepistemic
access to entities unable to exist in ourordinary, tangi‐
bleuniverse;
andstraightforward
antirealistepistemology
toinvolvedontology
with thecompelling
problem
ofsecuring
their existence andobjectivity.
Ingeneral,
realismgets
troubled
by
ouraccessibility
to abstractentities;
antirealismfacesachallenge
ofwarrant‐ing
their existence andobjectivity.
Putsimply,
an easierontology
ofsomething
often makes itsepistemology
moredifficult,
and viceversa.Something
seemsreversed betweenrealist/ontological
andantirealist/epistemological
worldviews. And this is where the idea ofduality
between realism and antirealismcomesinto theplay.
3
Duality
Everything,
from Truth andMeaning
toBeing
andMind,
has dual facetsas aforemen‐tioned.
Conceptually, duality
theory,
inturn,
is anattempt
to unitethemtogether
with the ultimate aim ofshowing
thatthey
are the two sides ofone and thesame coin. Put another way,duality
allows twothings
opposed
in dualism to be reconciled and unitedas
just
twodifferent appearances ofone and the same fundamentalreality;
in thissense,duality
is a sort of monism established on thetop
of dualism(cf.
Hegelian
dialectics asin Lawveresphilosophy
ofmathematics).
In Dummettsphilosophy
on thetheory
ofmeaning,
forexample,
he makes abinary
opposition
between realism and antirealism(cf.
Platonism and
Intuitionism/Constructivism);
what isatstake there isbasically
thelegit‐
imacy
ofrecognition‐transcendent
truthconditions,
which is allowed inrealism,
but notin antirealism. As in my recent works
[36,
38however,
the realist and antirealist con‐ceptions
ofmeaning
may be reconciled and unitedassharing
the samesort ofstructure,
evenifthey
areliterally opposed.
In viewof Dummetts constitutionthesis, according
to which thecontent ofmetaphysical
(anti)realism
is constitutedby
semantic(anti)realism
[18, 42],
this wouldarguably
count as aunification ofmetaphysics
as well as thetheory
of
meaning.
Duality
thus conceived is aconstructive canon to deconstruct dualism as it were.Whilst
having posed
theCartesiandualismasaforementioned,
Descartesalsodeveloped
analytic
geometry,
which is ina sense aprecursor ofduality
betweenalgebra
andgeometry,
eventhough
hemight
nothave beenaware thatsystems
ofequations
are dualto spacesof theirzeroloci
(logically
paraphrasing,
this amounts tothe fact thatsystems
of axioms aredual to spaces of theirmodels;
and suchcorrespondence
betweenlogic
andalgebraic
geometry
can be madeprecise
induality
theory).
Notwithstanding
that Galoistheory
may be seen as an instance of
duality,
Galois himself wouldnot have been aware of theessentially categorical duality underpinning
histheory,
either. Itwould, then,
be Riemann whofirst discoveredduality
betweengeometry
andalgebra
inamathematically
substantialform;
indeed he showed how toreconstruct(what
are nowcalled)
Riemann surfaces fromfunction
fields,
and vice versa,thereby establishing
the(dual)
equivalence
between them.Even earlier than
Riemann,
however,
Dedekind‐WeUer and Kronecker(mathematically)
gave a
purely
algebraic
conception
of space, a sort of precursor of what is now calledpoint‐free
geometry
(philosophically,
it would date backtoLeibnizasaforementioned).
The
history
ofduality
inmathematicalform, thus,
goes backtothe late 19thcentury
(it
ought
to be noted here thatduality
in mathematical formbasically
meanscategorically
representable duality,
and soduality
inprojective geometry,
forexample,
doescount as anorigin
ofduality
in this sense; it would not becategorically
representable,
at leastto my
knowledge).
Duality
then flourished in theearly
20thcentury,
asexemplified
by Hilberts, Stones, Gelfands,
andPontryagins
dualities(Hilberts
Nullstellensatz isessentially
adualequivalence
betweenfinitely
generated
reduced k‐algebras
and varieties overk foranalgebraically
closed field k).
