Vous êtes sur la page 1sur 34

GROUP KNOWLEDGE:

VIRTUE RELIABILISM, DISTRIBUTED COGNITION AND


STRONG EPISTEMIC ANTI-INDIVIDUALISM

ABSTRACT. The topic of ‘social knowledge’, broadly construed, has received considerable
attention from ethnographers, sociologists of science, philosophers of science, philosophers of
mind, and cognitive scientists alike. Yet the phenomenon of collective knowledge that is the
product of group collaborations has received virtually no attention from mainstream epistemologists.
This is a major gap in the literature that is mainly due to the discrepancy between the robustly
social nature of collaborative knowledge and the methodological individualism that underlies
mainstream epistemology. Nevertheless, the combination of virtue reliabilism from
mainstream epistemology with the hypothesis of distributed cognition from philosophy of
cognitive science can help rectify this. On this basis, the idea of epistemic group agents can be
introduced in order to account for group knowledge: i.e., cognitive success that has been
collaboratively arrived at by means of some irreducible, collective belief-forming process that
belongs to a group as whole. The outcome is the introduction of strong anti-individualism within
mainstream epistemology—i.e., the claim that, occasionally, the nature of knowledge can be
distinctively social. To flesh out this theoretical possibility the paper further focuses on two
concrete examples, which, in addition, can jointly reveal what is practically required for the
efficient design of future epistemic group agents, such as scientific research teams, social
machines and knowledge management systems.

1. INTRODUCTION
Collaborative work within science is nowadays the rule rather than the exception and
attributing knowledge to entire corporations or intelligence agencies is commonplace.
Similarly, the advent of the web and social machines,1 such as Wikipedia and Intellipedia
(CIA’s analogue of Wikipedia), highlight the progressively central role that collaborative
knowledge plays in modern society. A society, in which the means for communication and
distributed information processing become increasingly abundant, due to the rapid
advancement of technology.
Of course, as the history of science testifies (Giere 2002b, Giere and Moffatt 2003,
[REFERENCE SUPPRESSED]), close epistemic collaboration has always had a dramatic
effect on what we may know. Its full potential, however, is yet to be disclosed, with the advent
of the Social Web and Web 2.0 technologies being only some of the latest developments in
this exciting direction.
Nevertheless, if we set the current and possible future advantages to the side, the
prominence of collaborative knowledge within our society and the tendency to ascribe such
knowledge to the entirety of the relevant groups rather than to the individuals that comprise
them raises also two pressing theoretical questions: Are such ascriptions of collective knowledge
metaphorical shortcuts for referring to the sum total of the knowledge possessed by the

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1 ‘Social machines’ are “processes in which the people do the creative work and the machine does the
administration” (Berners-Lee, 1999).

! 1!
individual members of the relevant group? And if not how can mainstream epistemology
account for collaborative knowledge that is distinctively (i.e., irreducibly) social in nature?
Assuming that the answer to the first question is indeed negative, the topic of
collaborative knowledge has already been extensively discussed within the literature.
Indicatively, several sociologists (Bloor, 1991; Latour & Woolgar, 1986; Latour, 1987) and
ethnographers (e.g., Knorr-Cetina 1999;!Nersessian et al. 2003a, 2003b; Neresessian 2005;
2006) have accentuated the collaborative, social nature of scientific knowledge. Similarly,
prominent philosophers of science, such as Thagard (1993; 1994; 1997) and Giere (Giere &
Moffatt, 2003; Giere, 1990, 2002, 2006, 2007) have envisioned accounts of scientific
knowledge, whereby the scientific process is to be understood as a distinctively social process
in terms of distributed cognition. None of these authors, however, engages with the
mainstream account of knowledge in terms of justified true belief or gives an explanation of the
irreducibly social nature of collaborative knowledge.
Of course, the topic of social knowledge, broadly construed, has long made its
appearance within mainstream epistemology (indicatively, see Hardwig (1985), Tollefsen
(2006), Bird (2010), Lackey (2014) and Goldman (1999; 2004)). Again, however, such
discussions either avoid putting forward a positive account of social knowledge in terms of the
standard tripartite account of knowledge, or, if they do (as in the case of List and Pettit ((List &
Pettit, 2002, 2004, 2006, List, 2005, 2008, 2011), Tuomela (2004) and Gilbert (1994, 2007a,
2007b, 2010)),2 they fall short of addressing collaborative knowledge that is distinctively (i.e.,
irreducibly) social in nature.
There are two main reasons for this. First, the above accounts attempt to argue for
collective knowledge on the basis of collective true beliefs or acceptances. Within the
literature, however, there are considerable doubts as to whether collective beliefs and
acceptances (and the corresponding knowledge they are supposed to be components of) are
trully social in nature, or whether they can instead be reduced to the beliefs of the underlying
members of the relevant group (for an overview of the relevant debate, see Tollefsen 2004).
Secondly, even if the irreducibility of the relevant beliefs or acceptances can be established, it
is not clear that what we end up with is collective knowledge rather than mere collective true
beliefs: Given that the stadard way for distinguishing between mere true beliefs and
knowledge is to claim that the latter has also a justification component built into it, then,
contrary to the above approaches, any adequate account of collective knowledge (as opposed
to mere collective true beliefs) should focus on the irreducibly social nature of the process of

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2 Similarly, Wray (2007) and Rolin (2008) recently drew on Gilbert (2000) to provide an analysis of scientific

knowledge in terms of collective beliefs and plural subjects. For an overview of Rolin and Wray’s debate on
whether the focus should be on the entirety of the scientific community or merely on scientific research teams and
how their views relate to Gilbert’s, see Cheon (2014).

! 2!
justifying the relevant beliefs, rather than on the collective nature of the (true) beliefs
themesleves.3
In fact, not even Goldman’s (forthcoming) ‘social process reliabilism’, which is cashed
out in terms of specific processes for aggregating the justified true beliefs of the members of a
group, makes it clear that the relevant knowledge is distinctively social in the sense of being
irreducible to the propositional attitudes of the underlying individuals. As Goldman himself
notes, his view concerns the “special topic of justification aggregation”, where “the J-status
[i.e., justificatory status] of the [relevant] group belief will depend on the members' attitude
profile and on the J-statuses of those attitudes.” In other words, on Goldman’s account it is
hard to see how the group’s justification can be anything over and above the aggregate of the
justification possessed by every single one (or at least the majority) of the individual members
of the group. But, again, even if Goldman can satisfactorily establish the irreducibility of the
relevant aggregated justified true beliefs, it is obvious that his account is not concerned with
knowledge that is collaboratively produced. Instead, his account is concerned with aggregating
the justified true beliefs of individuals, where the relevant justificational attitudes have been
independently arrived at by the individual members of the group, deliberating on their own.4
To the contrary, collaborative knowledge whereby the relevant justification component
is the product of dense interactions between the members of some group.5 This is an important
difference, because such cooperative processes, as we shall see, make it impossible to use
“piecemeal decomposition and additive reassembly” (Clark, 2008, p. 116) in order to trace the
final product back to the isolated contributions of the members of the group. In other words,
collaborative knowledge is not amenable to an aggregative analysis: The final product is an
epistemic amalgam, in which it is impossible to isolate how knowledge of any single individual
was involved.

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
3 Arguably, the only possible exception here is (List 2005), which, following Nozick’s (1981) truth-tracking account

of justification, focuses on a variety of processes for aggregating individual judgments so as to track certain truths,
rather than on processes for aggregating the mere opinions of individuals. Truth-tracking conditions, however, are
not sufficient for knowledge (Pritchard 2010b) and, arguably, they are not necessary either ([REFERENCE
SUPPRESSED]. In any case, however, just as Goldman’s ‘social process reliabilism’ that I discuss in the
paragraph immediately below, List’s account focuses on collective justification that has been arrived at by
aggregating the independent judgments of individuals and not on the present paper’s target: collaborative collective
justification. See also the footnote below.
4 Both List’s (2005) and Goldman’s (forthcoming) accounts rely on the Condorcet Jury Theorem (CJT), according

to which, the members' beliefs (or judgments) must be independent of one another.
5 ‘Justification’ is an epistemological term of art that is open to several interpretations, depending on one’s
preferred account of knowledge. Perhaps, it would be more appropriate to here follow [NAME SUPPRESSED]
and [NAME SUPPRESSED]’s suggestion to employ a theory-neutral term and instead express myself by saying
that the difference of the present approach is that it focuses on the ‘epistemifying’ (or epistemicising, in American
English) component of knowledge. Should the reader’s understanding of ‘justification’ be fixed in a way that my
use of the term causes conceptual dissonance then, in what follows, s/he is welcome to substitute ‘justification’ with
the above theory-neutral term. Assuming, however, that the nature of justification (or whatever ‘epistemifies one’s
true beliefs) is still an open theoretical question, I here choose to employ the traditional term, because this way we
can be clear about how the existing approach differs from the alternative approaches to collective knowledge—i.e.,
by focusing on the justification component and not on the true belief component of knowledge—but also how it is
related to mainstream epistemology—i.e., by collectivizing the justification component of the standard tripartite
account of knowledge. !

! 3!
Accordingly, the focus of the present paper differs from the existing discussions of
social knowledge both in kind and in strength: The present account is not concerned with the
existence of collective beliefs, intentions or the process for aggregating judgments or justified
true beliefs. Instead, it is primarily concerned with the collaborative nature of the justificatory
processes of groups, and the accompanying claim that knowledge can be collective, because
collaborative (and not aggregative) justification can be collective. Moreover, collective knowledge
of this collaborative type is supposed to emerge (in a sense to be explained below) out of the
individuals’ dense interactions and it can therefore qualify as a particularly promising
candidate for knowledge that is irreducible to the sum total of the knowledge possessed by the
individual members of the relevant group.
Needless to say, of course, that this type of knowledge is particularly puzzling for the
traditionally individualistic methodology of mainstream epistemology—or, as Goldman (2010,
p. 3) puts it, of “epistemology ‘as we know it’ (in the analytic fashion).” On one hand, it is
important for mainstream epistemology to be able to account for such collaborative, collective
knowledge. On the other hand, epistemologists must keep well in mind that “the epistemic
agents of traditional epistemology are exclusively individuals” (ibid., 2): A perplexing
predicament indeed... though, perhaps, not for every mainstream approach to knowledge and
justification.
Specifically, in this paper we will see that it is possible to offer a promising account of
such irreducibly collective, collaborative knowledge in mainstream epistemological terms, by
focusing on the externalist approach of virtue reliabilism. Admittedly, however, virtue
reliabilism has a particularly strong individualist pedigree, and one may fairly wonder whether
the view is applicable to groups. Put simply, how is it possible to talk about groups as if they
are virtue epistemic agents in themselves? To prefigure the paper’s general line of
argumentation, a promising way for doing so is to combine virtue reliabilism with the idea of
group agency from the field of distributed cognition in order to introduce the idea of epistemic
group agents that can act as epistemic subjects in themselves. The overall result will be a
mainstream epistemological approach to collaborative knowledge that will allow us to account
for knowledge that is not possessed by any individual alone, but which is methodologically
individualistic all the while—i.e., cashed out in terms of the mainstream individualist account of
virtue reliabilism. If successful, this can then be a first significant step towards making sense of
strong epistemic anti-individualism within mainstream epistemology: i.e., the idea that, on certain
occasions, knowledge can be entirely social.
Before proceeding with the actual account, however, here is one final note for the
practically minded reader too: Focusing on the collaborative process of knowledge production (as
opposed to aggregates of beliefs, intentions or judgments), a further advantage of the present
approach is that it can be of direct interest, inform and be informed by extant theoretical,

