C – A Platform to Design, Implement and

Execute Social Computations

Flavio S. Correa da Silva

, David S. Robertson

and Wamberto W. Vasconcelos

Dept. of Computer Science, University of Sao Paulo, Sao Paulo, Brazil

School of Informatics, University of Edinburgh, Edinburgh, U.K.

Dept. Computing Science, University of Aberdeen, Aberdeen, U.K.

Keywords:

Interaction Models, Social Computation Models, Social Interaction Protocols.

Abstract:

Social computers have been characterised as goal oriented complex systems comprised of humans as well as

computational devices. Such systems can be found in natura in a variety of scenarios, as well as designed to

tackle speciﬁc issues of social and economic relevance. In the present article we introduce the Lightweight

Situated Social Calculus (LS

C) as a language to design executable speciﬁcations of interaction protocols for

social computations. Additionally, we describe a platform to process these speciﬁcations, giving them a com-

putational realisation. We argue that LS

C can be used to design, implement and execute social computations.

1 INTRODUCTION

Social computers have been characterised as complex

systems that harness the innate problem solving, ac-

tion and information gathering powers of humans and

the environments in which they live to tackle large

scale social and economic problems (Giunchiglia and

Robertson, 2010):

• The “hardware” of a social computer is supplied

by humans (taken as individuals as well as collec-

tively in the form of human-powered institutions)

and the environment where these humans live, in-

cluding all relevant artifacts which can be natural

or man-made, as well as computational devices.

• The “software” of a social computer is comprised

of human capabilities, organisational and social

rules and norms, social conventions, as well as

computer software.

• The “algorithms” of social computation are de-

ﬁned by socially accepted goals and correspond-

ing actions which can be taken to achieve local as

well as global goals.

• Finally, the “processing” of algorithms in so-

cial computers are collective, decentralised, goal-

oriented actions whose emergent results can be

iteratively evaluated and steered towards active

goals.

Evidently, one cannot program a social computer

the way conventional computational devices are pro-

grammed. Social computers are evolving social sys-

tems, whose components (i.e. their “hardware”, “soft-

ware”, “algorithms” and “processing”) are dynami-

cally and evolutionarily designed together with their

goals and available resources. The analysis and de-

sign of social computers require novel methodologi-

cal practices, blending existing techniques and experi-

ences from applied social sciences and computational

sciences (Correa da Silva et al., 2013).

In order to design, implement and continuously

monitor and steer the behaviour of social computers,

specialised languages are required to build speciﬁca-

tions, and corresponding computational platforms are

required to support, manage and provide a computa-

tional realisation of social computations. An essential

aspect to be represented in such languages is inter-

action between components of social computers, so

that the internal behaviour of these components can

be abstracted and the resulting systems can be anal-

ysed as a whole. Additionally, since these languages

should be used to communicate speciﬁcations as well

as processing results to participants in social comput-

ers (i.e. humans who behave as components in so-

cial computers), they should be concise and simple to

understand. Finally, in order to build social comput-

ers whose behaviour can be veriﬁed with respect to

desired requirements and attributes, these languages

196

Correa Da Silva F., Robertson D. and Vasconcelos W..

LS2C – A Platform to Design, Implement and Execute Social Computations.

DOI: 10.5220/0005237101960203

In Proceedings of the International Conference on Agents and Artiﬁcial Intelligence (ICAART-2015), pages 196-203

ISBN: 978-989-758-074-1

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

should have a formal underpinning and the corre-

sponding speciﬁcations and processing results should

be formally veriﬁable.

In the present article we present the ongoing de-

velopment of a language and a platform for social

computers. The proposed language and companion

computational platform – coined the Lightweight Sit-

uated Social Calculus (LS

C) is a fusion of two previ-

ously existing languages, respectively the Lightweight

Social Calculus (LSC) and the JamSession platform.

In section 2 we present some related work and the

preliminary concepts that have guided the develop-

ment of the LS

C language and platform. In section

3 we introduce in detail the LS

C language. In section

4 we brieﬂy describe the platform in which this lan-

guage shall be used. In section 5 we illustrate how this

platform can be used in practice. Finally, in section

6 we present some conclusions and proposed future

work.

2 RELATED WORK AND

PRELIMINARY CONCEPTS

The LS

C platform is a fusion of the Lightweight

Social Calculus (LSC) and the JamSession platform.

LSC, in turn, is an extension of the Lightweight Coor-

dination Calculus (LCC). In the following paragraphs

we brieﬂy introduce these languages and platforms.

The LCC is an executable speciﬁcation language

grounded on the notions of process algebras and ini-

tially proposed for the speciﬁcation and processing

of interaction models for distributed software compo-

nents (Robertson, 2004). It has been extended in a

variety of ways, e.g. for contextual reasoning about

distributed software systems (Sindhu et al., 2006), for

the speciﬁcation and execution of choreographies for

web services (Bai et al., 2012) and, more recently,

for the speciﬁcation of social computers, under the

name of Lightweight Social Calculus – LSC (Murray-

Rust and Robertson, 2014). It has also been success-

fully implemented using the logic programming lan-

guage Prolog, the object oriented programming lan-

guage Java and the object-functional programming

language Scala.