Thediscovery
of dualitieswasthereafterfollowedby applications
tofunctionalanalysis,
general topology,
and universalalgebra
ontheonehand,
and toalgebraic
geometry, representation
theory,
and numbertheory
onthe other(interestingly,
thePontryagin duality plays
avital rôle in numbertheory
asexemplified
by
André Weils Basic NumberTheory
[55]).
And it waseventually accompanied by
one to
identify
a universal form ofduality
(before
category theory
it wasonly vaguely
understood what
exactly
different dualities have in common, andno one was abletospell
out whatpresicely
duality
is).
Today
there are agreat
variety
of dualities found acrossquite
different kinds of science as in thefollowing figure
(and
even inengineering
such asoptimisation,
linearprogramming,
controltheory,
and electrical circuittheory;
duality
thus goes far
beyond
puremathematics,
and it may sometimes be ofgenuine
practical
use as in thoseengineering
theories):
These dualities are diverse at first
sight,
andyet
tightly
intertwined with each other in theirconceptual
structures. To pursue links between different dualities is indeedone ofthe
principal
aims ofduality theory
incategory
theory. Having
a look at the abovepic‐
ture of
dualities,
it isparticularly
notable that thephysics duality
between states and observablesis,
in a way, akin to the informationalduality
betweensystems
and observ‐ ableproperties/behaviours.
Howcould we,then,
shedlight
on structuralanalogies
anddisanalogies
between diverse dualities? What is thegeneric
structure or architecture ofduality
in the firstplace?
Suchquestions
propel
theinvestigation
ofcategorical duality
theory.
Theduality‐theoretical correspondence
betweenlogic
andalgebraic
geometry
il‐ lustrates what Ulam calls ananalogy
betweenanalogies
inmathematics;
note that the Stoneduality
isconcerned withequivalence
betweensyntax
andsemantics,
and it is actu‐ally
astrengthened
version of the Gödelscompleteness
theorem(to
reinforce thispoint,
it is named Gödel‐Stone in thepicture;
technically,
theinjectivity
ofan evaluation mapin the Stone
duality exactly
amounts tocompleteness,
whereas thesurjectivity,
though
computer
scienceduality
aboveis,
in itsmathematicalsubstance,
aform of Stoneduality,
and even the
physics
duality
between states and observables maybe formulated in theStone
duality style. Quite
somepart
of theduality
picture
above, therefore,
boils downto Stone
duality,
thebirds‐eye
view of which shall begiven
below. As evident in the abovepicture
ofcategorical dualities,
category
theory today
has foundwidespread appli‐
cations in diversedisciplines
of sciencebeyond mathematics;
it would now be more like foundations of science ingeneral
than foundations of mathematics inparticular.3
Duality
is even crucialfor Hilberts programme, asCoquand
etal.[13]
assert:A
partial
realisation of Hilberts programme hasrecently
proved
successful in commutativealgebra
Oneof thekey
tools isJoyals point‐free
versionof the Zariskispectrum
as a distributive latticeIn
[13]
they
contrive aconstructive version ofGrothendiecks schemesby replacing
theirbasespaces with
point‐free
onesthrough
theStoneduality
for distributive lattices. From acategorical
point
ofview,
wecould saythat thespectrum
functorSpec
:Al\mathrm{g}^{} \rightarrow \mathrm{S}\mathrm{p}\mathrm{a}
fromanalgebraic
category
Alg
toatopological
category
Spa
amounts tothe introduction of ideal elements in Hilbertssense, and itsadjoint
functor the elimination of them. Dual‐ity,
therefore,
has contributedtoHilberts programme and constructivism. Thepoint‐free
Tychonoff
theorem isconstructive;
this is classic. Yet the state‐of‐the‐art goes farbeyond
it, encompassing
notjust
general topology
but also somemainstream mathematics such as Grothendiecks schemetheory.