! 4!
empirical and mathematical research on the topic of collaborative work and knowledge
management, broadly construed. An indicative list of examples includes studies on computer
supported collaborative learning (e.g., (Stahl, 2006)), knowledge management (e.g., (Rulke &
Galaskiewicz, 2000; Bonabeau, 2009; McAfee, 2008; Bonabeau and Meyer, 2001)), socio-
technical systems engineering (e.g., (Baxter & Sommerville, 2011; Griffith & Dougherty, 2001)),
sociological and mathematical studies of group processes (e.g., (Arrow et al., 1995; McGrath, Arrow, &
Berdahl, 2000)), educational theory and the study of small-group problem based learning (e.g., (Mennin,
2007)), studies of the Web and social machines (e.g., (Black, Delaney, & Fitzgerald, 2007; E.
Goldman, 2009; Halfaker, Geiger, Morgan, & Riedl, 2012; Halpin, Clark, & Wheeler, 2010;
Halpin, 2013; Hendler & Berners-Lee, 2010; Shadbolt et al., 2013; Smart, 2012), intelligence
studies (Schoech et al., 2002; Rønn & Høffdig, 2013; Fry & Hockstein, 1993; Davies, 2004;
Fitzgerals & Lebow, 2006; Horn, 2003; Lahneman, 2004) and so on.6
As noted in the beginning, collaborative knowledge is a particularly broad (though
largely underexplored topic) and has the potential to exhibit a vast array of exciting, future
applications.

2. VIRTUE RELIABILISM AND DISTRIBUTED COGNITION


2.1 Virtue Reliabilism
According to virtue reliabilism, knowledge is creditable true belief, which is creditable because it
is true in virtue of the manifestation of cognitive ability. On this view, cognitive ability is understood as
a reliable belief-forming process that has been appropriately integrated into the agent’s
cognitive character, where the agent’s cognitive character mainly consists of the agent’s
cognitive faculties of the brain/central nervous system (CNS), including her natural
perceptual faculties, her memory, and the overall doxastic system. In addition, however, it can
also consist of “acquired skills of perception and acquired methods of inquiry including those
involving highly specialized training or even advanced technology” (Greco 1999, 287). Here is
a relatively weak formulation of virtue reliabilism we can work with:7

COGAweak

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
6 The above list of references is far from exhaustive, and the full potential of ‘group epistemics’ (to build on
Goldman’s (1987) terminology) could only be explored on the basis of a large interdisciplinary research project on
‘Group Knowledge’.
7 COGAweak stands for Weak COGnitive Agency. Despite the name, this is a particularly promising formulation of
virtue reliabilism that is able to accommodate most (if not all) of the problems facing its alternatives. For extensive
defenses, see (Pritchard 2010b, 2012), [REFERENCE SUPPRESSED] and [REFERENCE SUPPRESSED]. In
fact, the only reasons it is supposed to be a ‘weak’ formulation of virtue reliabilism are technical ones: Firstly, it is a
necessary (rather than both a necessary and sufficient) condition on knowledge: Several epistemologists hold that
virtue reliabilism is a necessary component, but to have an adequate theory of knowledge, they argue, it must be
further supplemented by either the safety or the sensitivity principle. For such an example, see (Pritchard 2012).
Secondly, COGAweak requires that one’s cognitive success be significantly, as opposed to primarily, creditable to one’s
cognitive agency.

! 5!
If S knows that p, then S’s true belief that p is the product of a reliable belief-forming process,
which is appropriately integrated within S’s cognitive character such that her cognitive success
is to a significant degree creditable to her cognitive agency (Pritchard 2010b, 136-7).

The reason why virtue reliabilists turn to an account of knowledge that stresses the creditable
nature of the cognitive success (i.e., believing the truth) as well as its origin in the agent’s
cognitive ability has to do with knowledge-undermining epistemic luck involved in Gettier
cases. As Gettier demonstrated, one’s justified belief may turn out to be true without thereby
counting as an instance of knowledge. In the typical scenario, one’s belief, which is the
product of a defective justificatory process, just happens to be true for reasons that are
extraneous to one’s justification: In a lucky turn of events, one’s belief, which would otherwise
be false (given it is produced in a defective way), turns out to be true. Contrast this with cases
of success through the manifestation of ability. “There is a sense of ‘luck’ on which lucky
success is precisely opposed to success through virtue or ability” (Greco 2007, 58). When one’s
true belief is the product of the manifestation of one’s ability then believing the truth cannot
have been lucky; of course, one may still be lucky to believe anything at all (because, say, one
could have easily been killed), but believing the truth is not lucky itself. Accordingly, and since
credit is normally attributed in cases of success through ability, virtue reliabilists hold that
when some agent knows, his belief must be true because of cognitive ability, such that the success be
creditable to him.
So, to be as clear as possible, virtue reliabilists accentuate the importance of the process
in which one gets things right: It is not enough that one forms one’s belief on the basis of virtue
(i.e., ability) and that one’s belief be true; instead, in order to know, getting to the truth of the
matter must be creditable to one and for that to be the case, one must believe the truth because
of one’s cognitive ability.
Now, as mentioned before, according to virtue reliabilism, in order for a process to
count as a cognitive ability it must be a proper part of the agent’s cognitive character. So what
could it be required in order for a process to be so integrated? As far as common-sense
intuitions are concerned, Greco (1999, 2010) has noted that the relevant belief-forming
process must be neither strange nor fleeting (i.e., it must be a normal, dispositional cognitive
process). Despite such broad intuitions, however, Greco has noted in later work (2010) that in
order for a process to be appropriately integrated within one’s cognitive character it must
interact cooperatively with it. Specifically, he writes: “cognitive integration is a function of
cooperation and interaction, or cooperative interaction with other aspects of the cognitive
system” (2010, 152).
But what precisely is the reason that Greco spells out ‘cognitive integration’ and
‘cognitive character’ in this way? The answer has to do with a minimal notion of epistemic
agency and responsibility. Specifically, Greco is after a notion of subjective justification that is

! 6!
inline with epistemic externalism in that it denies that in order to be subjectively
justified/epistemically responsible one needs to have access to the reasons for which one’s
beliefs are reliable. And accentuating the integrated nature of one’s cognitive character
provides just this: If one’s belief-forming process cooperatively interacts with other aspects of
one’s cognitive system, then it can be continuously monitored in the background such that if
there is something wrong with it, then the agent will be able to notice this and respond
appropriately. Otherwise—if the agent has no negative beliefs about his/her belief-forming
process—he/she can be subjectively justified in employing the relevant process by default, even
if he/she has absolutely no positive beliefs as to whether or why it might be reliable. In other
words, on virtue reliabilism, provided that (1) one’s belief-forming process in integrated to
one’s cognitive character, and (2) one is conscientious (i.e., motivated to believe what is true)
such that one would indeed be responsive were there something wrong with one’s process, one
can be subjectively justified in holding the resulting beliefs merely by lacking any negative
reasons against them.8

2.2 Distributed Cognition


The hypothesis of distributed cognition is perhaps the most radical version of active
externalism. As a general approach to the nature of cognition, active externalism (Clark and
Chalmers 1998; Clark 2007, 2008; Hutchins 1996; Theiner 2011; Wheeler 2005; Menary
2006, 2007; Rowlands 1999; Wilson 2000, 2004) is standardly contrasted with Putnam (1975)
and Burge’s (1986) content, or passive externalism, as it concentrates on the aspects of the
environment that drive one’s cognitive loops in an ongoing way. Focusing on the specifics,
active externalism has appeared in the literature under several labels and formulations—e.g.,
the extended mind thesis (Clark and Chalmers 1998), cognitive integration (Menary 2007),
environmentalism (Rowlands 1999), locational externalism (Wilson 2000, 2004), the
hypothesis of extended cognition (Clark and Chalmers 1998), the hypothesis of distributed
cognition (Hutchins 1996) and so on. Here, however, the emphasis will be only on the details
of the latter two.
Focusing on cognitive processing, the hypothesis of extended cognition is the claim that
“the actual local operations that realize certain forms of human cognizing include inextricable
tangles of feedback, feedforward and feed-around loops: loops that promiscuously criss-cross
the boundaries of brain, body and world” (Clark 2007, sec. 2). Cognitive processing can and
(under the appropriate conditions) literally extends to the agent’s surrounding environment.
Think about solving a mathematical problem by using pen and paper or perceiving a chair

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
8 For a detailed analysis of this minimal yet epistemically adequate notion of subjective justification/epistemic
responsibility and its relation to cognitive integration see [REFERENCE SUPPRESSED].

! 7!
through a tactile visual substitution system.9 According to the hypothesis of extended
cognition, the involved artifacts are proper parts of the ongoing cognitive processing.
However provocative the hypothesis of extended cognition may sound, the hypothesis
of distributed cognition (Barnier et al. 2008; Heylighen et al. 2004; Hutchins, 1996; Sutton et
al., 2010; Sutton, 2008; Theiner et al. 2010; Theiner, 2013a, 2013b; Tollefsen & Dale, 2012;
Tollefsen, 2006; Wilson, 2005) may sound more challenging still. According to this form of
active externalism, cognitive processing may not just be extended beyond the agent’s head or
organism but even distributed amongst several individuals along with their epistemic artifacts.
With respect to argumentative lines, active externalism, especially in the form of the
extended mind thesis, has been traditionally associated with common-sense functionalism
(Braddon-Mitchell & Jackson, 2006). It has been recently argued (Chemero 2009,
[REFERENCE SUPPRESSED], [REFERENCE SUPPRESSED]), however, that contrary
to the extended mind thesis, the focus of both the extended and distributed cognition
hypotheses is not on mental states (such as beliefs and desires, understood in common-sense
functionalist terms), but on extended (and distributed) dynamical cognitive processes and the
overall cognitive systems these processes give rise to. Accordingly, the extended and distributed
cognition hypotheses do not need to rely for their support on common-sense functionalism.
Instead, they can be motivated on the basis of Dynamical Systems Theory (DST), which is
perhaps the most powerful, if not the only, mathematical framework for studying the behavior
of dynamical systems, in general. 10
According to this conceptual framework, in order to claim that two (or more) systems
give rise to some extended or distributed process and, thereby, to an overall extended or
distributed system (either way, to a coupled system, in DST terms), what is required is the
existence of non-linear relations that arise out of continuous reciprocal interactions between the
contributing parts (Chemero 2009, Froese et al. 2013, Sutton et al. 2008, Theiner et al. 2010,
Wegner et al. 1985, Tollefsen & Dale 2011, [REFERENCE SUPPRESSED]). The
underlying rational for this is that the aforementioned non-linear relations give rise to an
overall non-decomposable system that consists of all the contributing subcomponents operating in
tandem.
In some more detail, there are two reasons for postulating the overall coupled system:
(1) The aforementioned non-linear interactions give rise to new systemic properties that
belong only to the overall system and to none of the contributing subsystems alone (therefore

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
9 See Bach-y-Rita and Kercel (2003) for a recent review on tactile visual substitutions systems.
10 See also (Shani, 2013), whose view—viz., moderate active externalism—is similar to what we here call the
hypothesis of extended cognition (though note that Shani’s arguments do not so heavily rely on DST, and his view
is stronger than the hypothesis of extended cognition in that it denies—instead of remaining silent on the matter—
the extension of (common-sense functionalist) mental states. For more details on why common-sense functionalism
is necessary for the extended mind thesis, but not the extended and distributed cognition hypotheses, see
([REFERENCE SUPPRESSED]). Again though, note that the hypotheses of extended and distributed cognition
are neither incompatible with common-sense functionalism, nor anti-functionalist on the whole. In so far as a
cognitive process is a function, these two hypotheses are compatible with functionalism.

