A moderate emergentist account of group agency

(1)

A moderate emergentist account of group agency

MSc Thesis (Afstudeerscriptie)

written by Flavio Tisi

(born July 10th, 1994 in Foggia, Italy)

under the supervision of Prof Dr Sonja Smets and Prof Dr Marija Slavkovik, and submitted to the Examinations Board in partial fulfillment of

the requirements for the degree of

MSc in Logic

at the Universiteit van Amsterdam.

Date of the public defense: Members of the Thesis Committee: August 28, 2020 Dr Benno van den Berg (Chair)

Prof Dr Sonja Smets (Supervisor) Prof Dr Marija Slavkovik (Supervisor) Dr Maria Aloni

Dr Federica Russo Dr Aybüke Özgün

(2)

(3)

Abstract

In my thesis, I argue in favour of a moderate emergentist account of group agency.

In Chapter 2, I survey several accounts of group agency that have been stud-ied in the literature, and I introduce formal frameworks that are used later in the thesis. In Chapter 3, I argue in favour of a specific definition of group agent, while refuting the claim that group agents’ belief sets need to be consistent. In Chapter 4, I explain why I believe that moderate emergentism is the right ac-count of group agency, by refuting both eliminativism and strong emergentism as theoretical options. Having argued in favour of moderate emergentism, I dedicate Chapter 5 to investigate possible specifications of the supervenience relation, as well as clarifying the kind of supervenience relation moderate emer-gentism requires us to subscribe to.

(4)

(5)

Acknowledgements

At the end of this intellectual journey, I would like to thank Prof. Sonja Smets and Prof. Marija Slavkovik for their careful supervision during my thesis work, and Prof. Francesco Berto for his precious advice.

(6)

(7)

Introduction

In this thesis, I explore the topic of group agency. Group agency has been thoroughly studied in philosophy, and especially in metaphysics, where the on-tological status of group agents and shared intentions has been analysed 1_{. It}

has also played a central role in AI studies which focused on trying to model the behaviour of collectives composed of several individual agents2_.

There are several reasons for studying group agency. First of all, it is a subject interesting in its own right. But it also played a central role in social science in general, especially in phenomena such as the madness of the crowd and group polarisation 3_{. Nowadays, phenomena of this kind are even more}

important than they used to be, given the central role that social networks such as Facebook, Twitter, Instagram have developed in recent years4_{. Nonetheless,}

while it would definitely be interesting to study the role of group agency with reference to social networks, this will not be the focus of my thesis. Instead, in this thesis I focus on the notion of group agency from a philosophical and meta-physical perspective. I am interested in what group agents are, which features we can ascribe to them and how these features relate with the individuals that groups comprise.

Specifically, I defend a strong emergentist account of group agency. What do I mean by a strong emergentist account ? A strong emergentist account of group agency is an account according to which it is legitimate to ascribe beliefs and attitudes to group agents, while at the same time these ascriptions supervene on the individuals ascriptions of the individual members the groups consist of.

In order to defend my thesis, I survey the scientific literature. I then offer and defend my definitions of group and group agency, while specifying whether all group agents should have coherent beliefs set or not. Then, I argue in favour

1_{See for instance (Epstein 2018), (Roth 2017), (A. Baltag et al. 2018)} 2_{See (Halpern and Moses 1990)}

3_{For some recent work on these see (Pedersen, S. Smets, and ˚}_{Agotnes 2019), (Baltag,} Hansen, and Smets 2013), (Achimescu, Alexandru Baltag, and Sack 2016)

4_{For recent references on logic and social networks see (Alexandru Baltag et al. 2019) and} (Dennis and Slavkovik 2020)

(10)

of the supervenience of group agent attitudes on the individual attitudes of the individual agents. Having done so, I explore some specific accounts of group beliefs, and which metaphysical kinds of supervenience relation can be applied to my supervenience thesis. Throughout my thesis, in order to justify my claims, I resort to philosophical arguments, and sometimes I employ thought experiments to derive counterexamples to universal claims I disagree with.

The thesis is structured as follows.

Chapter 2 is devoted to the survey of the scientific literature. My primary source throughout this thesis is (List and Pettit 2011). The reason why I rely primarily on (List and Pettit 2011) is that their book is a foundational text in the field of group agency, and therefore it blends together in a coherent framework theories and conceptual tools coming from different fields. However, while (List and Pettit 2011) remains my primary source in this chapter, I also rely on Stanford Encyclopedia entrees such as (O’Connor and Wong 2020), (Epstein 2018) and (B. McLaughlin and Bennett 2018) and other references. In fact, I strive to get a good understanding of the theories and conceptual tools presented by (List and Pettit 2011), so as to be able to formulate my own theories and arguments.

In Chapter 3, I provide my own definitions of groups and group agents, and I contrast these definitions with the ones offered by (List and Pettit 2011). Furthermore, I argue in favour of an alternative theory than the one advanced in (List and Pettit 2011) regarding the constraints group agents should satisfy. In fact, I argue that it is not necessary for groups to exhibit robust group rationality, and therefore to have coherent belief sets.

In Chapter 4, I argue in favour of the moderate emergentist account. In order to do so, I argue against eliminativism, the position according to which ascriptions of beliefs and attitudes to group agents are not legitimate, and strong emergentism, according to which group agents attitudes and beliefs do not su-pervene on the ones of the individual members the group is composed of.

In Chapter 5, I investigate instead the supervenience relation. In the first part, I present accounts due to (List and Pettit 2011) and (Gilbert 1987). In the second part of the chapter, my focus shifts instead to the metaphysical debate regarding the supervenience relation: I argue that only strong global supervenience and strong multi-domain supervenience can be applied to model the supervenience relation between group attitudes and the individual attitudes, and I explore whether these two versions of the supervenience relation can be considered equivalent.

Finally, Chapter 6 is devoted to the conclusion of my research, where I summarise what I have argued in the thesis and I lay down possible paths for future research.

(11)

Chapter 2

The group agency account

of List and Pettit

2.1 Introduction

In this chapter, I am going to analyse the scientific literature on group agency. First of all, I will analyse the concept of group and group agent. Then, I will shift my attention to the debate regarding the metaphysical status of group agents: the theories of eliminativism, strong emergentism and moderate emergentism. Having done so, I will move to consider several versions of the supervenience relations that have been studied in the scientific literature. Finally, I will intro-duce some formal results regarding the aggregation of individual attitudes due to (List and Pettit 2011).

2.2 The concept of group and group agent

2.2.1 Group and group agent definitions

The concept of group features prominently in the scientific literature on social ontology. As Epstein writes in the Stanford Encyclopedia of Philosophy entry on Social Ontology (Epstein 2018), an important subject of social ontology is rep-resented by social groups. In fact, as (Epstein 2018) points out, social ontology investigates the definitions of social groups and their agential counterpart, group agents. But social ontology does not restrict its investigation to the metaphysi-cal status of groups and group agents. As (Epstein 2018) writes, social ontology also tries to answer questions related to which features we could and should as-cribe to group agents, whether and how group agents form attitudes and beliefs, and on which bases they do so.