In
light
of richduality
theoriesdeveloped
sofar,
wegive
succinct answerstothethreequestions
onthe mechanismofduality
(for
detail,
seemy DPhil thesis[39]):
\bullet How does
duality
emerge?
It is when the dualaspects
ofasingle
entity
arein har‐mony
with eachother;
what I call theharmony
conditionexplicates
thisharmony.
Dualadjunctions
emerge whenalgebraic
structures areharmonious withtopological
structures,
according
to(the
harmony
conditionof)
theduality theory
viacategor‐
icaltopology
andalgebra.
In dualadjunctions
betweenalgebras
and spaces, theharmony
conditionbasically
means that thealgebraic
operations
induced on thespectra
ofalgebras
are continuous. Dualequivalences
are determinedby
the ratioof
existing
term functions over allfunctions, according
to(my
understanding
of)
natural
duality theory.
'This
is thethingcategorytheoryhas aimedat since itsearly days. Granted thatquitesomecategorytheorists had moreorlessfoundationalistdoctrines, nevertheless,itwouldbeappropriatetothink ofcategorytheoryaslocal relative foundations rather thanglobalabsolutefoundations,whichiswhatsettheory isaboutinthenatureof theuniverseor the cumulativehierarchyofsets,justasbasechangeisafundamental idea ofcategorytheory. Settheorycansupporta multiverseviewasshownin recentdevelopments,andyet categorytheory intrinsicallydoesso,Iwouldsay. Notealso that
thepracticeof mathematicsisconcerned with different combinations of set‐theoretical andcategory‐theoretical ideas,and thebinaryoppositionbetweensettheoryandcategorytheoryarenot veryconstructiveorfruitfulinpractice, apartfrom
\bullet Howdoes
duality
mutate? Dual structuresget
simplified
astermfunctionsincrease;
this is what naturalduality
theory
tells us. As alimiting
case, ifexisting
termfunctionsare all functions
(
\mathrm{i}.\mathrm{e}., functionalcompleteness
inlogical
terms),
thendualspaces are Stone
(aka. Boolean)
spaces(this
is theprimal duality theorem;
extrastructures onspace are
indispensable
in thequasi‐primal duality
theorem).
Ifcon‐ tinuous functions coincide with termfunctions,
then dual structures are coherentspaces
(this
could be called continuous functionalcompleteness,
which entailsStoneduality
withrespect
to coherentspaces).
\bullet How does
duality
break? It is causedby
either an excess of the ontic or an ex‐ cess of theepistemic,
as shall be discussed below. There are someimpossibility
theorems known in non‐commutative
algebra
[6],
which exhibits an excess of theepistemic.
Thiscanhowever beremediedby
meansof sheaftheory.
Theidea ofnon‐commutative
duality theory
via sheaftheory
issimple:
we take the commutative coreofanon‐commutativealgebra,
dualiseit,
andequip
the dual space with asheafstructure to account for the non‐commutative
part.
The same methods works for abroad
variety
of non‐commutativealgebras, including
operator
algebras,
quantales,
and substructurallogics.
In substructurallogics,
the method is further extended in such a way that ingeneral
we take the structural core of a substructurallogic,
dualise
it,
and endowasheafstructurewith ittotakecareof thesubstructuralpart.
This process may be
expressed by
means of thegeneral
concept
of Grothendiecksituations.
(For
detail,
seemy DPhil thesis[39].)