! 8!
one has to postulate the overall extended or distributed system); (2) Said interactions also make
it impossible to decompose the two systems in terms of distinct inputs and outputs from the
one subsystem to the other (therefore one cannot but postulate the overall system).11
Accordingly, the claim, on the basis of dynamical systems theory, is that in order to have an
extended or even distributed cognitive system—as opposed to merely an embedded cognitive
system (cf. Adams & Aizawa, 2001, 2010; Rupert, 2004, 2009)—all that is required is that the
contributing members (i.e., the relevant cognitive agents and their artifacts) interact
continuously and reciprocally with each other.12
Now, to turn our focus on distributed cognition alone, the general claim is that
“groups have the potential to display emergent cognitive properties that no individual
member has, or might even be capable of having” (Theiner et al., 2010, p. 381). But how
exactly should such emergent, irreducible, collective properties be understood? Briefly, the
answer is that emergent, collective properties refer to regularities in the behavior of the group as
a whole. Each token instance of any such behavior may still, in principle, be performed by a
single individual or at least by a random collection of them. But in order for such behavior to
be regular, the group entity must be in place.
For example, in some strange turn of events, it is possible for processes identical to the
processes of some time slice of an experiment in CERN to be temporarily realized by a single
individual alone or, at least, by an unrelated collection of individual scientists that fall short of
forming a research team. Such a lucky arrangement of events, however, would only resemble a
real experiment, as it is impossible for the processes constitutive of proper CERN experiments
to be sustained and regularly performed in the absence of some research team as a whole.
In other words, any behavior that could be classed as the manifestation of some
system’s (cognitive) properties (such as the set of processes giving rise to a scientific research
team) cannot count as such if it is merely the product of all the necessary ingredients
momentarily coming together in a fleeting way. Instead, the relevant behavior needs to arise
out of the cooperative and (thereby) self-regulatory activity of some appropriate collection of
units that will allow it to be (at least potentially) regular behavior. According to the dynamical
systems theory arguments for postulating coupled systems, however, the existence of the
requisite cooperative (non-linear) interactions between the individual members of the group

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
11 To preempt a possible worry, here, the relevant reciprocal interactions need only be continuous during the
operation of the relevant coupled cognitive system and the unfolding of any processes related to it. For example, if,
as part of her job and during normal working hours, individual S participates in distributed cognitive system X, S
does not need to continuously interact with the other members of X, when she is at home. However, whenever X is
in operation, S must continuously and reciprocally interact with the rest of the X-members. For a detailed
explanation of why the existence of non-linear relations that arise out of reciprocal interactions between agents and
their artifacts ensures the existence of extended cognitive systems see [REFERENCE SUPPRESSED].
12 For more details on how dynamical systems theory can help distinguish between the hypothesis of extended
cognition and the hypothesis of embedded cognition as well as avoid several other worries with respect to the
hypothesis of extended cognition (e.g., the ‘cognitive bloat’ worry and the ‘causal-constitution’ fallacy), see
([REFERENCE SUPPRESSED]).

! 9!
and the properties these interactions give rise to renders the postulation of the group entities
necessary.
Briefly then, the main idea of how the emergent order comes about in the case of
group entities is this: When individual members coordinate on the basis of reciprocal
interactions, they adapt mutually to each other by restricting their actions in such a way so as to
reliably—that is, regularly—achieve ends that they would only luckily—if ever—bring about
were they to act on their own. Via the application of such positive mutual constrains, which
result from, and further guide, the members’ coordinated activity, new collective properties
(i.e., regular behaviors) emerge and the collective achieves a stable configuration that is
necessary for its survival and further development. This process of “self organization and
further evolution of the collective” as Heylighen et al. put it (2004, p. 6), “effectively creates a
form of ‘social’ organization in which agents help each other so as to maximize the collective
benefit.”
Overall then, in such cases, short of postulating the relevant collective (group) entity,
it is impossible to account for the individual members’ restrained behavior. A behavior that
results from the members’ coordinated activity and which gives rise to emergent properties in
the form of unprecedented regularities in the behavior of the group as a whole.13

2.3 Virtue Reliabilism and Distributed Cognition


It has been previously argued that reading virtue reliabilism along the lines suggested by the
extended cognition hypothesis is not only an available option (see Pritchard 2010b;
[REFERENCE SUPPRESSED), but perhaps even necessary for accounting for many
instances of knowledge acquired via the employment of epistemic artifacts ([REFERENCE
SUPPRESSED]; [REFERENCE SUPPRESSED]). Here, however, a few remarks about the
strong compatibility of active externalism (both in the form of the extended and the
distributed cognition hypothesis) with virtue reliabilism should be sufficient, as the present
goal is to move further by demonstrating how the combination of the two views can provide a
positive account of group knowledge.14
To start with, the first thing to notice is that there is nothing in the formulation of
COGAweak or in the concepts involved thereof that restricts knowledge-conducive cognitive
abilities to processes within the agent’s head. To the contrary, the idea of a cognitive
character that may consist of “acquired methods of inquiry including those involving highly

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
13 For a detailed defense of group properties and entities on the basis of a naturalized version of emergence, see
([REFERENCE SUPPRESSED]).
14 It should be noted that any claim about the compatibility between two theories might carry any of the following

two meanings. The two theories may be weakly compatible in the sense of being merely consistent with each other.
Additionally, however, they may also be strongly compatible by being coherent with each other on the basis of
mutually reinforcing claims. In what follows, it is the latter kind of compatibility between active externalism and
virtue reliabilism that I have in mind.

! 10!
specialized training or even advanced technology” seems to be compatible with, or even
prefigure, active externalism.
If we focus on the details of the two theories, however, a much stronger claim can be
advanced. Specifically, both theories put forward the same condition in order for a process to
count as cognitively integrated (and thereby, by the lights of virtue reliabilism, as knowledge-
conducive): Just as proponents of extended cognition claim that a cognitive system is
integrated when its contributing parts engage in reciprocal interactions (independently of where
these parts may be located), so Greco claims that cognitive integration of a belief-forming
process (be it internal or external) is a matter of cooperative interaction with other parts of the
cognitive system.15
The upshot is that both in epistemology and philosophy of mind and cognitive
science, satisfaction of the same criterion (cooperative interaction with other aspects of the
agent’s cognitive system) is required for a process to be cognitively integrated and thereby
knowledge-conducive. Accordingly, there is no principled theoretical bar disallowing
extended or even distributed belief-forming processes from counting as knowledge-conducive.
Cognitive characters, therefore, may be extended or even distributed.
So, to give an example of an extended cognitive character, it is possible, in this way,
to explain how a subject might come to know the position of a satellite on the basis of a
telescope, while holding fast to the idea that knowledge is belief that is true in virtue of cognitive
ability.16 Even though the belief-forming process in virtue of which the subject believes the
truth is for the most part external to his organismic cognitive agency, it still counts as one of
his cognitive abilities, as it has been appropriately integrated into his cognitive character.17
Moreover, the subject satisfies COGAweak, since his cognitive success is significantly creditable
to his cognitive agency (i.e., his organismic cognitive apparatus): It is the subject’s organismic
cognitive faculties that are first and foremost responsible for the recruitment, sustaining, and
monitoring of the extended belief-forming process (i.e., telescopic observation), in virtue of
which the truth with respect to the satellite’s position is eventually arrived at.
In cases like this, therefore, even though it is the external component that accounts for
the truth-status of the agent’s belief, the agent’s cognitive agency—i.e., his organismic
cognitive faculties—is still significantly creditable for having appropriately integrated the
relevant external component into his cognitive system. In other words, even though believing
the truth is the product of some process that is for the most part external, the agent’s cognitive
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
15 Elsewhere ([REFERENCE SUPPRESSED], [REFERENCE SUPPRESSED]), it has been argued that both
theories put also forward the same broad, common sense functionalist intuitions on what is required from a process
to count as a cognitive ability. Briefly, both views state that the process must be (a) normal and reliable, (b) one of
the agent’s habits/dispositions and (c) integrated into the rest of the agent’s cognitive character/system.
16 Making observations through a telescope does qualify as a good case of cognitive extension as it is a dynamical
process that involves ongoing reciprocal interactions between the agent and the artifact. Moving the telescope
around, while adjusting the lenses, generates certain effects (e.g., shapes on the lens of the telescope), whose
feedback drives the ongoing cognitive loops along. Eventually, as the process unfolds, the coupled system of the agent
and his telescope is able to identify—that is, see—the target satellite.
17 On the basis of the feedback loops between the agent and his artifact (see the footnote above).

! 11!
success is, in accordance with the demands of COGAweak, still significantly creditable to his
organismic cognitive faculties. As Clark suggests, cognitive characters may extend, but we
should also keep in mind that “the organism (and within the organism the brain/CNS)
remains the core and currently the most active element. Cognition is organism centered
[even] when it is not organism bound” (Clark 2007, sec. 9).
These are significant considerations with respect to extended knowledge and we will
be returning to them as we go along. The present aim, however, is to account for group rather
than extended knowledge. Accordingly, we need to move further by considering propositional
knowledge, which has been arrived at not by an extended but by a collective belief-forming
process that cannot be fully reduced to the set of cognitive abilities possessed by the individual
members of the relevant group. The reason is that true beliefs that have been arrived at on
the basis of such irreducible, collective processes, will constitute cognitive success that can be
neither reduced nor thereby attributed to any particular individual member’s cognitive
agency.
By the lights of COGAweak, of course, this may initially seem like a troubling
possibility, since it leads to the following dilemma: (a) Either no one knows or (b) knowledge is
not possessed by any individual alone. Taking up option (a) would make mainstream
epistemology appear deeply problematic as it would preclude many collaboratively produced
true beliefs (such as the results of scientific research teams) from counting as knowledge. Even
worse, option (b) would render mainstream epistemology inapplicable as it goes against its
fundamentally individualistic methodology.
Fortunately, however, there is a promising way to digest the above dilemma: As we
shall see, being able to recognize, on the basis of the hypothesis of distributed cognition,
groups as cognitive subjects in themselves, can in fact render GOGAweak legitimately
applicable at the group level, thereby allowing for the claim that p is known by S, even though
it is not known by any individual alone (thus embracing the second of the two options above,
and which amounts to strong epistemic anti-individualism in mainstream epistemology).
By way of giving flesh to this picture, it is worth examining two specific cases: (1)
Transactive Memory Systems as studied by Wegner at al. (1985), and (2) the case of
knowledge acquired by research teams within scientific laboratories. Both cases involve
knowledge that is acquired at the group level, because in both cases true beliefs are acquired
on the basis of reliable collective/distributed belief-forming processes that arise out of
reciprocal interactions between the members of the relevant group and (if applicable) their
artifacts as well.
Moreover, these two examples of epistemic group agents have several properties in
common that can delineate how COGAweak can be applied to both of them as well as to any
other possible case. Specifically, in order to employ COGAweak at the group level it is

! 12!
necessary to understand who is S in ‘S knows that p’; what is the ‘reliable belief-forming
process’ that gives rise to S’s true believing; what is S’s ‘cognitive character’ into which the
collective belief-forming process must have been ‘appropriately integrated’; what is the nature
of the relevant belief component in cases of group knowledge; and, finally, what is the group’s
‘cognitive agency’, to which the ‘cognitive success must be significantly creditable’, such that S
can know that p? To set the stage for answering the above set of theoretical questions, it
should be helpful to focus on the details of the aforementioned two concrete cases first.