LCC and its variations – particularly LSC – ful-

ﬁll most of the requirements to be a language for the

speciﬁcation, implementation and processing of so-

cial computations. LSC is a compact formal language

that can be used to specify and to mediate ongoing so-

cial interaction protocols. The syntax of LSC (as well

as all other variations of LCC), however, can lead to

lengthy speciﬁcations which can be difﬁcult for hu-

man reading and understanding. Moreover, the exten-

sion of LCC to manage contexts (coined Ambient LCC

(Sindhu et al., 2006)) departs from the lightweight

approach and becomes more complex than the other

variations of LCC, resulting in a not so concise and

effective platform for the speciﬁcation and execution

of interaction protocols by human system designers.

The explicit management of contexts can be a

powerful technique to help in the analysis and de-

sign of social interactions, given that many of these

interactions are context-dependent (e.g. business ne-

gotiations must occur in adequately equipped meet-

ing rooms, to ensure privacy and the availability of

required communication resources; healthcare must

occur in hospitals and clinics, to ensure the avail-

ability of required specialised equipment and person-

nel; bank transactions must occur over the appropri-

ate counters; the automated interactions among com-

municating portable devices in an Internet of Things

scenario must be context sensitive to ensure privacy

and reliability of interactions; and so on). Therefore,

a useful feature in a language for social computers is

the explicit representation and management of con-

texts, which can be abstracted as locations in which

certain interactions are allowed to occur.

The JamSession platform is a language developed

for purposes similar to those of LSC. It was initially

conceived as an executable speciﬁcation language to

manage the interactions between human controlled

and synthetic characters in Second Life-style virtual

worlds and multiplayer computer games, and later

employed to mediate business interactions between

organisations in cross-organisational workﬂows (Cor-

rea da Silva, 2011; Correa da Silva et al., 2012). A

simpliﬁed prototype of JamSession has been imple-

mented in Prolog, and a cloud-based prototype of

JamSession has been implemented using the func-

tional language F], based on which sample demos of

applications have been developed (David, 2012). The

fundamental concept in JamSession is the notion of

situated interaction protocols, which determine how

and where agents can interact with each other and

with the environment. The semantics of situated in-

teraction protocols can be formally characterised in

terms of Nested Petri Nets, which are an extension

of coloured Petri nets to handle recursion (Fernan-

dez Venero and Correa da Silva, 2013a). Nested Petri

Nets, in turn, can be translated into the speciﬁca-

tion language Promela and veriﬁed using the model

checker SPIN with respecto to properties of their op-

erational behaviour (such as liveness and termina-

tion) (Fernandez Venero and Correa da Silva, 2013b).

There have been initiatives by other authors to

analyse social interactions based on formal languages

capable of capturing the dynamics of interactions,

LS2C-APlatformtoDesign,ImplementandExecuteSocialComputations

197

in many cases grounded on the notions of dynamic

modal logics, preference logics and public announce-

ment logics (Christoff and Hansen, 2013; Hansen,

2014; Seligman et al., 2011; Zhen and Seligman,

2011). Our work distinguishes from these initiatives

in two relevant senses:

1. We focus on systems design as well as analysis,

whereas those initiatives focus primarily on anal-

ysis of existing social networks grounded on for-

mal theories.

2. Since we are interested in the design of sys-

tems for goal-oriented social interactions, we have

taken into account scalability and computational

performance issues, as well as interaction design

issues. Previously existing initiatives have mostly

focused on theoretical issues, accounting for com-

putational and system level performance as sec-

ondary issues. Scalability in LS

C shall be en-

sured by the appropriate use of asynchronous state

management based on Linda-style tuple spaces

(Gelernter, 1985), following the implementation

practices used in JamSession.

3 THE LS

C LANGUAGE

The Lightweight Situated Social Calculus (LS

C) is

based on the notion of situated social interactions.

A situated social interaction is comprised of actions

which are permitted to occur if performed by spe-

ciﬁc agents at speciﬁc locations, together with mes-

sages exchanged among agents in order to enable and

trigger further actions, and with migrations of agents

across locations. In LS

C, locations are an abstraction

used to represent a variety of concepts, such as:

• Actual physical locations, e.g. the counter in a

bank where ﬁnancial transactions are permitted to

occur.

• Contextual information, e.g. characterising the

collective acceptance of interaction protocols by

agents in a business transaction (buy, sell, le-

gal procedures, and so on) so that it is common

knowledge that the transaction can be carried out

as long as all actions in all protocols – the “set-

ting” for the transaction – are fulﬁlled.

Locations are represented as nodes in a directed

graph, in which edges represent accessibility rela-

tions, characterising allowed transitions between lo-

cations or contexts. We denote the set of nodes in a

graph of locations as S = {s

,...,s

}

We abuse notation and also refer to the graph of loca-

tions itself as S.

Each location can host an unlimited number of

agents. An agent is capable of:

• Moving between directly connected locations.

• Performing allowed actions while in speciﬁc loca-

tions.

• Reading, writing or deleting messages in loca-

tions.