Toelucidate how
duality changes
inlogical
contextsinparticular,
forexample,
whenyouweaken/strengthen
yourlogic
or extend it withoperators,
let us alsopresent
abirds‐eye
view ofdifferent
logical
dualities in arough
andyet
intuitivemanner.Stone‐type
dualitiesbasically
tellus that thealgebras
ofpropositions
are dual to the spacesof models in thefollowing
fashion:\bullet Classical
logic
isdual tozero‐dimensional Hausdorffspaces.‐
Propositions
areclosedopens,for which the law of excluded middle(LEM)
holds,
sincethe union ofaclosed open and its
complement,
which is closedopenagain,
is
equal
to the entire space.\bullet Intuitionistic
logic
is dual to certain non‐Hausdorffspaces, thatis, compact
soberspaces such that its
compact
opens form abasis,
and the interiors of their booleancombinations are
compact.4
4Thisdefinition ofHeytingspacescameoutofmyjointwork with K.Sato;LuriesHigher Topos Theory[33]gives yet
‐
Propositions
arecompact
opens. Thetopological
meaning
of LEM is zero‐dimensionality.
Ingeneral
itdoesnothold because thecomplement
ofacompact
open is not
necessarily
compact
open.\bullet Modal
logic
isdual toVietoriscoalgebras
overtopological
spaces.‐ Modal
operators
amount toKripke
relations or Vietorishyperspaces.
This iswhat is called
Abramsky‐Kupke‐Kurz‐Venema duality
inthethesis,
relating
topowerdomain
constructions in domaintheory
aswell.Note that the existence of unit ensures that duals spaces are
compact
(all
elements ofafinitary algebra
concernedyield
compact
subspaces,
andso,if there isaunitelement,
the entire space iscompact);
otherwisethey
areonly locally
compact.
Thesameholds for theGelfand
duality
aswell. There are, of course,evenmorelogical
systems
youcanthinkof: \bullet First‐orderlogic
may be dualisedby
twoapproaches: topological
groupoids
(\mathrm{i}.\mathrm{e}.,
spaces of models with
automorphisms)
andindexed/fibrational
topological
spaces(i.e.,
duals of Lawverehyperdoctrines).
‐ The latter
approach
extends tohigher‐order logic,
thusgiving
duals oftriposes
or
higher‐order hyperdoctrines.
Itjust
topologically
dualise thepropositional
value
category
ofahyperdoctrine
ortripos.
\bullet
Infinitary logic
forcesustotakenotevenlocally
compact
spacesintoaccount,
just
likethe
duality
for frames(aka. locales).
And theresulting
duality
is adualadjunction
in
general,
rather thanadualequivalence.
‐
There maynot be
enough
models orpoints
toseparate
non‐equivalent proposi‐
tions. There isno need for the axiom of choice thanksto
infinitary
operations,
i.e.,
no need toreduce infinitaries on thetopological
side into finitaries on thealgebraic.
Note that all the otherdualitiesrequire
the axiomof choiceto warrantthe existence of
enough
points.
\bullet
Many‐valued logics
are diverse. Itdepends
what sortof dual structure appears. Itis,
e.g., rationalpolyhedra
for Lukasiewiczlogic.
For otherlogics,
dual structures often includemulti‐ary
relations on spaces asinnaturalduality theory.
‐ Dualities for
many‐valued logics
aremostly
subsumed under the framework ofdualities induced
by
Janusian(aka.
schizophrenic)
objects
$\Omega$, or Chuduality
theory
on valueobjects
$\Omega$,whichmaybemultiple‐valued.
(For
detail,
see[39].)
You can combine some of
these,
andthereby
obtain morecomplex
dualities for moreare
usually
required,
andyet
there is nogeneral
method togenerate
them so far. Thestructureof
duality
combinations andcoherency
conditions thusrequired
would be worth further elucidation. Note that this is arough
picture
of dualities inlogic,
and there are some inaccuracies and omissions. Notice also that not all of these dualities are inducedby
Janusian(aka.
schizophrenic)
objects, including
those for intuitionistic and modallogics,
in whichimplication
andmodality, respectively,
are notpointwise
operations
on theirspectra
(for
mode detailed accounts see[39]).
4
Disduality
The absence of
duality,
what is nameddisduality
in thepresent
article,
isjust
as in‐teresting
on its ownright
as the presence ofduality.