3. EPISTEMIC GROUP AGENTS


3.1 Transactive Memory Systems
It should be rather uncontroversial to claim that memory is a reliable belief-forming process.
Memory, that is, has a high propensity to deliver true rather than false beliefs. This is mainly
the reason why memory is very often cited as a justificatory process on the basis of which an
individual can claim to know some proposition p that he has encountered in the past.
Interestingly, however, literature originating from cognitive psychology suggests that memory
may be instantiated by more than just one individual on the basis of transactive
communication processes between two or more people. Accordingly, the first candidate for an
epistemic group agent—i.e., a collective agent who can come to know on the basis of a belief-
forming process that is not fully reducible to the cognitive abilities possessed by its individual
members—will be a Transactive Memory System (TMS): i.e., a group of individuals who
collaboratively encode, store and retrieve information.
The reason why TMSs are good candidates for group minds—and thereby for
epistemic group agents—is because, as Sutton et al. observe (2007), such systems are likely to
involve skillful interactive simultaneous coordination of people. In other words, the members of
TMSs engage in continuous reciprocal interactions, which means that they satisfy the
suggested criterion of cognitive integration. Moreover, as Wegner et al. claim, TMSs can be
used in order “to conceptualize how people in close relationships may depend on each other
for acquiring, remembering, and generating knowledge” (1985, 253). “The observable
interaction between individuals entails not only the transfer of knowledge, but the
construction of a knowledge-acquiring, knowledge-holding and knowledge-using system that
is greater than the sum of its individual member systems (1095, 256). Consider the following
example:

Suppose we are spending an evening with Rudy and Lulu, a couple married for several years.
Lulu is in another room for the moment, and we happen to ask Rudy where they got that
wonderful staffed Canadian goose on the mantle. He says “we were in British Columbia…,”
and then bellows, “Lulu! What was the name of that place where we got the goose?” Lulu
returns to the room to say that it was near Kelowna or Penticton—somewhere along lake

! 13!
Okanogan. Rudy says, “Yes, in that area with all the fruit stands.” Lulu finally makes the
identification: Peachland (257).

As Wegner at al. explain, during the discussion between Rudy and Lulu, the various
ideas they exchange lead them through and elicit their individual memories. “In a process of
interactive cueing, they move sequentially toward the retrieval of a memory trace, the
existence of which is known to both of them. And it is possible that without each other,
neither Rudy nor Lulu could have produced the item” (1985, 257).
Of course, transactive processes do not only take place during the retrieval of
memories, but they may also occur during the process of encoding and storage as well. For
instance, when partners perceive some event, each one may form some individual, private
memory of it, but they may as well discuss about it along the way. This discussion, far from
being a mere rehash of the original event, could be much more: For example, it can lead them
to the encoding of a memory, which is qualitatively different from the memories they would
have acquired, had they been on their own; similarly, if the dyad engages in transactive
processes, not during the actual event, but later on, when it is on its own, it may qualitatively
alter the members’ initial, individual memories of the event and thereby have an effect on the
process of storage as well.
So, all encoding, storage and retrieval may profit from transactive processes, in effect
giving rise to group memories and knowledge that are qualitatively different from the
memories and knowledge that one would have normally encoded, stored and retrieved were
they on their own. In order for these transactions and their effects to occur, however, Wegner
et al. note that certain specific conditions must be in place.
As they explain, “to build a transactive memory is to acquire a set of communication
processes whereby two minds can work as one” (Wegner et al. 1985, 263). The first step in
acquiring such a transactive memory system is to ensure that its candidate members will share
a common culture and language so that they can adequately understand each other and, thus,
communicate. In other words they must possess a common set of background assumptions. If
this set of common knowledge is in place, the members of the group can begin a relationship, even
as strangers, with a certain sense that each knows something that the other knows.
In turn, this will allow them to take the second step of differentiation. Couples typically
begin a relationship by revealing information about themselves to each other. Thus, in trading
knowledge of their life goals, personality traits, emotional investments and so on, they are
building the differentiation of their transactive memory: Each fact about the self that is
revealed to the other lends the other a sense of one’s expertise and experience. As each
member becomes more cognizant of the specialties of the other, the dyad’s memory as a
whole grows in differentiation.

! 14!
Finally, once common knowledge and differentiation are in place, the dyad is ready
for the last and most crucial step towards the acquisition of a TMS, i.e., the formulation of an
integrated structure. According to Wegner et al., an integrated structure effectively comes
about when and only when its members can efficiently take advantage of their common
knowledge and differentiated structure so as to create new knowledge on the basis of effective
communication feedback loops. Specifically, differentiation and common knowledge allow the
members of the group to know when they must defer to the other members’ knowledge and
expertise, and when it is time for them to take action themselves. In other words, once
differentiation and common knowledge are in place, the dyad can effectively coordinate so as
to combine the individual knowledge of its members in order to produce an integrated
understanding of some topic, on the basis of transactive encoding, storage, and retrieval
processes. The existence and the effects of this coordinated and integrated structure are the
final indication that the group has successfully formed a TMS.
So with the above details in mind, it should be apparent that TMSs do manifest
continuous reciprocal interactions between their members at almost every dimension of their
operation—i.e., during encoding, retrieval, storage, differentiation, and integration.
Accordingly, TMSs can clearly qualify as distributed cognitive systems that exhibit emergent
cognitive properties, which do not belong to their individual members, but, instead, to the
group entity as a whole.18 Moreover, given that memory does, in general, count as a
knowledge-conducive belief-forming process, TMSs are strong candidates for counting as
epistemic group agents (i.e., epistemic agents who come to know on the basis of collective belief-
forming processes that are possessed only at the group level). Accordingly, forming true beliefs
on the basis of a transactive memory process can also qualify as a first case of knowledge that
is produced by a collective (memory) process, and which is not reducible to the cognitive
processes or knowledge possessed by the individual members of the group.

3.2 Scientific Research Teams


Inspired by Knorr-Cetina’s (1999) ethnographic study of high energy physics (HEP)
experiments in CERN, Hutchins’ work (1995) on ship navigation, as well as Clark and
Chalmers’ (1998) hypothesis of extended cognition, Giere (Giere 2002a; 2002b; 2006; 2007;
Giere & Moffat 2003) has proposed to understand scientific experiments in terms of
distributed cognition. The main reason he offers for this is that, in much of scientific
experimentation, “completing the task requires coordinated action by several different
people” (2006, 711), and this coordinated action “makes possible the acquisition of knowledge
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
18 It is worth reminding here that emergent collective properties are supposed to count as such, because they
constitute the (actual or potential) regular behavior of some relevant system. Going back to the example of Rudy
and Lulu as encountered above, for any given piece of knowledge they may in fact retrieve as a dyad, it is always
possible, at least in principle, to imagine either of them retrieving that very same item individually. In the absence
of their TMS, however, it is impossible to imagine how they could individually retrieve such items on a regular basis.

! 15!
that no single person, or a group of people without instruments, could possibly acquire”
(2003, 305). Accordingly, he suggests understanding scientific research teams as emergent
distributed cognitive systems that produce knowledge that no individual could produce on
his/her own. Moreover, Giere notes, in order to understand this collective process of
knowledge acquisition and production, in order, that is, to “understand the workings of the
big cognitive system one has to consider the human-machine interactions as well as the
human-human interactions” (Giere 2002b, 292). The examples that Giere proposes are HEP
experiments (2002a), the Hubble Space Telescope (2006), and Latour’s (1999) (Giere & Mofat
2003) example of a scientific investigation that seeks to determine whether the Amazonian
rainforest is encroaching on the adjacent savannah or if the savannah is encroaching on the
rainforest.
Giere’s focus on human-machine and human-human interactions indicates that
scientific research teams are good candidates for counting as epistemic group agents. Giere, of
course, does not go into much detail when he describes the workings of the distributed
cognitive systems he considers and this is hardly surprising, given their immense complexity.
Similarly, even though we will here attempt to provide some more detail, we still need to
streamline the discussion by focusing on the interactive processes involved in an imaginary
scientific experiment. Even though this may be a simplification, the example is specifically
designed so as to capture the kind of complexity involved in real scientific investigations. The
fact that real cases might be much more complex than a simplified scenario can then add to,
rather than undermine, the present argument. This added complexity and how it may further
contribute to the overall discussion will become clear, later on, by examining certain
considerations from Knorr-Cetina’s (1999) ethnographic study of HEP experiments.
To start with, imagine a scientific experiment that consists of a laboratory and three
experts: an experimentalist, a mathematician, and a theoretical physicist. Before they even
enter the laboratory, the three of them must come together in order to decide what
instrumentation and analytical approaches they are going to employ. In other words, given
their pre-experimental understanding of the relevant theoretical domain and the available
software and hardware tools (i.e., the available mathematical and theoretical models and
instrumentation, respectively), they must brainstorm in order to choose how they should tailor
their scientific laboratory so as to most effectively study some proposition p, or a series of
them.
Once agreement on these matters is achieved, the three experts can start the actual
experimental work (for an illustration, see the workflow of figure 1). The starting point of the
experiment is the calibration of the instrument. In this first stage, the experimentalist runs
several cycles of data collection in order to check whether they correspond to data in the
literature that have been available from previous, similar experiments. In order to check the

! 16!
correspondence of course, she needs to collaborate closely with the physicist who will interpret
the available data for her. If the data turn out to correspond, the experimentalist will proceed
to the second stage of collecting data relevant to the actual experiment. Otherwise,
experimentalist and physicist will further collaborate in order to better calibrate the
instrument. Once the instrument has been calibrated and a sufficient number of data has been
collected, the experiment may then proceed to its third stage.