What is a group? (List and Pettit 2011) offer the following definition:

(12)

Definition 1 ((List and Pettit 2011) definition of Group) A group is a collection that can survive change of membership

This definition is important because it clearly specifies that groups are col-lections, but collections of a specific kind: collections that can survive change of membership. Not all collections can do it: for instance, the set of people who happen to be in a certain place at a certain time (for instance: all the people who gathered in Times Square, NY on New Year’s Eve 2020) is a set that cannot survive change of memberships. On the other hand, the collection of MPs of the House of Commons in the UK is a collection that can survive change of membership: and indeed it does, since the members of this assembly are changed routinely at every General Election.

As we can notice, (List and Pettit 2011)’s definition is framed in terms of collection. The concept of collection of agents is taken as primitive in the framework of (List and Pettit 2011).

The idea that a defining feature of groups is their ability to survive change of membership is far from being uncommon. (Epstein 2018) writes:

“What are social groups? One debate in the literature concerns the kind of entities that social groups are: collections, classes, sets, fusions, structures, or some other kind of entity. It may seem natural to think of a group as a set of people in the mathematical sense. However, groups can persist through changes in membership, while sets are generally understood as having their members essentially” Another very important concept that features prominently in (List and Pettit 2011)’s analysis is the idea of agency. In fact, as they write in (List and Pettit 2011), in their book their aim is “to defend the logical possibility of group agents”. Therefore, they (List and Pettit 2011) “introduce basic conditions of agency, investigate how individual inputs can be aggregated into group-level outputs, and explore different possible structures of a group agent”.

But what, then, is a group agent? This concept has been elusive so far. However, (List and Pettit 2011) provide a clear definition of group agent. Their definition of group agent follows as an application of the concept of agent to the concept of group. This means that a group agent is a group which exhibits the features of agency, i.e. it is a group that satisfies the necessary and sufficient conditions for being an agent. The concept of agent is so defined in (List and Pettit 2011):

Definition 2 ((List and Pettit 2011)’s definition of Agent) An agent is a system with these features: it has representational states, motivational states, and a capacity to process them and to act on their basis.

1. First feature. It has representational states that depict how things are in the environment.

2. Second feature. It has motivational states that specify how it requires things to be in the environment.

(13)

2.2. THE CONCEPT OF GROUP AND GROUP AGENT 13 3. Third feature. It has the capacity to process its representational and moti-vational states, leading it to intervene suitably in the environment when-ever that environment fails to match a motivating specification.

We next focus on the concept of agency.

2.2.2 An analysis of agency

According to (List and Pettit 2011), only groups can be agents, and not mere collections. This is an important feature of groups that distinguishes them from collections. As they write (List and Pettit 2011):

“Could mere collections, and not just groups, count as agents? We do not think so. Any multi-member agent must be identifiable over time by the way its beliefs and desires evolve. So there must be a basis for thinking of it as the same entity, even as its membership changes due to someone’s departure or the addition of new members. Thus it is a candidate for being taken as a group, not just a mere collection.”

If we want to represent formally the argument they provide in favour of their thesis, we could do it as follows:

Argument 1 ((List and Pettit 2011)’s Group-only Agency) 1. If a col-lection is an agent, then it is identifiable over time by the way its beliefs and desires evolve

2. If a collection is identifiable over time by the way its beliefs and desires evolve, then there is a basis for thinking of it as the same entity, even as its membership changes

3. If there is a basis for thinking of a collection as the same entity, even as its membership changes, then that collection is a group

4. Therefore, if a collection is an agent, then it is a group

However, the definition of agency provided so far is still not sufficiently clear. In fact, several terms employed have to be defined better: what are representational states? What are motivational states? (List and Pettit 2011) offer the following definitions:

Definition 3 (Representational and motivational state) A state is ‘representational’ if it plays the role of depicting the world, and ‘motivational’ if it plays the role of motivating action.

However, in subsequent parts of their writing, (List and Pettit 2011) come to the conclusion that the original wording of the Third feature of Agent that they provide is too strong, and they opt instead for a more ‘relaxed’ wording:

(14)

Definition 4 ((List and Pettit 2011)’s relaxed definition of Agent) An agent is a system that has representational and motivational states such that in favorable conditions, within feasible limits, it acts for the satisfaction of its mo-tivations according to its representations.

We have seen that in the framework of (List and Pettit 2011) representational states play the role of depicting the world, whereas motivational ones play the role of motivating actions. But in order to gain a better understanding of (List and Pettit 2011)’s framework, we have to consider the mirror concept of attitudes:

“An agent, we have seen, is a system that forms representations and motivations about the world and acts so as to realize its motivations according to its representations. The agent’s representations and motivations can be modeled as attitudes held towards certain propo-sitions. An agent’s representational attitude towards a proposition (or scenario) captures its belief in that proposition, its motivational attitude towards it its desire for it.”

Therefore, a group’s attitude too can be representational or motivationa (List and Pettit 2011):

“The attitude can be either representational or motivational; the object encodes what is being depicted or specified in the represen-tation or motivation. Although there is room for some theoretical maneuver here, we take the object to be a proposition. If the at-titude towards it is representational, the proposition is depicted as being true; if the attitude is motivational, it is specified as one the agent is motivated to make true through its actions.”

(List and Pettit 2011) specify that it does not matter which interpretation of the concept of proposition we choose to rely on, since the standard inter-pretations mostly fit with their account: propositions, however, come in many shapes.

More specifically, (List and Pettit 2011) distinguish between binary attitudes and non-binary attitudes. They go on to define representational binary attitudes in the following way:

“Consider representational attitudes. A binary representational at-titude towards some proposition p is what we call a judgment on p, where we assume that an agent either judges that p or doesn’t judge that p, with no halfway houses. ”

Non-binary representational attitudes are instead credences (List and Pettit 2011):

“By contrast, a typical non-binary representational attitude towards p is what decision theorists call a credence or degree of belief the agent assigns to p, which may take any value between zero and one. ”

(15)

2.2. THE CONCEPT OF GROUP AND GROUP AGENT 15 According to (List and Pettit 2011) there is no sharp distinction between binary and binary attitudes, and it is easily possible to transform non-binary attitudes in non-binary ones. For instance, degrees of probability assigned to a proposition can be transformed in judgements regarding the probability of a that proposition, and vice versa. If as an agent I assign a degree of probability 1/6 to the proposition ‘the die will show number 3’, I can equivalently state that I judge that ‘there is a probability of 1/6 that the die will show number 3’.

In a similar vein, the distinction between representational and preferential attitudes (where preferential attitudes are the attitudinal correspondent of moti-vational states) is not as strong as someone might think, because in this case too we can rephrase preferential attitudes into representational ones (List and Pettit 2011). In this case too it is possible to paraphrase judgements with preferences, and viceversa.

For instance, the statement ‘I prefer to eat vegetables tonight’ can be effec-tively paraphrased as ‘I judge that it is preferable to eat vegetable tonight’, and the converse applies as well.