According
to the discussionsofar,
duality
is about therelationships
between theepistemic
and the ontic. What is disdual‐ity
then? In anutshell, disduality
is about an excess of theepistemic
or theontic;
theduality correspondence collapses
when either of the ontic and theepistemic
is excessive. To articulate what isreally
meanthere,
let us focus upon two casesofdisduality
in thefollowing:
one is causedby incompleteness
and the otherby
non‐commutativity
as inquantum
theory.
The former shallgive
a case of the excessof theontic,
and the latter a case of theexcess of theepistemic.
As mentioned
above, completeness
maybe seen as aform ofduality
between theoriesand models. What Gödels first
incompleteness
theorem tells us is that there are notenough
formal theories to characterise the truths of intendedmodel(s)
concerned,
or toput
itdifferently,
there are some models which are unable to be axiomatised via formaltheories,
where theories are, of course, assumed to befinitary
(or
recursively
axiomatis‐able)
andstronger
than the Robinson arithemetic(the
technicalstatementof this is that theset ofstronger‐than‐Robinson
truths is notrecursively
enumerable).
If you allow forinfinitary theories,
you can nonetheless obtain acomplete
characterisation,
forexample,
of arithmetical
truths,
andyet
this is notacceptable
from anepistemological
point
ofview,
such asHilbertsfinitism. This is a caseofdisduality
duetothe excessof the ontic.Wenowturn to the otherkind of
disduality.
Let us have a look at a case of the excess of the
epistemic.
There is some sort ofincompleteness
inquantum
algebra.
The Gelfandduality
tellsus there is adualequiva‐
lence between
(possibly nonunital)
commutativeC‐algebras
andlocally
compact
Haus‐ dorff spaces. There have been differentattempts
togeneralise
it so as to include non‐commutative
algebras,
inparticular algebras
of observables inquantum
theory,
andyet,
as
long
as the duals of non‐commutativealgebras
arepurely topological,
this isactually
the
quantum
realm ofnon‐commutativity.
This is indeed a case ofdisduality
due tothe excess of theepistemic:
there are too many non‐commutativealgebras, compared
totheavailable amount of
topological
spaces. Thedisduality
may be remedied to extend thenotionofspaceso asto
include,
forexample,
sheaves ofalgebras
inadditiontotopological
spacesperse
(just
as Grothendieckindeed did in his scheme‐theoreticalduality);
in such a case,however,
both sides ofduality
get
moreor lessalgebraic
(the
same may be saidabout the Tannaka
duality
for noncommutativecompact
groups,inwhichcase duals arecategories
ofrepresentations,
and sofairly
algebraic).
There is another
thought
on the notion ofdisduality.
No canonicalagreement
exists on whatduality
meansinthefirstplace
even amongcategory
theorists aswell asamongphilosophers.
Forexample,
some sayduality
is dualequivalence,
whilst others say it isdual
adjunction
ingeneral.
A weaker notion ofduality
could count asa kind of(weaker)
disduality.
In that case we can seehow far dualadjunctions
arefrom,
andyet
howthey
(technically
always
butpractically
sometimes)
transforminto,
dualequivalences.
The different between dualequivalences
and dualadjunctions
do matterfromaphilosophical
point
of view.Think,
forexample,
ofphysics,
whichmaybe seen aspursuing
theduality
between
reality
and observation(recall
the state‐observableduality
above).
If there is aperfect
balance betweenreality
and observation that means there is a dualequivalence
between them. Yet if there is more
reality
than can be reconstructed from observationthen it is adual
adjunction
which is not a dualequivalence.
Likewise ifthere are moreobservationalor
epistemic
differences thanreality
canmetaphysically
accommodate thenit
is, again,
a dualadjunction
which is not a dualequivalence.
This is notjust
aboutphysics,
and there are, forexample,
subtle theories of different balances between statesand observation in theoretical
computer
science. What this sort ofstorytells us is thatthere can still be some sort of weaker
duality
(e.g.,
adjunction)
even in the presence ofdisduality
(e.g., non‐equivalence).