[ENTER FIGURE 1 HERE]

This is the stage of data validation, whereby the mathematician confirms that the
processed data are accurate and of good quality (i.e., there are no overlaps or weak peaks). If
the results of data validation are not satisfactory, he will send them back to the experimentalist
in order to perform a more accurate calibration, again, along with the physicist. If, however,
the data are satisfactory, the mathematician will go on to analyze them. This is the process of
identifying which theoretical entities are responsible for the observed signals in the sample. By
its very nature, this stage cannot be performed by the mathematician alone; the contribution
of the physicist is crucial at this point as well.
As soon as the data have been validated and analyzed by both the physicist and the
mathematician, the former can then study them on his own in order to come up with a model
for them. During this stage, the physicist will use his knowledge and training in order to ‘make
the data fit’ in mathematical models that are consistent with, or similar to models that have
been previously used in the relevant theoretical domain—very rarely, but occasionally, he
may even try to come up with an entirely new and surprising model. As soon, however, as the
data have been satisfactorily modeled, the physicist will check his work with the
mathematician. If the mathematician discovers any problems—say, with the calculations, or
the arguments involved—he will send the model back to the physicist along with comments
and suggestions, or the two of them may further collaborate in order to improve the model.
Then, after the model passes satisfactorily the ‘model check’ stage, the physicist can work on
an interpretation of it, thus, generating a working hypothesis with respect to the proposition(s)
under consideration.
Once a working hypothesis has been generated the experiment can go through to its
final stage, i.e., testing the hypothesis. At this stage, having the model at her hands—thereby
knowing how the theoretical entities are supposed to behave with respect to varying
conditions—the experimentalist can collect several data in order to check the validity of the
working hypothesis. Should the data, after analysis and validation, match the predictions of
the model—which will hardly ever happen from the first run of the whole experimental
cycle—the three experts can finally come together in order to write an article and submit it

! 17!
for publication. Otherwise, the whole cycle begins all over again on the basis of more data
that could be used in order to refine the proposed model and working hypothesis, or give rise
to entirely new ones.
It is therefore easy to see why such experiments are good candidates for counting as
group agents. Firstly, for all stages of instrument calibration, data analysis, and publication to
complete, the direct collaboration of at least two experts is required (as indicated by the two-
and three-patterned boxes in the preceding workflow). Secondly, we can draw our attention to
the direct feedback loops between the paired stages of ‘data validation’-‘instrument
calibration’; ‘testing hypothesis’-‘data collection’; and finally, ‘model check’-‘data modeling’.
Thirdly, and contrary to the initial impression the workflow may give, not all the above stages
need to follow each other in discrete steps (for instance, the experimentalist may be
continuously sending data to the mathematician for validation). It should be obvious, then,
that there is a direct influence and continuous reciprocal interactions between the activities of
all the experts involved. Therefore, as far as the process of experimentation is concerned, the
direct cognitive interdependence of these three individuals brings them together into an
irreducible emergent group agent. Put another way, as far as the cognitive processes of
justifying and producing the final working hypothesis are concerned, they are not performed
by any individual alone, but by the group agent as a whole, comprising of all the three experts
and the instruments they use to collect and process data, as well as communicate between
them.19
Of course one possible worry that might be raised against this imaginary experiment
is that it is an oversimplification. Experiments are usually much more complex, involving
more than just three experts and one experimental apparatus. It should be fairly
straightforward, however, that adding to the complexity of the experiment will actually
generate more mutual interactions between scientists and their equipment, thus, reinforcing
the idea that they form a collective entity. The question, however, is ‘to what extent is this the
case?’ What about HEP experiments like UA2 and ATLAS, which have around 100, or even
2000 participants, respectively? Due to the amount of the participants involved, it may appear
implausible that everyone, or at least most of the scientists taking part will be able to directly
influence each other.
To overcome this worry, it is instructive to focus on Knorr-Cetina’s ethnographic
study of HEP experiments. As Cetina explains, large collaborations are not run by any
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
19 With respect to the constitutive role instruments play in producing the final hypothesis, consider Heylighen et al.

(2007) who explain that when a group uses some instrument in order to communicate or process information,
altering the instrument by using it, which, in turn, alters the capabilities of the group members and the way they
interact, the involved “external media are increasingly coopted into the social organization making the
organization’s functioning even more dependent on them. As a result, the cognitive system is extended into the
physical environment and can no longer be separated by it” (8). In the previous example, this process of mutual
dependence between the individuals and their instruments is most apparent during the instrument calibration, data
collection, and data validation stages.

! 18!
individual alone and no individual is responsible for their management and organization.
Instead, such experiments are managed by discourse, which “channels individual knowledge in
the experiment, providing it with a sort of distributed cognition […], which flows from the
astonishingly intricate webs of communication pathways” (173). There is a well-established
formal grid of discourse spaces created by intersections between participants and “this grid
was and is today perhaps the most important vehicle of experimental coordination and
integration” (ibid. 174). Think of research and development meetings, panel meetings,
institute meetings, steering group meetings, collaboration meetings, referee meetings,
accelerator meetings, fixed committee meetings, special workshops dedicated to detector
complexes, submeetings of some of the former, and the very important “meetings after
meetings” (the informal exchanges that occur after scheduled events) (174).
Moreover, Cetina notes, these meetings do not take place randomly, but they rather
follow a specific sequence and temporal structure:

[This] sequential order suggests a passing of knowledge and technical decisions from the
expert group where the responsibility lies to wider circles that take note of these details and
play them back—through discussions, questions, and comments. Several rounds of this feed-
forward and feedback result in the major technical decisions that are made by institute
meetings (e.g., choices between competing detector technologies) after months of discussion.
These rounds of discussion include panel feedback and panel recommendations, which are
also channeled through working group meetings, submeetings, and plenary meetings (175).

It is apparent then that the timing of those schedules is very important for the above feedback
and feed-forward communication loops to occur. “Schedules pace, phrase, and state the work,
allocating turns within which certain points must be made or else points, and turns may be
lost. For someone to hold their turn in the collective conversation, other activities (a study, a
check, a calculation, an assembly task, a panel recommendation) must be performed on time
and the results exhibited in status reports” (190). Schedules in HEP experiments, however, are
not in any individual’s hands, but they “originate from disperse and diverse sources, among
which object requirements and the expertise of the researchers whom they regulate play a
central role” (191). What this means, in Heylighen et al.’s (2007) terminology, is that
experiments ‘self-organize’ on the basis of schedules that derive from the workings of the
experiment itself. These schedules lock experimental activities firmly into deadlines and time
slots, and they give rise to a ‘common time ordering’ of the individual activities, which
constitutes a strong coordinating force that stitches the whole collaboration together.
As collaborations grow bigger and bigger, then, their participants do not lack the
opportunity to interact with each other. To the contrary, as Cetina’s ethnographic study of
HEP experiments indicates, distant collaborators actually inform about and affect each other’s
work on the basis of formal avenues of discourse, ordered on a common time scale that
coordinates everyone’s work. Even in HEP experiments, then, and at least in most of their

! 19!
aspects, the majority of participants seem to engage in continuous reciprocal interactions, the
presence of which is here taken to be both a necessary and sufficient condition on the
emergence of collective properties and entities.
Before closing this section, however, it is also worth noting that, given the preceding
discussion, any collaborative scientific research team, just like TMSs, is most likely to also
manifest the following characteristics: First, the participants of a scientific experiment must
share some common knowledge on the basis of which they can communicate. Of course,
contrary to TMSs, this set of background knowledge does not have to be a common culture or
language understood in commonsensical terms; participants in such experiments may come
from very disparate cultural backgrounds. Instead, the members of scientific experiments must
share something close to what Kuhn (1962) termed a ‘paradigm’: a set of metaphysical and
methodological assumptions about what constitutes good scientific practice, as well as an
agreement on and understanding of the broader set of theoretical tools and equipment they
may employ. In other words, they must speak a common scientific language. Second, it must
be clear enough that any scientific experiment, just like TMSs, will have a differentiated
structure. Participants will hold beliefs about each other’s expertise and will know with whom
they must communicate in order to retrieve any information that might be relevant to their
own work. Despite the existence of the aforementioned common knowledge and differentiated
structure, however, a scientific collaboration will count as integrated only when its
participants can take advantage of the above characteristics in order to interactively create
new bits of knowledge, which differ from the knowledge each one could regularly produce on
one’s own, and which they may later use to progress their experiment further.

4. VIRTUE RELIABILISM, GROUP KNOWLEDGE AND STRONG


EPISTEMIC ANTI-INDIVIDUALISM
To recapitulate, the previous section examined two examples of groups of individuals that
form a distributed/collective cognitive agent, mainly in virtue of irreducibly collective,
cognitive, belief-forming processes that emerge out of the non-linear interactions between the
participant members and the artifacts they use. Moreover, given that the relevant collective
belief-forming processes are supposed to be knowledge-conducive, the above two cases can
qualify not just as group agents simpliciter, but as bona fide cases of epistemic group agents too.20

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
20 One common objection here is to deny that groups can qualify even as agents (let alone as epistemic group agents)
on account of lacking properties such as intentionality and consciousness—i.e., properties that are usually closely
associated with the ideas of mindedness and agency. Even though there are several ways to deflate this worry with
respect to the properties of intentionality and consciousness in particular, it is sufficient to note that the denial of
agency or mindedeness to groups would go against much of the current practice within cognitive science where
several authors are willing to ascribe mind and agency to groups (e.g., Theiner et al., 2010; Theiner & O’Connor,
2010; Theiner and Wilson, 2013; Barnier et al., 2008; Tollefsen, 2006; Wegner, 1985; Wegner; 1986; Sutton,
2008). Wilson (2001, p. 267) has cashed out this approach in terms of what he calls ‘minimal mindedness’: “X has a
minimal mind, just in case X engages in at least one psychological process or has at least one psychological ability.”

! 20!
Now to see how virtue reliabilism may be applied to such cases of epistemic group
agents, here is COGAweak again, for ease of reference:

If S knows that p, then S’s true belief is the product of a reliable belief-forming process, which
is appropriately integrated within S’s cognitive character such that her cognitive success is to a
significant degree creditable to her cognitive agency (Pritchard 2010b, 136-7).

Clearly, the first question that needs to be addressed is who S is, according to a strongly anti-
individualistic reading of COGAweak. By way of answering this, recall that group agents can be
individuated on the basis of the collective processes their component parts give rise to. The
reason is that such collective processes emerge out of the continuous mutual interactions of
the relevant components—a behavior that, according to DST, necessitates the postulation of
an overall system, in the absence of which the corresponding collective properties cannot be
accounted for. Since, however, not only human agents but also artifacts may take part in the
relevant interactions, the upshot is that the epistemic group agent, S, may be comprised of
both individuals and the artifacts they use in order to process information and communicate.
Next come the notions of a ‘reliable belief-forming process’ and S’s ‘cognitive
character’. In the discussion of virtue reliabilism in §2, it was noted that, according to Greco,
one’s cognitive character is the integrated structure of reliable and stable dispositions (i.e.,
belief-forming processes) as well as one’s memories and overall doxastic system that one
manifests when one is conscientious (i.e., motivated to believe what is true). For reasons of
parity and to remain faithful to the pursued analogy between individual and group epistemic
agents, there is no reason why the cognitive character of epistemic group agents should be
understood any differently. Notice, however, that epistemic group agents usually come
together to form just one collective belief-forming process, in virtue of which they also exist. It
may be assumed then that, usually, the cognitive character of epistemic group agents will be
identical to just that one collective belief-forming process. Greco, however, also includes in an
individual’s cognitive character one’s memories and overall doxastic system as well. Perhaps,
then, it is appropriate to add within the cognitive character of groups the shared common
knowledge of their members (i.e., their paradigm in cases of research teams)—including their
common knowledge with respect to the differentiated structure of the group—which provides
the framework for the relevant collective belief-forming process to arise, and, maybe, even
make sense.21
Now to turn to the reliability of the collective belief-forming process and the group’s
overall cognitive character as well as the general attitude of being a conscientious epistemic

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
21 Additionally, it is reasonable to assume that, in most cases, the group’s common knowledge will also include
beliefs about the member’s shared intentions, values and commitments to explicit norms (see Tollefsen 2006 and
Bratman 1999, 2006). Even though this may not be a necessary condition for every (epistemic) group agent (see
also Tollefsen and Dale 2012) it may play a largely facilitative effect to the formation of efficiently operating
collectives, especially when the members’ sub-plans are likely to clash.