While according to (List and Pettit 2011), it is not very important which conceptualisation of propositions we rely on, it is more important the language the propositions are expressed in. As (List and Pettit 2011) point out, the lan-guage on which propositions are defined can vary according to what we need to express. We could resort to the expressiveness of propositional logic, first-order logic, modal logic and deontic logic, or alternatively, to using a metalanguage that “can express propositions that assign properties to other, ‘object-language’ propositions.” This way we would be able to express things like ‘Agent A be-lieves that p’.

It is important to highlight that according to (List and Pettit 2011) it is possible to attribute beliefs also to groups that do not constitute group agents. This consideration allows (List and Pettit 2011) to lay down a further feature of group agency: the attitudes of a group agent cannot be inconsistent or irrational. As they write (List and Pettit 2011):

“It is methodologically defensible to regard a group as an agent only if it makes sense to ascribe intentional attitudes to it. But we often also ascribe intentional attitudes to groups that are not agents. For example, we ascribe beliefs and desires, on the basis of majority polling, to various random samples and the populations they exemplify. So how do we draw the line between non-agential and agential groups? Our discussion suggests that we regard a group as an agent just when we think something is amiss if those attitudes are inconsistent, or otherwise irrational. ”

The paragraph above is important, because from the definition of Agency that I have given above, it is not immediately clear that the set of attitudes of a group agent cannot be inconsistent. But as (List and Pettit 2011) point out, this is indeed the case in their account. The importance of consistency is further reiterated (List and Pettit 2011):

(16)

“Achieving consistency is of special importance. It matters, first, when the agent’s actions respond to the attitudes involved and, sec-ond, when the actions of others respond to those attitudes: for ex-ample, because others form expectations based on the attitudes they ascribe to the agent.”

Furthermore, (List and Pettit 2011) specify some standards attitude should follow. They believe that these standards can be grouped in two categories: attitude-to-fact standards and attitude-to-action standards. According to (List and Pettit 2011), attitude-to-fact standards are standards that “rule in favour of representations that fit with how things are and rule against those that don’t”, whereas attitudetoaction standard “rule in favor of actions that are required -or at least permitted – by the agent’s representations and motivations and rule against those that aren’t”.

This concludes the introduction to the most important elements of (List and Pettit 2011)’s conceptualisation of groups and group agency. Apart from the definitions that they provide of these concept, the main lessons seem to be twofold:

1. There is extreme flexibility in the language we can use for ascribing at-titudes to group agents. This flexibility can be obtained through the recourse to metalanguages as we have hinted before, as well as through the paraphrasing mechanisms that we have discussed

2. However, there are some constraints that group agents should satisfy. Among them, probably the most important one for (List and Pettit 2011) is the consistency of their attitudes

2.3 Classical accounts: emergentism and

elimi-nativism

2.3.1 Eliminativism about group agents

The ascription of attitudes and beliefs to group agents presupposes that these ascriptions are legitimate, that is, that it makes sense to ascribe beliefs and attitudes to group agents. But that is far from being a non contentious position. In fact, it is possible to identify two traditions.

According to the first tradition, generally labelled as eliminativism, the as-criptions of beliefs and attitudes to group agents are not legitimate, and should be understood as metaphorical at best. According to the other tradition, emer-gentism, not only these ascriptions are perfectly legitimate, but they are not easily reducible to the ascriptions of beliefs of the agents the group comprises.

As summarised by (List and Pettit 2011):

“There are two extreme and starkly opposed traditions of think-ing about group agents in the literature[. . . ]. One is the emergentist

(17)

2.3. CLASSICAL ACCOUNTS: EMERGENTISM AND ELIMINATIVISM 17 tradition in law and history, sociology, and political theory that pros-pered in the late nineteenth and early twentieth centuries. The other is a debunking, eliminativist tradition, associated with analytical philosophy and economics, which arose in the nineteenth century and became mainstream from about the middle of the twentieth. The one tradition held that group agents emerge as new phenomena over and above the individuals constituting them. The other held that group agents can be eliminated from any serious inventory of the world; they are nothing but individual agents acting in concert.” In this section, I devote my attention to the eliminativit tradition.

As I have already anticipated, a typical trait of eliminativism is regarding attitude ascriptions to group agents as purely metaphorical. The idea is that while it is true that we normally ascribe beliefs to group agents, these ascription of beliefs can be paraphrased by using complex descriptions that employ exclu-sively properties ascribed to the single, individual agents the group comprises (List and Pettit 2011).

Eliminativism is sometimes referred in the scientific literature with different names. For instance, Margaret Gilbert uses the term singularism (List and Pettit 2011):

“ We [. . . ] describe any non-realist view of group agents as elimi-nativist. We could also re-describe such a view more positively as singularism, using a term introduced by Margaret Gilbert [. . . ]. Sin-gularism asserts that there are only individual agents and that any talk of group agents is either metaphorical or wrong.”

The philosopher Anthony Quinton expresses similar views under the label of ontological individualism (Quinton 1975). He characterises ontological indi-vidualism in the following way:

“Ontological individualism simply denies that any special mode of access to social facts is required for knowledge of them to be ob-tained. Since statements about social objects are equivalent and reducible to statements about individual people, it must follow that every law expressed in social terms is expressible as a law about individuals.”

He contrasts ontological individualism with its physical counterpart, that is, “physical reductionism”, which “takes theoretical entities to be less real than the common objects to which statements about such entities ultimately refer. It regards them as logical constructions or abstractions, as useful fa¸cons de parler or symbolic conveniences.” (Quinton 1975)

(Quinton 1975) admits that often we ascribe “beliefs, emotions and atti-tudes” to groups. Nonetheless, these ascriptions are to be taken as “metaphor-ical”, since “to ascribe mental predicates to a group is always an indirect way of ascribing such predicates to its members. ” (Quinton 1975)

(18)

(Epstein 2018) instead refers to eliminativism as individualism and to emer-gentism as holism:

“Individualism is the somewhat vague thesis that the social is built exclusively out of individual people. Holism is the even vaguer thesis that social entities are “sui generis”, or ontologically fundamental in some sense.”

As (List and Pettit 2011) write, some of the “support” in favour of elimi-nativism comes from its connection with a certain conception of society which is labelled methodological individualism. (List and Pettit 2011) characterise methodological individualism as “the view that good explanations of social phe-nomena should not postulate any social forces other than those that derive from the agency of individuals: that is, from their psychologically explicable responses to one another and to their natural and social environment.”

2.3.2 Emergentism about group agents

The other standard theoretical account of group agents is the emergentist ac-count.

The emergentist account is the application of the philosophical notion of emergence to the concept of group agency. In the Stanford Encyclopedia En-try on Emergent properties, the following characterisation of emergent entity is offered (O’Connor and Wong 2020):

“Emergent entities (properties or substances) ‘arise’ out of more fundamental entities and yet are ‘novel’ or ‘irreducible’ with respect to them”

Emergentism is a term that arose in the British philosophical debate (O’Connor and Wong 2020). The classical British philosophers discussing emergentism in the 19th Century were J.S. Mill and C.D. Broad (O’Connor and Wong 2020).