From thispoint
ofview,
the differencebetweenduality
anddisduality
may be considered relative andcontinuous,
the transition between thembeing gradual.
Disduality
is notanything
uncommon. If you have more models thantheories,
orif you have more theories than
models,
you havedisduality.
If you have more spacesthan
equations
canrepresent
as solution spaces, or if you have moreequations
thanspaces can
distinguish,
you havedisduality.
Ifyou have morereality
thanlanguage
canexpress, oryouhavemore
language
thanreality
candifferentiate,
youhavedisduality.
Theentire
enterprise
of scienceis,
in away, aboutelucidating duality
ordisduality
betweenformulae and solutions
(
\mathrm{i}.\mathrm{e}., substantival entitiessatisfying
them),
just
asphilosophy
hascentred around the dualism between the
epistemic
and the ontic. Theduality/disduality
between formulae and solutions iscrystal‐clear
inlogic
andgeometry,
as seen in theformal
correspondence
betweenlogic
andalgebraic
geometry,
and it further holds up foranalysis
andphysics
aswell. Given theSchrödinger
equation,
forexample,
youcanthinkof the Hilbert space of
solutions,
which inform us of the microstructureof thequantum
universe. What if theSchrödinger
equation
is non‐linear? You havejust got
an infinite‐ dimensionalsimplectic
manifold as the solution space. Given the Einsteinequation,
you can think of the manifold ofsolutions,
which tells you about the macro structure of the relativisticuniverse.5
If there are notenough
solutions to realiseequations,
or ifthereare not
enough
equations
to formalisesolutions,
you havedisduality
(otherwise
youhave
complete
duality),
andknowing
about that is again
in theenterprise
ofscience,
as Gödelincompleteness
served as a fruitful theorem for laterdevelopments
in differentfields.
Disduality
is ageneral
idea of the limit of theepistemic
or the ontic. In Godelincompleteness,
what isincomplete
is theepistemic,
andyet
inprinciple,
it can be theotherway around. And indeed itis thecase in
quantum
theory
that what isincomplete
isthe ontic as
non‐commutativity
tellsus(that
istosay,reality
isincomplete
rather thanquantum theory
isincomplete).
Duality
anddisduality
are notfancy
rhetoric,
butthey
dopin
down the fundamentalmeaning
and limit of the scientificenterprise.
Although
this sounds like aboldclaim, nonetheless,
it isarguably supported,
and to some extentjustified, by
numerouscases of science in whichduality
anddisduality play
central rôles.We can evenshed new
light
onthe so‐called frameproblem
in(philosophy of)
artificialintelligence
from thedisduality
point
of view. It is concerned with the fundamental limitation of thecomputational theory
of mind.My
abstract formulation of the frameproblem
is asfollows:\bullet Dimensions of
reality
arepossibly infinite;
\bullet Need \mathrm{a}
(finitary)
frame to reducepossibly
infinite dimensions and toidentify
thefinitary
scope of relevantinformation;
\bullet Need \mathrm{a}
(finitary)
meta‐frameto chooseaframe because there arepossibly infinitely
many
frames;
\bullet This meta‐frame determination process continues ad
infinitum.
Hereeveryframe is assumedtobe
finitary,
asevery formalsystem
isassumedtobefinitary
(i.e.,
recursively
axiomatisable)
inthe standard formulation ofincompleteness
theorem. Thisargument
applies
to any sort offinitary
entity,
and so, if the human is afinitary
entity
then itapplies
tothe humanaswellasthe machine. What is essential in the frameproblem
is the finitude ofbeings.
From thispoint
ofview,
theframeproblem
isabout the'What thestructureof
aphysical theoryishas beenacentralissue in recentdevelopmentsof structural realisminthe
analytic philosophyofscience. There could be different solutions. Foronething,the structureofaphysicaltheorymaybe