! 21!
agent, recall that, as Heylighen et al. (2007) note, individuals form collective entities on the
basis of the process of self-organization, so as to bring about some desired result that maximizes
the collective benefit. That is, collections of people tend to interact until they evolve to a stable
configuration of states. Once the system has achieved this stable configuration, its component
parts have mutually adapted by restricting their interactions to those that allow them to
accomplish their end (the end, amongst other things, could be fitness, profit, or, in the present
case, objectively reliable true beliefs). Now, given that the focus here is on epistemic group
agents whose members are supposed to be conscientious (i.e., motivated to believe what is
true) it follows that the group agent’s process of self-organization, which gives rise to the
relevant collective belief-forming process, will also ensure the reliability of the latter.
Otherwise, the collective agent would have not accomplished its end of reliably leading to
truths, thereby dissolving, or would have given rise to another internal configuration and,
thus, to a different belief-forming process that would have been more appropriate (i.e., more
objectively reliable). Moreover, as in the case of individual epistemic agents, in order for the
group to be subjectively justified, it does not need to have any beliefs regarding the reliability
of its collective belief-forming process. Instead, provided (1) that the group’s distributed
cognitive character is integrated on the basis of the members’ reciprocal interactions and their
interactions with their equipment and (2) that the members are conscientious (i.e., motivated
to believe what is true) such that they would be responsive, were there something wrong with
some aspect of their overall process, we can make the following claim: The group as a
collection of individuals, or any one of its members individually, can be subjectively justified in
holding any beliefs that result from their collective belief-forming process by default, provided
that no member has expressed any negative reasons against it.
But how exactly should we understand the belief component of group knowledge?
Can the relevant beliefs be ordinary individual beliefs of (perhaps only some of) the members
of the group, or should the relevant propositions be collectively believed (or accepted) (Gilbert
1994, 2007a, 2007b, 2010, Rolin 2008, List 2011, Tuomela 2004)? And if they are collectively
believed, should we be summativists or non-summativists about them (for an overview of the
relevant debate, see Tollefsen 2004)? The answer is that whether the relevant beliefs (or
acceptances) count as the beliefs of the group, because they are of the summative or non-
summative type, the beliefs of some operative members of the group, or merely the beliefs of a
single representative (Lackey 2014) is an issue we do not here need to take a clear stance on.
In point of fact, on the present account, any of these possibilities with respect to group beliefs
may give rise to collective knowledge.
To see why, remember that, as we noted in §2.1, virtue reliabilists accentuate the
importance of the process via which one gets to the truth of the matter: One’s true belief is
creditable to one, such that it can thereby constitute knowledge, only if one arrives at the truth

! 22!
in virtue of the belief-forming process (i.e., the cognitive ability) one employed to form one’s
belief. Accordingly, on the basis of virtue reliabilism, which accentuates the importance of the
cognitive process via which one arrives at the truth, collective knowledge can be motivated on
the basis of cases where arriving at the truth of some matter is the product of a collective
belief-forming process. In other words, on the present, virtue reliabilist approach to collective
knowledge, claiming that a group, as a whole, can have knowledge of a proposition p is not
because the relevant proposition is collectively and irreducibly believed—i.e., independently
of whether this is the case or not, the relevant proposition may still qualify as collective
knowledge that is irreducible to the sum total of the individual knowledge possessed by the
members of the group. This is because, on the present account, a proposition p counts as the
knowledge of some group as a whole, if and only if getting to the truth of the matter as to
whether p (or not-p) was collectively and irreducibly achieved, on the basis of the relevant
group’s integrated cognitive character.
So, finally, with that said, it only remains to address the idea of ‘appropriate
integration’, which can also be used to understand how to think about the epistemic group
agent’s ‘cognitive agency’. Back in §2.1, where we discussed the integration of an external
element to one’s cognitive character, we claimed that in order for this to be the case, the
external element must be continuously and reciprocally interacting with the agent’s
organismic cognitive faculties. This requirement on cognitive integration clearly recognizes
the external components’ constitutive role in driving the ongoing cognitive loops, but it also
does justice to the centrality of the internal, organismic components in a way that is well in
line with the demands of COGAweak—specifically, it is in agreement with the requirement that
the cognitive success must, in the end, be significantly creditable to one’s cognitive agency: It
is the agent’s organismic cognitive faculties (i.e., the brain/CNS) that are first and foremost
responsible for the appropriate employment, recruitment and monitoring of the external
elements, on whose basis the agent will deliver outputs, which recycled as inputs will drive her
cognitive character further along, so as to eventually form a true belief with respect to some
proposition p.
Now, an analogous understanding of the process of integration and of the notion of
cognitive agency could also be provided in the case of epistemic group agents. The case of
distributed cognition and group agency, however, does not involve merely an artifact being
employed by an agent, but is, instead, about many agents and their artifacts coming together
to form a cognitive system comprising of all of them. It is not apposite, therefore, to speak of
any agent or artifact being integrated into anyone else’s cognitive character. Instead, the
question should be about what is required for the whole group of individuals, along with their
artifacts, to count as having an integrated structure. In other words, the question is ‘when does
a potentially shared belief-forming process count as having an integrated structure that can

! 23!
give rise to S’s collective cognitive character, and thereby to S?’ And the answer, given what
has been said about group agents so far, is that this won’t happen unless S’s members engage
in reciprocal interactions between them and their instruments.
Moreover, the existence of such reciprocal interactions so as to have a shared
integrated cognitive character points out how to also understand a group’s cognitive agency.
Since reciprocal interactions within a group are manifested and maintained primarily due to
the organismic cognitive faculties of its members, it is the assembly of these organismic
cognitive faculties that is first and foremost responsible for the emergence of distributed
cognition. Therefore it is the collection of these organismic cognitive faculties as a whole that
should count as the group’s cognitive agency.
Additionally, it may be recalled that agency, according to virtue reliabilism, is a rather
weak notion that manifests itself in the actions of initiating, sustaining, and monitoring the
relevant belief-forming process. Specifically, according to virtue reliabilism, agency is
manifested in the following weak, (epistemically) externalist sense of epistemic responsibility: If
there is something wrong with the relevant belief-forming process then the agent will be able
to spot this and respond appropriately. Otherwise—if there is nothing wrong—the agent can
be by default responsible in employing the relevant belief-forming process and its resulting
beliefs without even being aware that she/he does so or that the process is reliable at all.
Now, in the case of group agents, it is the collection of the individual members of the
group as a whole that initiates and sustains the relevant collective belief-forming process.
Moreover, it is the same collection as a whole that is epistemically responsible, as it is the
reciprocal interactions between the members of this assembly that allow it to respond
appropriately in cases where there is something wrong with some part of the overall belief-
forming process. Accordingly, to paraphrase Clark (2007, sec. 9), we can here say that
‘cognition is organism centered even when it is distributed’: An epistemic group agent’s
shared reliable belief-forming process—which may consist of both humans and instruments—
is primarily the product of dense interactions between the organismic cognitive faculties of its
members. Consequently, and in line with COGAweak, the collective cognitive success must be
significantly creditable to the group’s cognitive agency, which, to repeat the claim, consists of
the collection of the organismic cognitive faculties of its individual members as a whole. It is
the collection of these organismic cognitive faculties, as a whole, that is first and foremost
responsible for the emergence, appropriate employment and monitoring of the collective
belief-forming process so as to eventually accomplish its own cognizing.
Interestingly, this should also indicate why knowledge, in such cases, cannot be
attributed to any individual alone, pointing thereby towards strong anti-individualism within
mainstream epistemology. As noted in §2.2, whenever components interact reciprocally, they
give rise to a non-decomposable system consisting of all of them, without any single component

! 24!
causally standing out of the whole. In other words, the activity of the system depends on the
system as a whole and it is to the system as a whole that the success of its activity must be
attributed. Therefore, in cases of collective cognitive success that is the product of the
interactions of the organismic cognitive faculties of the members of the group, the success
must be attributed to the collection of the members’ cognitive agencies as a whole (i.e., to the
cognitive agency of the group), and to none of the individual members alone. Accordingly (to
foreshadow the paper’s final conclusion) given that virtue reliabilism understands knowledge
as creditable and reliably formed true belief, collaboratively produced true beliefs won’t be
known by any individual alone, but by the group agent as a whole, because it is only to the
group agent’s cognitive agency that such collective cognitive success can be significantly
attributed.

5. GROUP KNOWLEDGE: DISCUSSION AND CONCLUSION


In summary, there are cases where individual epistemic agents come together with the
motivation to believe what is true. This leads them to interact as a group, which self-organizes
so as to generate true beliefs. For the sake of truth and on the basis of interactive self-
organization there emerges a reliable collective belief-forming process in virtue of which the
epistemic group agent exists, and which constitutes for the most part the group’s shared
cognitive character. The reason this is a non-reducible epistemic group agent is that its
cognitive character emerges out of dense, non-linear interactions between its individual
members and their artifacts. According to dynamical systems theory, such complex
interactions and the properties they give rise to (e.g., the group’s cognitive character) cannot
be accounted for in the absence of the system (i.e., the epistemic group agent) as a whole: A
system that is more than the sum of its parts, exactly because it exhibits emergent properties in
the form of regular behaviors that none of its constituent parts has. Accordingly, any cognitive
success achieved on the basis of emergent collective belief-forming processes cannot be fully
reduced to the sum of the cognitive abilities or knowledge possessed by the individual
members of the group. Instead, it must be attributed to the collection of all the participating
individual members of the group as a whole, as it is the collection of the members’ cognitive
agencies (i.e. their organismic cognitive faculties) as a whole that is first and foremost
responsible for the emergence, maintenance and monitoring of the group’s cognitive
character.
Now to assess the situation, we need to recall two things: (1) Knowledge, according to
virtue reliabilism, is creditable true belief. (2) As noted above, any cognitive success that is the
product of the group’s cognitive character won’t be significantly creditable to any individual
alone. So given (1) and (2) the crucial question is: Should we conclude that such
collaboratively produced beliefs cannot count as knowledge at all, or should we claim that