How could we sum up the different accounts comprised by British emergen-tism? (O’Connor and Wong 2020) offer the following synthesis:

“Common to all these theorists is a layered view of nature. The world is divided into discrete strata, with fundamental physics as the base level, followed by chemistry, biology, and psychology (and possibly sociology). To each level corresponds a special science, and the levels are arranged in terms of increasing organizational complex-ity of matter, the bottom level being the limiting case investigated by the fundamental science of physics.”

According to (O’Connor and Wong 2020), it is possible distinguish between two traditions regarding the phenomenon of emergence: an epistemological tra-dition and an ontological tratra-dition. According to the epistemological tratra-dition, which, according to (O’Connor and Wong 2020), is in the vein of Samuel Alexan-der work, emergent properties are not metaphysically primitive (O’Connor and

(19)

2.3. CLASSICAL ACCOUNTS: EMERGENTISM AND ELIMINATIVISM 19 Wong 2020). I will not explore this option further, because it is remote from the standard interpretation of emergence in the context of group agency.

Instead, it is more rewarding to focus our attention to the concept of onto-logical emergence. In (O’Connor and Wong 2020) the following characterisation is offered:

“Ontological emergentists see the physical world as entirely consti-tuted by physical structures, simple or composite. But composites are not (always) mere aggregates of the simples. There are layered strata, or levels, of objects, based on increasing complexity.” And regarding the laws, (O’Connor and Wong 2020) state:

“ Emergent laws are fundamental; they are irreducible to laws char-acterizing properties at lower levels of complexity, even given ideal information as to boundary conditions”

However, we still need a definition of emergent properties. Quoting McLaugh-lin, (O’Connor and Wong 2020) offer the following definition of emergent prop-erties:

[Emergent property] “ If P is a property of w, then P is emergent if and only if (1) P supervenes with nomological necessity, but not with logical necessity, on properties the parts of w have taken sepa-rately or in other combinations; and (2) some of the supervenience principles linking properties of the parts of w with w’s having P are fundamental laws. [. . . ] (A law L is a fundamental law if and only if it is not metaphysically necessitated by any other laws, even together with initial conditions.)”

In general, the idea of emergent property exposed in this definition does not fit easily in the framework developed by (List and Pettit 2011), because the notion of emergence about group agents considered by (List and Pettit 2011) is not strictly nomological (see the definition of Supervenience offered by (List and Pettit 2011) below). In fact, their characterisation of emergence is broader, and focuses on the supervenience relation. The notion of supervenience is very important for (List and Pettit 2011), because it is the tool that allows us to understand more clearly the difference between eliminativism and emergentism. On the one hand, in fact, emergentism has a clear non-supervenient stance, while recognising that group agents are real entities, and therefore their properties theoretically eligible for being supervenient (List and Pettit 2011):

“The difference between these two traditions can be illuminated with the help of the notion of supervenience. Taken at its word, emer-gentism holds that group agents are real entities whose properties as agents do not supervene on the contributions of their individual members.”

(20)

On the other hand, since according to eliminativism group agents are not real entities, it is impossible to make sense of the idea that their properties could supervene on anything (List and Pettit 2011):

“ If emergentism makes group agents into non-supervenient realities, eliminativism denies the reality of group agents altogether and so denies that the issue of supervenience even arises in any interesting way. Where emergentism makes group agents into hyper-realities, eliminativism makes them into non-realities. On this view, group agents are not entities of any kind and thus not even candidates for an emergent or non-supervenient status.”

In order to understand these consideration, we have to take into account the notion of supervenience:

Definition 5 (Supervenience according to (List and Pettit 2011)) Formally, we say that one set of facts, say B, ‘supervenes’ on another, say A, if and only if, necessarily, fixing the A-facts also fixes the B-facts.

Given the definition above, (List and Pettit 2011) go on to specify their supervenience thesis:

“The notion of supervenience enables us to express formally our the-sis about the relationship between what goes on at the level of the group and what goes on at the level of its members: The superve-nience thesis. The attitudes and actions of a group agent supervene on the contributions of its members.”

It is important to highlight that the supervenience thesis refers to the con-tributions of the members, and not necessarily the individual attitudes of the members (List and Pettit 2011). This means that in determining the group at-titude it is possible that the supervenience thesis might even ignore altogether the attitudes of the single agents the group comprise, and focus instead only on the individual “actions and dispositions” (List and Pettit 2011).

(List and Pettit 2011) explicitly recognise how the emergentist account arose from the original emergentism referred to organic entities, the kind of emergen-tism the British emergentist I have named above were keen on developing (List and Pettit 2011). Writing about the philosophers who applied emergentism to group agents, (List and Pettit 2011) clearly acknowledge that these philosophers were trying to draw an account of group agents similar to the one the original emergentists draw regarding organic entities, and therefore an account accord-ing to which “collective entities had a life and mind of their own” (List and Pettit 2011).

2.4 Pettit and List’s emergentism

In this section, I present (List and Pettit 2011)’s moderate emergentist account of group agency. The importance of this account for our analysis lies in the

(21)

2.4. PETTIT AND LIST’S EMERGENTISM 21 fact that I too endorse and argue for a moderate emergentist account of group agency in my thesis.

In fact, (List and Pettit 2011) describe their account with the following words:

“The moderate position in relation to these extremes [eliminativism and emergentism] is that consciousness exists but that it is superve-nient on the contributions of the neurons. In almost exact parallel, our view about group agency avoids emergentist and eliminativist ex-tremes and asserts that group agency exists but that it materializes superveniently on the contributions of group members. Moderation rules and moderation makes no big news.”

While they firmly embrace their moderate emergentist account, this does not mean that every ascription of beliefs to group agents should be taken seriously. In fact, they explicitly concede that sometimes, in certain cases, our talk of group agents is metaphorical (List and Pettit 2011).

The idea that it is the supervenience relation the building tool of group attitudes is not something peculiar to (List and Pettit 2011). For instance, (Epstein 2018) writes:

“One of the more precise ways of clarifying claims about the building of social entities is to use various forms of the supervenience relation” Why is the account developed by (List and Pettit 2011) different from emer-gentism? Because according to the emergentists there is something more to group agents than the attitudes and relations of the members the groups com-prise. As (List and Pettit 2011) write:

“If a collection of individuals needs some add-on factor to become a group agent, then it must be logically possible that a duplicate collection of agents, indiscernible in attitudes, actions, and relations, might not have that add-on factor and might thus fail to constitute a group agent. ”

Why should we accept (List and Pettit 2011)’s account? In their book, List and Pettit argue that both eliminativism and emergentism have some serious flaws.

Regarding emergentism, they write (List and Pettit 2011):

“The opposition to any reification of groups was supported in An-glophone twentieth-century philosophy and related disciplines by the association between the emergentist, organicist tradition and the ap-pearance of fascism”

However, the political connotations are definitely not the only ground to reject emergence. Another objection to emergence comes from the alleged lack of intelligibility of the idea of an emergent force (List and Pettit 2011):

(22)

“It is hard to make sense of the idea that group agency requires not just the coordination of individual members but also the presence of a mysterious, individualistically inaccessible force.”

What are List and Pettit’s arguments against eliminativism?

The strongest argument they offer is that there is no easy translation from the individual attitudes to group attitudes (as we will see in the section Pettit and List group agency results) (List and Pettit 2011). But if there is no easy translation, we are justified in seeing group attitudes as metaphysically legitimate entities. In the words of (List and Pettit 2011):

“The formal arguments we have presented in this chapter show that there is no easy translation between some of the things we can know about group agents and the things we know about their members.” “While the agency achieved by a group supervenes on the contribu-tions of its members – while it is not ontologically autonomous – it is autonomous in another, related sense. The agency of the group relates in such a complex way to the agency of individuals that we have little chance of tracking the dispositions of the group agent, and of interacting with it as an agent to contest or interrogate, persuade or coerce, if we conceptualize its doings at the individual level.”

Sometimes, if we want to have a coherent agent, we should be even willing to ascribe an attitude to a group agent which is not shared by any of the individual agents the group is comprised of. This is further proof of a lack of an easy translation from individual attitudes to group attitudes, and therefore further proof that eliminativism can be seriously challenged (List and Pettit 2011):

“Drawing on recent work in the theory of judgement aggregation, we show that the attitudes – beliefs and desires – we need to ascribe to any group that meets the conditions of agency are not readily reducible to the attitudes of individuals. A group agent, like an in-dividual, has to satisfy certain minimal requirements of consistency, and to avoid breaching them, it may be constrained to form atti-tudes on some issues that depart from the attiatti-tudes of its members. In the limit, it may even have to form an attitude on some issue that all its members individually reject. Since the ascription of attitudes is needed to make sense of such a group’s behavior, the lack of an easy translation of group-level attitudes into individual-level ones requires us to recognize the existence of group agents in making an inventory of the social world.”

Furthermore, according to (List and Pettit 2011), some of the common em-ployed arguments used in favour of eliminativism are less strong than they seem to be at first glance. In fact, while eliminativism gains a lot of support from its

(23)

2.5. AN ANALYSIS OF THE SUPERVENIENCE RELATION 23 compatibility with methodological individualism, it is still the case, as (List and Pettit 2011) point out, that eliminativism is a sufficient condition for method-ological individualism, but not a necessary one.

The fact that their theory does not posit psychological mysterious forces is the proof that their account is compatible with methodological individualism (List and Pettit 2011). In fact, in their work, (List and Pettit 2011) reject elim-inativism, because they believe that at least in some cases we can legitimately ascribe “ascriptions of agency ”to group agents. At the same time, their account rejects strong emergentism too, because there is no recourse to “psychologically mysterious forces”, since, as they write, “the agency of group agents depends wholly on the organisation and behaviour of individual members”.

2.5 An analysis of the supervenience relation

As we have seen, the supervenience relation plays a key role in the moderate emergentist account of (List and Pettit 2011). However, they do not delve deep into the metaphysics of supervenience. I believe that given the importance of the notion of supervenience for articulating a moderate emergentist account, such a notion is worthy of further analysis in the context of this thesis.

It is important to highlight that supervenience is considered a technical term of art. According to some authors the word originates with the British emer-gentists of the early 20th century; however, there is a wide discussion regarding where the term originated (B. McLaughlin and Bennett 2018).

In (List and Pettit 2011), the supervenience relation is not so strictly defined. As we have seen, (List and Pettit 2011) define the supervenience relation only in broad terms.

How could we get a more precise definition? In this section, I will focus on the Stanford Encyclopedia of Philosophy entry on supervenience (B. McLaughlin and Bennett 2018) in order to survey possible answers to this question.

The broad definition of supervenience offered in (B. McLaughlin and Bennett 2018) is not too different from the definition offered by (List and Pettit 2011):

Definition 6 (Broad Supervenience according to (B. McLaughlin and Bennett 2018)) A set of properties A supervenes upon another set B just in case no two things

can differ with respect to A-properties without also differing with respect to their B-properties

When discussing the supervenience relation, we must keep in mind that a clear feature of the supervenience relation is that it is a modal relation. But there are several ways in which we can specify the modal relation:

“ the ‘cannot’ in “there cannot be an A-difference without a B-difference” comes in different strengths. For example, it can mean “cannot as a matter of logic,” or it can mean “cannot consistently with the laws of nature”.”

(24)

In their analysis, Brian McLaughlin and Karen Bennett distinguish between: 1. Individual and global supervenience

2. Weak and strong supervenience

Let’s turn out attention to individual supervenience. In terms of possible worlds, we have the following definitions:

Definition 7 (Weak supervenience with possible worlds) A-properties weakly supervene on B-properties if and only if for any possible world w and any in-dividuals x and y in w, if x and y are B-indiscernible in w, then they are A-indiscernible in w.

Definition 8 (Strong supervenience with possible worlds) A-properties strongly supervene on B-properties if and only if for any possible worlds w1 and w2 and

any individuals x in w1 and y in w2, if x in w1 is B-indiscernible from y in w2, then x in w1 is A-indiscernible from y in w2.

We can also consider the definitions in terms of modal operators:

Definition 9 (Weak supervenience with modal operators) A weakly su-pervenesm on B if and only if necessarily, if anything x has some property F in A, then there is at least one property G in B such that x has G, and everything that has G has F

Definition 10 (Strong supervenience with modal operators) A strongly supervenesm on B if and only if necessarily, if anything x has some property F in A, then there is at least one property G in B such x has G, and necessarily everything that has G has F

At first the possible worlds and modal definitions were thought to be equiv-alent to the aforementioned ones, but today there is much more debate, and the consensus tends to be that they are different (B. McLaughlin and Bennett 2018).

Let’s now turn our attention to global supervenience. We have the following definitions:

Definition 11 (Weak global supervenience) A-properties weakly globally su-pervene on B-properties iff for any worlds w1 and w2, if there is a B-preserving isomorphism between w1 and w2, then there is an A-preserving isomorphism between them.

Definition 12 (Strong global supervenience) A-properties strongly globally supervene on B-properties iff for any worlds w1 and w2, every B-preserving iso-morphism between w1 and w2 is an A-preserving isoiso-morphism between them.

Of course, such a definition requires the definition of preserving isomorphism, which again is provided in the article:

(25)

2.5. AN ANALYSIS OF THE SUPERVENIENCE RELATION 25 Definition 13 (Preserving isomorphism) An isomorphism I between the in-habitants of worlds w1 and w2 preserves f-properties if and only if for any x in w1, x has a f-property P in w1 just in case the image of x under I (the individual to which I maps x) has P in w2.

Another theoretical option is constituted by the definition of supervenience for multiple domains:

Definition 14 (Multiple domain supervenience) Let D1 and D2 be non-empty domains of individuals, R be a relation between D1 and D2, and R|x be the subset of D2 to which x is R-related. We can define the following notions:

(A, D1) weakly multiple domain supervenes on (B, D2) relative to R just in case necessarily for any x and y in D1, if R|x and R|y are B-indiscernible, then x and y are A-indiscernible.