! 25!
there can be knowledge that is not known by any individual alone? What the above discussion
indicates is that, despite the anti-individualistic ring of the latter option, it is actually a route
that mainstream epistemologists can now take: Since the cognitive success of collaboratively
produced true beliefs can be at least significantly attributed to the group agent’s cognitive
agency, such beliefs, according to COGAweak, can be known by S, the epistemic group agent. In
other words, the distributed cognition hypothesis makes it possible to recognize group agents
as epistemic subjects in themselves, such that COGAweak can now be applied to such strongly
anti-individualistic cases equally well. Therefore, we can now make sense of the claim that p is
known by S, the group agent, even though it is not known by any individual alone.
Before closing, however, it is important to repeat what this does and does not mean
on the present account. Recall that virtue reliabilism accentuates the importance of the process
via which one gets to the truth of the matter: One’s true belief is creditable to one, such that it can
thereby constitute knowledge, only if one arrives at the truth in virtue of the belief-forming
process (i.e., cognitive ability) one employed to form one’s belief. Accordingly, virtue
reliabilism can motivate group knowledge on the basis of cases where arriving at the truth of
some matter is the product of a collective belief-forming process. On the present approach,
therefore, it is possible to claim that a group, as a whole, can have knowledge of a proposition
p, not because the relevant proposition is collectively believed (cf. Gilbert, 2007a; 2007b;
1994; 2010), but because getting to the truth of the matter as to whether p (or not-p) can only
be collectively produced and is thereby creditable only to the group as a whole.22
To conclude, then, knowledge can be produced in several ways and, occasionally, it
may even be the product of the collective, collaborative belief-forming process of some
irreducible epistemic group agent. The present paper is an attempt to outline an account of
this distinctively social type of collaborative knowledge in mainstream epistemological terms,
despite the fact that the very concept of group knowledge points towards strong epistemic
anti-individualism and away from the letter (but not the spirit) of one of the core tenets of
traditional epistemology—i.e., that epistemic subjects are exclusively individuals: When
knowledge is considered from the point of view of virtue reliabilism and collaborative groups
are treated as cognitive subjects in themselves on the basis of distributed cognition, it is
possible to account for collaborative group knowledge as if it were the knowledge of a single
individual agent, thereby preserving the methodological individualism that underlies mainstream
epistemology. Besides the philosophical importance of these considerations, however, the
present account is also an attempt to highlight what are the central features one should pay

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
22 To be clear, though, it should be noted again that the present claim is not that group knowledge does not involve
belief. Just as any other type of knowledge, group knowledge will always involve belief in the proposition known,
and the relevant belief must also, on some appropriate construal, qualify as the belief of the group. Nevertheless,
whether the relevant belief (or acceptance) counts as the belief of the group, because it is of the summative or non-
summative type, the belief of some operative members of the group or merely the belief of a single representative
(for an overview, see Tollefesen 2004) is an issue we do not here need to take a clear stance on. On the present
account, any of these possibilities with respect to group belief may give rise to collective knowledge.

! 26!
attention to when constructing epistemic group agents—an endeavor that, given current
social trends, is beyond doubt of considerable practical importance.
In brief, given that the defining feature of group agents in general and epistemic
group agents in particular is the existence of dynamic reciprocal interactions between the
participating members and their artifacts, future research on collaborative group knowledge
must direct its attention to the modeling techniques of dynamical systems theory and the
concepts involved thereof. Moreover, focusing on the processes of self-organization and
cognitive integration, as well as on the prerequisite properties of common knowledge and
differentiated structure can prove particularly useful in the effective design of future
distributed cognitive systems and epistemic group agents, such as social machines, intelligence
agencies and knowledge management systems for business, academic and scientific
institutions.

REFERENCES
Adams, F. (2010). The Bounds of Cognition (1 edition.). Malden, MA: Wiley-Blackwell.
Adams, F., & Aizawa, K. (2001). The bounds of cognition. Philosophical Psychology, 14(1), 43–
64. doi:10.1080/09515080120033571
Arrow, H., Berdahl, J. L., Bouas, K. S., Craig, K. M., Cummings, A., Lebie, L., Schlosser, A.
(1995). Time, technology and groups: An integration. Computer Supported Cooperative
Work (CSCW), 4(2-3), 253–261. doi:10.1007/BF00749749
Barnier, A. J., Sutton, J., Harris, C. B., & Wilson, R. A. (2008). A conceptual and empirical
framework for the social distribution of cognition: The case of memory. Cognitive
Systems Research, 9(1–2), 33–51. doi:10.1016/j.cogsys.2007.07.002
Baxter, G., & Sommerville, I. (2011). Socio-technical Systems: From Design Methods to
Systems Engineering. Interact. Comput., 23(1), 4–17.
doi:10.1016/j.intcom.2010.07.003
Berners-Lee, T. (1999). Weaving the Web: The Original Design and Ultimate Destiny of the World Wide
Web by its Inventor (1 edition.). San Francisco: HarperOne.
Bonabaeu, E. (2009). Decisions 2.0: The Power of Collective Intelligence. MIT Sloan
Management Review. Vol. 50, No. 2. pp. 45-52.
Boanabeau, E. and Meyer, C. (2001). Swarm Intelligence: A Whole New Way to Think
About Business. Harvard Business Review.
Bird, A. (2010). Social Knowing: The Social Sense of Scientific Knowledge. Philosophical
Perspectives, 24(1), 23–56. doi:10.1111/j.1520-8583.2010.00184.x
Black, P. J., Delaney, H., & Fitzgerald, B. F. (2007). Legal issues for wikis: the challenge of
user-generated and peer-produced knowledge, content and culture. ELaw Journal,
14(1), 245–282.

! 27!
Bloor, D. (1991). Knowledge and Social Imagery. University of Chicago Press.
BonJour, L. (1985). The Structure of Empirical Knowledge. Harvard University Press.
Braddon-Mitchell, D., & Jackson, F. (2006). Philosophy of Mind and Cognition: An Introduction (2nd
Edition edition.). Malden, MA: Wiley-Blackwell.
Bratman, M. (1999). Shared cooperative activity. In Faces of Intention. Cambridge: Cambridge
University Press, pp. 93-109.
Bratman, M. (2006). Dynamics of Sociality. Midwest Studies in Philosophy, vol. 30, Shared
Intentions and Collective Responsibility, pp. 1-15.
Burge, T. (1986). ‘Individualism and psychology’, Philosophical Review, 95: 3-45.
Chemero, A. (2009). Radical Embodied Cognitive Science. MIT press.
Cheon, H. (2014). In What Sense Is Scientific Knowledge Collective Knowledge? Philosophy of
the Social Sciences, 44(4), 407–423. doi:10.1177/0048393113486523
Chisholm, R. M. (1977). Theory of knowledge. Prentice-Hall.
Clark, A., & Chalmers, D. (1998). ‘The Extended Mind’. Ananlysis 58, no. 1: 7-19.
Clark, A. (2007). ‘Curing Cognitive Hiccups: A Defense of the Extended Mind’, The Journal of
Philosophy, 104: 163-192.
Clark, A. (2008). Supersizing The Mind. Oxford University Press.
Davies, P. (2004). 'Intelligence, culture and intelligence failure in Britain and the United
States.' Cambridge Review of international Affairs, 17:3, pp. 495-520.
Fitzgerald, M., Lebow, R. (2006). 'Iraq: The Mother of all intelligence failures'. Intelligence and
National Security, 21:5, pp. 884-909.
Froese, T., Gershenson, C., & Rosenblueth, D., A. (2013). ‘The Dynamically Extended
Mind’, available at: http://arxiv.org/abs/1305.1958.
Fry, M. Hochstein, M. (1993), Epistemic Communities: Intelligence Studies and International
Relations, Intelligence and National Security, 8:3, pp. 14-28.
Giere, R. N. (1990). Explaining Science: A Cognitive Approach. Chicago: University Of Chicago
Press.
Giere, R. N. (2002a). Discussion Note: Distributed Cognition in Epistemic Cultures. Philosophy
of Science, 69(4), 637–644.
Giere, R. N. (2002b). ‘Scientific Cognition as Distributed Cognition’. In Cognitive Bases of
Science, eds. Peter Carruthers, Stephen Stitch and Michael Siegal, Cambridge:
Cambridge University Press.
Giere, R. N. (2006). The Role of Agency in Distributed Cognitive Systems. Philosophy of Science,
73(5), 710–719.
Giere, R. N. (2007). Distributed Cognition Without Distributed Knowing. Social Epistemology,
21(3), 313–320.

! 28!
Giere, R. N., & Moffatt, B. (2003). Distributed Cognition: Where the Cognitive and the
Social Merge. Social Studies of Science, 33(2), 301–310.
doi:10.1177/03063127030332017
Gilbert, M. (2000). Sociality and Responsibility: New Essays in Plural Subject Theory. Rowman &
Littlefield.
Gilbert, M. P. (2007a). Collective Epistemology. doi:10.3366/epi.2004.1.2.95
Gilbert, M. P. (2007b). Modeling Collective Belief. Synthese, Vol. 73, pp. 185-204.
Gilbert, M. P. (1994). Remarks on Collective Belief. In Socializing Epistemology: The social
Dimensions of Knowledge. F., F., Schmitt (ed.) Rowman & Littlefield
Gilbert, M. P. (2010). Belief and Acceptance as Features of Groups. Protosociology: An
International Journal of Interdisciplinary Research Vol. 16, pp. 35-69
Goldman, A. (2010). Why Social Epistemology is Real Epistemology. In Social Epistemology (pp.
1–29). Oxford University Press, Usa.
Goldman, A. I. (1987). Foundations of social epistemics. Synthese, 73(1), 109–144.
doi:10.1007/BF00485444
Goldman, A. I. (1999). Knowledge in a social world. Oxford; New York: Clarendon Press ; Oxford
University Press.
Goldman, A. I. (2004). Group Knowledge Versus Group Rationality: Two Approaches to
Social Epistemology. Episteme, 1(01), 11–22. doi:10.3366/epi.2004.1.1.11
Goldman, E. (2009). Wikipedia’s Labor Squeeze and its Consequences (SSRN Scholarly Paper No.
ID 1458162). Rochester, NY: Social Science Research Network. Retrieved from
http://papers.ssrn.com/abstract=1458162
Greco, J. (1999). ‘Agent Reliabilism’, in Philosophical Perspectives 13: Epistemology (1999). James
Tomberlin (ed.), Atascadero, CA: Ridgeview Press, pp. 273-296.
Greco, J. (2007) ‘The Nature of Ability and the Purpose of Knowledge’, Philosophical Issues 17,
pp. 57- 69.
Greco, J. (2010). Achieving Knowledge: A Virtue-Theoretic Account of Epistemic Normativity. Cambridge
University Press.
Griffith, T. L., & Dougherty, D. J. (2001). Beyond socio-technical systems: introduction to the
special issue. Journal of Engineering and Technology Management, 18(3–4), 207–218.
doi:10.1016/S0923-4748(01)00034-0
Halfaker, A., Geiger, R. S., Morgan, J. T., & Riedl, J. (2012). The Rise and Decline of an
Open Collaboration System: How Wikipedia’s Reaction to Popularity Is Causing
Its Decline. American Behavioral Scientist, 0002764212469365.
doi:10.1177/0002764212469365