(A, D1) strongly multiple domain supervenes on (B, D2) relative to relation R just in case for any x and y in D1 and any worlds w1 and w2, if R|x in w1 is B-indiscernible from R|y in w2, x in w1 is A-indiscernible from y in w2. (Kim 1998)

Several other definitions of supervenience are offered in (B. McLaughlin and Bennett 2018), but I do not believe that they are relevant in the context of this thesis. I do not delve too much deeper into the debate regarding the way supervenience relation differs from other relations like entailment and grounding. I just consider the two most important arguments advanced by (B. McLaughlin and Bennett 2018) in this regard:

“A second way to see that supervenience is not identical to either grounding or ontological dependence is to note that the latter two relations are widely (though not universally) thought to be irreflexive and asymmetrical.”

“A third way to see that supervenience is not the same as either grounding or ontological dependence is that the following condition-als are fcondition-alse:

if A supervenes on B, B grounds A if A supervenes on B, A onto-logically depends on B”

A question that is sometimes asked is whether supervenience is ontologically innocent, that is, whether the properties which supervene can be considered ontologically autonomous. As they write (B. McLaughlin and Bennett 2018):

“The issues about entailment and reduction are related to a ques-tion about whether supervenience with metaphysical necessity is on-tologically innocent—whether the A-properties are anything “over and above” the B-properties.

Some think that supervenience–at least when accompanied by ground-ing [. . . ] is indeed ontologically innocent in this sense. After all, if the A-properties supervene with metaphysical necessity on the B-properties, then they come along automatically given the B-properties.”

(26)

2.6 Pettit and List group agency results

2.6.1 A definition of joint intention

Before showing some formal results from (List and Pettit 2011) regarding group agency, I have to specify that the account of group agency List and Pettit have in mind when drawing their results is an account of jointly intentional group agents. However, according to (List and Pettit 2011) it is possible for group agents not to be jointly intentional. They identify at least two ways in which non-jointly intentional group agents might emerge. The first one “would involve a process of natural or cultural evolution in which members are selected for possessing traits that lead them to act as required for group agency”, whereas the second one “would involve one or several organizational designers co-opting others into a structure underpinning group agency, without making them aware of their agency at the group level”.

(List and Pettit 2011) offer the following definition of joint intention: Definition 15 (Joint intention) We say that a collection of individuals ‘jointly intend’ to promote a particular goal if four conditions are met:

1. Shared goal. They each intend that they, the members of a more or less salient collection, together promote the given goal.

2. Individual contribution. They each intend to do their allotted part in a more or less salient plan for achieving that goal.

3. Interdependence. They each form these intentions at least partly because of believing that the others form such intentions too.

4. Common awareness. This is all a matter of common awareness, with each believing that the first three conditions are met, each believing that others believe this, and so on

A second important feature in the account of (List and Pettit 2011) lies in the fact that they restrict their attention to binary attitudes (although, as we have seen in previous sections, it is possible to translate non-binary attitudes into binary ones):

“One assumption underlying our account of group agency – specif-ically, jointly intentional group agency – must be made explicit at this stage. We assume that, whatever the group’s organizational structure is, it must lead the group to form binary attitudes, not attitudes that come in degrees of strength.”

Why do (List and Pettit 2011) restrict their attention to jointly intentional groups? A possible answer could be that according to some authors it is joint intention that allows group to be truly social objects. For instance, (Epstein 2018) writes:

(27)

2.6. PETTIT AND LIST GROUP AGENCY RESULTS 27 “Margaret Gilbert, for instance, hesitates to attribute “sociality” to crowds: sociality, she argues, arises from norms and commitments, which many crowds lack. According to Gilbert, it is with joint com-mitment that a group is genuinely social.”

A wide literature exists on joint actions and group intentions 1. There are several theories regarding how groups end up forming intentions. (Epstein 2018) observe that according to Bratman “shared intentions” ca be analysed “in terms of the common knowledge of group members”. As we will see, the idea that common knowledge features in group belief ascriptions plays an important role in the last chapter of my thesis. In (Epstein 2018), a distinction used in the scientific literature in order to analyse group agent attitudes is mentioned: that is, the distinction between “I-attitudes” and “we-attitudes”. The first ones are attitude that relate to the individual agents without reference to the other members of the group, in contrast to the second ones. However, this distinction will not feature in my research.

2.6.2 Some formal results and constraints

The first problem discussed by (List and Pettit 2011) in order to highlight how aggregating the individual attitudes of the agents in a group is problematic is the doctrinal paradox (List and Pettit 2011):

“ Our first example is the ‘doctrinal paradox’[. . . ] Suppose a three-member court has to make a decision in a breach-of-contract case. The court seeks to make judgments on the following propositions:

1. The defendant was contractually obliged not to do a certain action (the first premise).

2. The defendant did that action (the second premise).

3. The defendant is liable for breach of contract (the conclusion). According to legal doctrine, obligation and action are jointly nec-essary and sufficient for liability; that is, the conclusion is true if and only if both premises are true. Suppose, [. . . ] judge 1 believes both premises to be true; judge 2 believes the first but not the sec-ond premise to be true; and judge 3 believes the secsec-ond but not the first to be true. Then each premise is accepted by a majority of judges, yet only a minority, that is, judge 1, individually considers the defendant liable.”

As the paradox clearly shows, how to aggregate the individual attitudes into a group attitude is a difficult question2_{. And the doctrinal paradox is far from}

being the only puzzling result. Some equally puzzling formal results can be proved. (List and Pettit 2011) write:

1_{For instance, see (Tuomela 2007)}

(28)

“Formally, an individual’s ‘attitudes’ towards the propositions on the agenda are modeled as an assignment of Yes’s and No’s to these propositions, where ‘Yes’ stands for a positive attitude and ‘No’ for a negative one. A combination of attitudes across all individuals in the group is called a ‘profile’.[. . . ]

To move from a profile of individual attitudes to group-level atti-tudes, the group requires an ‘aggregation function’. [. . . ] An aggre-gation function is a mapping that assigns to each profile of individual attitudes towards the propositions on the agenda (assuming the pro-file is admissible, more on this later) the collective attitudes towards these propositions, which are also modeled as an assignment of Yes’s and No’s to them.”

Let’s now consider this apparently obvious set of requirements for the ag-gregation function (List and Pettit 2011):

“

1. Universal domain. The aggregation function admits as input any possible profile of individual attitudes towards the propo-sitions on the agenda, assuming that individual attitudes are consistent and complete.

2. Collective rationality. The aggregation function produces as output consistent and complete group attitudes towards the propositions on the agenda.

3. Anonymity. All individuals’ attitudes are given equal weight in determining the group attitudes. Formally, the aggregation function is invariant under permutations of any given profile of individual attitudes.

4. Systematicity. The group attitude on each proposition depends only on the individuals’ attitudes towards it, not on their at-titudes towards other propositions, and the pattern of depen-dence between individual and collective attitudes is the same for all propositions.

”

Then the following theorem can be proved:

Theorem 1 There exists no aggregation function satisfying universal domain, collective rationality, anonymity, and systematicity (List and Pettit 2002)

Interestingly, it is possible to satisfy any three of the conditions simultane-ously. Therefore, the question shifts to which feature should be ditched. Ac-cording to (List and Pettit 2011), the best option is renouncing systematicity. The discussion regarding the problems we run into when we aggregate indi-vidual attitudes into group attitudes is further expanded in Chapter 5, where some theses regarding the way we should translate individual attitudes to group attitudes are exposed.

(29)

Chapter 3

What are groups? What

are group agents?

3.1 What are groups? What are group agents?

3.1.1 Definitions of group and group agents

In this chapter, I am going to discuss what groups and group agents are. I will argue in favour of a definition of group that differs from the one offered in (List and Pettit 2011). Furthermore, I will argue that group agents need not be coherent, while adopting a definition of coherence that identifies coherence with logical consistency.

The term group is employed in several different contexts: its scope is uni-versal. It is employed in mathematics as it is employed in sociology (Epstein 2018). In the context of this thesis, I am interested into groups of agents, and not groups of any other kind.

Still, we have to come up with a definition of group in this context. In this thesis, I will assume the notion of collection as primitive, similarly to what is done in (List and Pettit 2011).

We can start by considering what definitely does not constitute a group: a random collection of agents. If we consider a collection made of me, the people who were in Times Square, NY on New Year’s Eve 2020 and all the people who were wearing blue socks last Monday, it is very difficult to consider this collection a group.

What is then that distinguishes a group from a collection? (List and Pettit 2011) argues that the differentiating feature is that groups can survive change of membership, whereas collections cannot. Therefore, as we have seen, they argue in favour of the following definition: a group is a collection that can survive change of membership.

This tentative definition has certainly some virtues. For instance, it allows us to understand why the collection that I came up with before, that is, the

(30)

collection made of me, the people who were in Times Square, NY on New Year’s Eve 2020 and all the people who were wearing blue socks last Monday is not a group. In fact, a group consisting of this collection would not survive any change of membership. As soon as a member of this collection is excluded, we would not be talking about the same group.

While the definition offered by (List and Pettit 2011) definitely looks promis-ing, I do not believe that it is adequate. In fact, I believe that it captures too much, that is, too many things are comprised in this definition that are not groups.

As soon as we can find a counterexample to (List and Pettit 2011)’s defi-nition, we can conclude that the definition of group offered in (List and Pettit 2011) is inadequate, because it ends up capturing too much.

What could constitute such a counterexample? I believe the following one qualifies as a counterexample:

Example 1 Consider the collection that comprises the following agents: all the agents whose first name is ‘Mark’. This collection can definitely survive change of membership, as the set of people whose first name is ‘Mark’ changes everyday, with newborns and old people dying.

However, our intuition tells us this is not really a group. After all, apart from sharing the same name, what do all these people really have in common? Most of them would not know each other. It is difficult to say they have much in common apart from an arbitrary property like having the same first name. If we want to pin down what they seem to be lacking, this is the fact that they do not think of themselves as a unitary agent.

Of course, it could still be possible to abstract from the individual prefer-ences/beliefs of the agents to try to ascribe preferences and beliefs to the entire group. Consider, in fact, the following refined example: consider the group singled out by the property having ‘Mark’ as your first name and being a cit-izen of the USA. Then we could start enquiring into the political preferences of the agents that are comprised by this definition. How many of them lean Republican? How many of them lean Democrat ? How many of them do not have any particular political leaning? For the sake of our example, let’s assume that a vast majority of them (let’s say, 80% of them) believe that a Republican president would be the best fit for the country. Would we consider this enough ground to ascribe to this group the corresponding belief? I believe the answer to such a question is no. Yes, it is true, the opinion shared by the vast majority of the members of this group is indeed that the Republican president would be better, but this does not seem enough to ascribe such a belief to the group as a whole. The reason why this is the case is that we do not perceive the group to constitute an unitary element of any kind. We do not perceive it to be a unitary agent of any sort because the agents that this definition comprises do not have expressed any intention, neither explicit nor implicit to be part of that collection. They are comprised just because they happen to share a certain property, but not because they believe that sharing the same property is a

(31)

rea-3.1. WHAT ARE GROUPS? WHAT ARE GROUP AGENTS? 31 son to coordinate with other members, even only implicitly, to express common attitudes and reach common goals.

Which lesson could we derive from this example? Which could be a more promising definition of group? I believe that such a definition could be the following one:

Definition 16 Group

A group is a collection of agents that recognise that the collection they belong to constitutes an agent

Let’s try to see how my definition works in the example above. I argue that such a definition would rule out the above example as a group. In fact, the people comprised in the collection of agent whose first name is ‘Mark’ and who live in the US do not see the collection they belong to as constituting an agent. That is the reason why such a collection does not constitute a group.

On the other hand, it is easy to understand why, according to my definition, a corporation does constitute a group instead. In fact, if we consider the members of a corporation, we can easily see that a corporation meets the demands of my definition of group: the agents of a corporation recognise that the corporation they belong to constitutes something more than the aggregation of its members. In fact, it is an agent in its own right.

It would be interesting to square these considerations against the theoretical account laid out by Pettit in (List and Pettit 2011). (List and Pettit 2011) develop the following account: they distinguish between mere collections and groups. Regarding groups, they distinguish between mere groups and group agents. Again, regarding group agents, they distinguish between non-jointly intentional and jointly intentional.

How does this taxonomy compare with mine? It is not so simple to answer this interesting question. The first, crucial question would be the following one: according to my definition, are all groups group agents? In order to tackle this question, we need to provide a definition of group agent. I propose the following definition:

Definition 17 (Group agent) A group agent is a group to which it is possible to attribute attitudes, such as beliefs and knowledge

The example that I have provided regarding the corporation is clearly an example where the group constitutes a group agent. In fact, we can easily attribute beliefs and knowledge to a corporation. However, this does not answer our question. The fact that a group like a corporation is also a group agent of course does not imply by itself that all groups are group agents, but only that some are.

We can move from the particular to the universal by analysing the definition of group that I have advanced previously. According to my definition, a group is a collection of agents that recognise that the collection they belong to consti-tutes an agent. But if that is the case, how can we have an agent without the possibility of ascribing to it knowledge or beliefs? It is difficult to envisage such

(32)

a possibility. Arguably, being an agent requires having knowledge and beliefs. But if that is the case, we can conclude the following : all groups are group agents.

Someone might object that I have gone too fast in my argument. In fact, my argument relied heavily on the premise that an agent cannot fail to have beliefs and knowledge. Are we overlooking possible objections to this premise? Could the case be made that it is possible to have agents without beliefs and knowledge?