! 29!
Halpin, H. (2013). Does the Web Extend the Mind? In Proceedings of the 5th Annual ACM Web
Science Conference (pp. 139–147). New York, NY, USA: ACM.
doi:10.1145/2464464.2479972
Halpin, H., Clark, A., & Wheeler, M. (2010). Towards a Philosophy of the Web:
Representation, Enaction, Collective Intelligence. Presented at the WebSci10:
Extending the Frontiers of Society On-Line, Raleigh, NC: US. Retrieved from
http://journal.webscience.org/324/
Hardwig, J. (1985). Epistemic Dependence. Journal of Philosophy, 82(7), 335–349.
Hendler, J., & Berners-Lee, T. (2010). From the Semantic Web to social machines: A research
challenge for AI on the World Wide Web. Artificial Intelligence, 174(2), 156–161.
doi:10.1016/j.artint.2009.11.010
Heylighen, F., Heath, M., & Van, F. (2004). The Emergence of Distributed Cognition: a
conceptual framework. In Proceedings of Collective Intentionality IV.
Horn, E. (2003). 'Knowing the Enemy: the Epistemoogy of Secret Intelligence. Grey Room, No.
11, pp. 58-85. Translated by Sara Ogger.
Hutchins, E. (1996). Cognition in the Wild (New edition edition.). Cambridge, Mass.: MIT Press.
Knorr, K. K. (1999). Epistemic Cultures: How the Sciences Make Knowledge. Cambridge, Mass:
Harvard University Press.
Kuhn, T. S. (1962). The Structure of Scientific Revolutions. Chicago: The University of Chicago
Press.
Lackey, J. (2014). Socially extended knowledge. Philosophical Issues 24.
Lackey, J. (2014). A deflationary Account of Group Testimony. In Essays in Collective
Epistemology. Lackey, J (ed.). Oxford University Press.
Latour, B. (1987). Science in Action: How to Follow Scientists and Engineers Through Society. Harvard
University Press.
Latour, B., & Woolgar, S. (1986). Laboratory Life: The Construction of Scientific Facts. Princeton
University Press.
Lahneman, W. (2004). 'Knowledge Sharing in the Intelligence Community After 9/11'.
International Journal of Intelligence and Counterintelligence, 17:4, 614-633.
List, C. (2005). Group knowledge and group rationality: a judgment aggregation perspective.
Episteme, 2(01), 25-38.
List, C. (2008). Distributed cognition: a perspective from social choice theory. In M. Albert,
D. Schmidtchen, & S. Voigt (Eds.), Scientific Competition: Theory and Policy (pp. 285–
308). Tübingen, Germany: Mohr Siebeck. Retrieved from
http://www.mohr.de/index_e.html
List, C. (2011). Group Agency: The Possibility, Design, and Status of Corporate Agents. Oxford ; New
York: OUP Oxford.

! 30!
List, C., & Pettit, P. (2002). Aggregating sets of judgments : an impossibility result. Economics
and Philosophy, 18(1), 89–110.
List, C., & Pettit, P. (2004). Aggregating sets of judgments : two impossibility results
compared. Synthese, 140(1-2), 207–235.
List, C., & Pettit, P. (2006a). Group Agency and Supervenience. The Southern Journal of
Philosophy, 44(S1), 85–105. doi:10.1111/j.2041-6962.2006.tb00032.x
McAfee, A. (2008). Enterprise 2.0: The Dawn of Emergent Collaboration. MIT Sloan
Management Review. Vol. 47, No. 3. pp. 21-28.
McGrath, J. E., Arrow, H., & Berdahl, J. L. (2000). The Study of Groups: Past, Present, and
Future. Personality and Social Psychology Review, 4(1), 95–105.
doi:10.1207/S15327957PSPR0401_8
Menary, R. (2006). ‘Attacking the Bounds of Cognition’, Philosophical Psychology. Vol. 19, No.
3, June 2006, pp. 329-344.
Menary, R. (2007). Cognitive Integration: Mind and Cognition Unbound. Palgrave McMillan.
Mennin, S. (2007). Small-group problem-based learning as a complex adaptive system.
Teaching and Teacher Education, 23(3), 303–313. doi:10.1016/j.tate.2006.12.016
Nersessian, N. J., Newstetter, W. C., Kurz-Milcke, E. & Davies, J. (2003). A Mixed-method
Approach to Studying Distributed Cognition in Evolving Environments.Proceeedings
of the International Conference on Learning Sciences. pp. 307 - 314.
Nersessian, N. J., Kurz-Milcke, E., Newstetter, W. C., & Davies, J. (2003).Research
laboratories as evolving distributed cognitive systems. Proceedings of The 25th Annual
Conference of the Cognitive Science Society. pp.857-862.
Nersessian, N. J. (2005). Interpreting scientific and engineering practices: Integrating the
cognitive, social, and cultural dimensions. In Scientific and Technological Thinking, M.
Gorman, R. Tweney, D. Gooding, & A. Kincannon, eds. (Erlbaum). pp. 17-56.
Nersessian, N. J. (2006). The Cognitive-Cultural Systems of the Research Laboratory.
Organization Studies, 27(1), pp. 125-145
Nozick, Robert (1981). Philosophical Explanations. Cambridge/MA: Harvard
University Press.
Pritchard , D. (2010a). ‘Knowledge and Understanding’, in A. Haddock, A. Millar & D. H.
Pritchard, The Nature and Value of Knowledge: Three Investigations, Oxford: Oxford
University Press.
Pritchard, D. (2010b). ‘Cognitive Ability and the Extended Cognition Thesis’. Synthese. 175:
133-151.
Pritchard, D. (2012). Anti-Luck Virtue Epistemology. Journal of Philosophy, (109), 247–279.
Putnam, H. (1975). ‘The Meaning of “Meaning”’. In Language, Mind and Knowledge. K.
Gunderson (ed.). Minneapolis: University of Minnesota Press.

! 31!
Rolin, K. (2008). Science as collective knowledge. Cognitive Systems Research, 9(1–2), 115–124.
doi:10.1016/j.cogsys.2007.07.007
Rowlands, M. (1999). The Body in Mind: Understanding Cognitive Processes. New York: Cambridge
University Press.
Rulke, D. L., & Galaskiewicz, J. (2000). Distribution of Knowledge, Group Network
Structure, and Group Performance. Manage. Sci., 46(5), 612–625.
doi:10.1287/mnsc.46.5.612.12052
Rupert, R. D. (2004). Challenges to the Hypothesis of Extended Cognition. Journal of
Philosophy, 101(8), 389–428.
Rupert, R. D. (2009). Cognitive Systems and the Extended Mind (First Edition edition.). Oxford ;
New York: OUP USA.
Rønn, K., Høffding, S. (2013). 'The Epistemic Status if Intelligence: An Epistemological
Contribution to the Understanding of Intelligence, Intelligence and National Security,
28:5, pp. 694-716.
Schoech, D., Fitch, D., Machtadden, R., Schkade, L. (2002). From data to Intelligence.'
Administration in social work, 26:1, pp. 1-21.
Shadbolt, N. R., Smith, D. A., Simperl, E., Van Kleek, M., Yang, Y., & Hall, W. (2013).
Towards a classification framework for social machines. Workshop presented at the
SOCM2013: The Theory and Practice of Social Machines. Retrieved from
http://eprints.soton.ac.uk/350513/
Shani, I. (2013). Making it mental: in search for the golden mean of the extended cognition
controversy. Phenomenology and the Cognitive Sciences, 12(1), 1–26.
doi:10.1007/s11097-012-9273-z
Smart, P. R. (2012). The Web-Extended Mind. Metaphilosophy, 43(4), 446–463.
doi:10.1111/j.1467-9973.2012.01756.x
Stahl, G. (2006). Group Cognition: Computer Support for Building Collaborative Knowledge. Boston,
Mass.: MIT Press.
Steup, M. (1999). A Defense of Internalism. In The Theory of Knowledge: Classical and Contemporary
Readings, 2nd edition. Wadsworth Publishing.
Sutton, J. (2008). Between Individual and Collective Memory: Interaction, Coordination,
Distribution. Social Research, 75(1), 23–48.
Sutton, J., Harris, C. B., Keil, P. G., & Barnier, A. J. (2010). The psychology of memory,
extended cognition, and socially distributed remembering. Phenomenology and the
Cognitive Sciences, 9(4), 521–560. doi:10.1007/s11097-010-9182-y
Thagard, P. (1993). Societies of minds: Science as distributed computing. Studies in History and
Philosophy of Science Part A, 24(1), 49–67. doi:10.1016/0039-3681(93)90024-E

! 32!
Thagard, P. (1994). Mind, Society, and the Growth of Knowledge. Philosophy of Science, 61(4),
629–645.
Thagard, P. (1997). Collaborative Knowledge. Noûs, 31(2), 242–261. doi:10.1111/0029-
4624.00044
Theiner, G. (2011). Res Cogitans Extensa: A Philosophical Defense of the Extended Mind Thesis, Bern,
Switzerland: Peter Lang GmbH, Europaischer Verlag der Wissenschaften.
Theiner, G. (2013a). Onwards and Upwards with the Extended Mind: From Individual to
Collective Epistemic Action. In L. Caporael, J. Griesemer, & W. Wimsatt (Eds.),
Developing Scaffolds (pp. 191–208). MIT Press.
Theiner, G. (2013b). Transactive Memory Systems: A Mechanistic Analysis of Emergent
Group Memory. Review of Philosophy and Psychology, 4(1), 65–89.
doi:10.1007/s13164-012-0128-x
Theiner, G., Allen, C., & Goldstone, R. L. (2010). Recognizing group cognition. Cognitive
Systems Research, 11(4), 378–395. doi:10.1016/j.cogsys.2010.07.002
Theiner, G. & O’Connor, T. (2010). ‘The Emergence of Group Cognition’, In A. Corradini
& T. O'Connor (eds.), Emergence in Science and Philosophy. Routledge. 6—78.
Theiner, G. & Wilson, R. (2013). 'Group Mind.' Encyclopedia of Philosophy and the Social Sciences.

Tollefsen, D. (2006). Group Deliberation, Social Cohesion, and Scientific Teamwork: Is


There Room for Dissent? Episteme: A Journal of Social Epistemology, 3(1), 37–51.
doi:10.1353/epi.0.0008
Tollefsen, D., & Dale, R. (2012). Naturalizing joint action: A process-based approach.
Philosophical Psychology, 25(3), 385–407. doi:10.1080/09515089.2011.579418
Tollefsen, D. P. (2006). From extended mind to collective mind. Cognitive Systems Research, 7(2–
3), 140–150. doi:10.1016/j.cogsys.2006.01.001
Tollefsen, D. P. (2006). Group Deliberation, Social Cohesion, and Scientific Teamwork: Is
There Room for Dissent?. Episteme. 3, pp. 37-51.
Tollefsen, D. P. (2004). Collective Intentionality. Internet Encyclopedia of Philosophy
Tuomela, R. (2004). ‘Group Knowledge Analyzed’. Episteme, 1 (2), pp. 109-127.
Wegner, M., Giuliano, T., Hertel, P. (1985). ‘Cognitive interdependence in close
relationships’. In W. J. Ickes (Ed.), Compatible and incompatible relationships (pp. 253–
276). New York: Springer-Verlag.
Wegner, D. (1986). Transactive Memory: A Contemporary Analysis of the
Group Mind. Theories of Group Behavior. Eds. B. Mullen and G. R.
Goethals. New York: Springer-Verlag.
Wheeler, M. (2005). Reconstructing the Cognitive World. MIT Press, Cambridge, Massachusetts.
Wilson, R. (2001). 'Group-Level Cognition.' Philosophy of Science, 68 (3), pp. 262-273.

! 33!
Wilson, R., A. (2004). Boundaries of the Mind: The individual in the Fragile Sciences: Cognition. New
York: Cambridge University Press.
Wilson, R. A. (2005). Collective memory, group minds, and the extended mind thesis. Cognitive
Processing, 6(4), 227–236. doi:10.1007/s10339-005-0012-z
Wray, K. B. (2007). Who has Scientific Knowledge? Social Epistemology, 21(3), 337–347.
doi:10.1080/02691720701674288

!
!
!
!

! 34!

Vous aimerez peut-être aussi