It seems implausible to answer the affirmative. What would be an agent without beliefs and knowledge? The word agent is associated with the verb ‘to act’. The agent is the one who acts. Now, the contrarian might say: it is possible to act without knowing and without believing. How could we answer this objection? I might point out that while indeed it is true that it is possible to act without knowing and without believing, my ideal concept of an agent is the one of somebody who acts in response to his knowledge and his beliefs. That is the reason why I believe that the inference that I have laid out in the previous paragraphs is justified, and we can safely conclude that all groups are group agents.

Once we have established this, how does my account compare with the tax-onomy developed by (List and Pettit 2011)? We can say that this does not affect the distinction between mere collections and groups, which still stands. Instead, it does collapse the distinction between group and group agent : if all groups are group agents, then there is no reason to contemplate this distinction in my taxonomy. What about the other distinction drawn by (List and Pettit 2011), that is to say, what about the distinction between jointly intentional and non-jointly intentional group agents? The results that we have articulated so far do not rule out such a distinction.

An objection that purports to rule out this distinction could have the follow-ing form. Let’s recall that my definition of group is the followfollow-ing one: A group is a collection of agents that recognise that the collection they belong to constitutes an agent. Is this enough to guarantee that all the beliefs and knowledge that we attribute to the group agents are the result of a joint intention? The reasoning could go along the following lines. If the group is a collection characterised by the fact the the agents it comprises see it as an agent, the argument goes, this could only mean that insofar the group is a group agent, all the beliefs and, in general, the attitudes that we ascribe to it emerge due to the recognition from the agents constituting the group. And in this context, what is recognition, if not joint intention?

This argument hinges strongly on the idea that recognition is joint intention. Of course, if we could advance a clear-cut example of a non-jointly intentional group agent this would suffice to show that the distinction is salient. Unfortu-nately, the corporation example does not qualify as an instance of non-jointly intentional group agent. However, I will not develop this issue further, and it will be the job of future research to decide whether the argument I have considered is sound after all, or if a clear-cut counterexample can be advanced.

(33)

3.1. WHAT ARE GROUPS? WHAT ARE GROUP AGENTS? 33

3.1.2 A counterexample to the consistency requirement

So far, I have discussed what could be the definitions of group and group agent. But it is still left to discuss, as stated in the introduction, why and how my definition of group agency differs from the one developed by (List and Pettit 2011). Let’s recall that my definition of group agency postulates that a group agent is a group to which it is possible to attribute attitudes such as beliefs and knowledge. By my definition, this is enough for the group to be a group agent. However, there is no specification regarding further qualities that the set of beliefs should have. Consider the following definition of consistency of belief sets:

Definition 18 (Consistency of belief sets) A set of beliefs is consistent if and only if the classical deductive closure of the set of propositions believed by the agent does not contain a proposition φ and a proposition ¬φ 1

Now, the definition of group agent that I have provided in the previous paragraphs does not posit any requirement in terms of consistency. It is perfectly possible to attribute a belief p and a belief non − p to our hypothetical group agent. Therefore, our hypothetical group agent can be inconsistent.

However, not all authors believe that it is possible for group agents to be inconsistent. For instance, according to (List and Pettit 2011) a group agent should exhibit what they label robust rationality. It is a bit difficult to in-troduce the concept of robust rationality, since I have not yet inin-troduced the concept of supervenience. If I am allowed to anticipate the discussion regarding supervenience, then I can give the following definition of robust rationality : Definition 19 A group agent exhibits robust rationality if and only if the su-pervenience relation determines consistent and complete group attitudes on the relevant propositions for any possible profile of consistent and complete member attitudes on these propositions.

Now, of course some parts of this definition remain obscure, because we have not explored the supervenience relation yet. But what interests us is that if a group agent exhibits robust rationality, such a group agent cannot be inconsistent. Therefore, if we claim that a group agent can be inconsistent, we are also claiming that a group agent can fail to exhibit robust rationality.

The natural question arises: how can a group agent fail to be consistent? It is not so easy to give an answer to this puzzling question. In fact, especially if we collapse the joint-intentional group agent to the group agent in general, the question would be even more difficult to answer. If group agent are jointly intentional, how is it possible to ascribe contradictory beliefs? The single agents

1_{It is important to pay attention to the subtlety of this definition. If the belief set of an} agent comprises exclusively the belief p, the belief q, and the belief p → ¬q, it does not matter that in the the belief set the belief ¬q is not comprised. The set of beliefs is non-consistent, because the classical deductive closure of the set of propositions believed by the agent contains q and ¬q.

(34)

that form the group do so, as the definition stipulates, by recognising that they belong to a collection to which it is possible to attribute knowledge and beliefs. But this does not force the agents to attribute a coherent set of beliefs to the group agent. Consider the following example: a football team.

Example 2 A football team is definitely a group and a group agent. Consider, for the sake of our example, a football team with no captain and all peers, like a football team made of friends. In such a scenario, it does not make any sense to prioritise the beliefs of any agent over the ones of some other agent of the group, since they are all peers.

Furthermore, it seems obvious to accept the following condition regarding the ascription of beliefs to the group: prima facie, as long as the majority of the group believes that p, the group believes that p.

Now, let’s imagine that the members of the group have not discussed which strategies to pursue for winning the next football match. Let’s assume that the group consists of 10 members. Therefore, as long as 6 members believe a certain proposition p, we can conclude that a majority of the group believes in proposi-tion p. Now, let’s consider the following beliefs. Let p be the proposiproposi-tion ‘The team should follow a defensive strategy’, let q be the proposition ‘The goalkeeper should be member A’. Of course, given these constraints, we have that p → q means If the team should follow a defensive strategy, the goalkeeper should be member A. Now, assume the following situation. We have 6 members who be-lieve that p and 6 members who bebe-lieve that q is false. The two sets are distinct, apart from one member who believes both p and q are true. Furthermore, except for this member, everybody else believes p → q. In such a scenario, if we follow the criterion that we have laid out in the previous paragraphs, we have that the group believes that p, the group believes that non − q, and the group believes that p → q. Therefore, the set of beliefs of the group agent is inconsistent.

What to do? Someone could say one’s modus ponens is one’s modus tollens. It is possible to say that what we have come up with is actually a counterexample to the assumption that in this instance we could attribute a belief to the group as long as the majority of the group members believed that p. The objector might point out that in order to attribute a belief to a group, such a belief should be the object of a deliberation, and in the example I have come up with, this is definitely not the case. Again, the objector might say, proceeding along these lines, that in this case we should abstain from attributing beliefs to the group. However, I do not believe that this the case. Even though the beliefs that we attribute to the group are not the result of a deliberative process, yet it makes sense to ascribe these beliefs, given that they reflect the reality of what the agents the group is comprised of believe. It seems preposterous to rule out the possibility of ascribing beliefs to these groups just because it shows that robust rationality fails to hold in certain circumstances. On the contrary, the example I contrived serves as a sign that it is the robust rationality thesis to be wrong. After all, the agents know that they belong to a group (in this case, the football team). In absence of a mechanism by which they discuss the

A moderate emergentist account of group agency