We have a set of agents A = {a

,...,a

} whose

behaviour is constrained and determined by each role

that they adopt, according to each location to which

they move. Actions and message types are available

only to agents bearing speciﬁc roles at speciﬁc loca-

tions. The positioning of agents in locations is the

way to control the processing of algorithms in social

computers represented using LS

Actions can be enabled by and inﬂuence or trans-

form objects that can be found in the environment. We

have a set of objects B = {b

,...,b

}, which are sub-

ject to the actions of agents. Objects can be physical

objects as well as their digital counterparts.

We build mappings pointing to agents and objects,

so that we can refer to them indirectly through built

connections between them (such as FatherOf(X) to re-

fer to an agent by naming another agent). For this

reason, we also include in the language a set F of n-

ary functions, 0 < n < ∞, such that if f ∈ F has arity

j, it can be used to build or point to an element of

A ∪ B given j elements of A ∪ B ∪ S . In other words,

f : (A ∪ B ∪ S )

7→ A ∪B.

In order to be able to build terms as in ﬁrst order

logics, we also include a countable set of variables

X = {x

,...}. Every formula built in this calculus is

assumed to be existentially closed, i.e. free variables

are implicitly bound to existential quantiﬁers.

We build relations involving agents and objects,

representing information that can be known by agents

and action statements. We have three sets of n-ary

predicates to represent each of these relation types:

• P : set of n-ary knowledge predicates, 0 ≤ n < ∞.

• Q : set of n-ary action predicates, 0 ≤ n < ∞.

• R : set of n-ary protocol names, 0 ≤ n < ∞.

Predicates are preﬁxed by modal operators as fol-

lows, in which p ∈ P ,q ∈ Q , r ∈ R ,s ∈ S and a ∈ A:

• [k]

p denotes a knowledge modality – agent a

knows fact p ∈ P at location s.

• [e]

q denotes an engagement modality – agent a

performs action q ∈ Q at location s.

• [i]

q denotes a location-speciﬁc computation

modality – action q ∈ Q is processed at location

ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence

198

• [i]

r denotes a location-speciﬁc interaction proto-

col – protocol r ∈ R can be started from location

In order to avoid unnecessary complications in our

proposed language, we allow modal operators to only

preﬁx a single predicate, i.e. no nesting of modali-

ties is allowed, nor it is allowed to have a modality

preﬁxing arbitrary formulae.

Communicative actions are deﬁned as follows, in

which p ∈ P ,a,a

∈ A and s,s

∈ S :

• null: a void message.

• [e]

write(p,a

): agent a writes message in loca-

tion s, which is then stored as predicate p known

by agent a

in s, i.e. [k]

p. In other words, agent

a tells p to a

in s.

• [e]

del(p, a

): agent a deletes message which was

previously stored in location s as predicate p

known by a

in s, i.e. the piece of knowledge

[k]

p is retracted from location s.

In order to continue with the deﬁnition of LS

we need to deﬁne two connectives:

• Non-commutative conjunction: given two exis-

tentially closed formulae ϕ and ψ, the conjunction

ϕ u ψ is evaluated as > if:

1. ϕ is evaluated as > AND

2. the variable bindings performed during the

evaluation of ϕ are used to bind the values of

variables in ψ, producing the instatiated for-

mula

ψ AND

ψ is also evaluated as >.

Otherwise, the conjunction ϕ u ψ is ⊥.

• Non-commutative disjunction: given two existen-

tially closed formulae ϕ and ψ, the disjunction

ϕ t ψ is evaluated as > if:

1. ϕ is evaluated as >, in which case ψ is never

evaluated OR

2. ϕ is evaluated as ⊥, and ψ is evaluated as >. In

this case, the variable bindings performed dur-

ing the evaluation of ϕ are not used to bind the

values of variables in ψ.

Otherwise, the disjunction ϕ t ψ is ⊥.

We deﬁne an atomic event AE as one of the fol-

lowing expressions, in which p ∈ P , q ∈ Q ,r ∈ R ,a ∈

A and s,s

∈ S :

• [k]

• [e]

• [i]

→s

mv, in which the special predicate mv is

used to state that agent a is being moved from lo-

cation s

to location s

• a communicative action M.

We deﬁne an event E as a conjunction of atomic

events, i.e. E = u

Finally, we deﬁne an interaction protocol as a pair

h[i]

r,t

i, in which [i]

r is a location-speciﬁc inter-

action protocol and t

is a non-commutative dis-

junction of events.

The interaction protocol h[i]

r,t

i is triggered

by a formula that uniﬁes with the left hand side ex-

pression [i]

r. Variable bindings are applied to the

right hand side expression t

, which is then com-

puted. Each event E

is an alternative course of ac-

tions that can be tested. If one of the events E

re-

turns >, then the interaction protocol succeeds and

the corresponding variable bindings are presented. If

all alternatives in tE

return ⊥, then the interaction

protocol fails and variable bindings are discarded.

It should be observed that, since location-speciﬁc

interaction protocol expressions [i]

can occur as

atomic events in the right hand side of interaction pro-

tocols, recursive interaction protocols are allowed in

C is a coordination language. Knowledge is

encoded in the platform using communicative actions

that update knowledge predicates p ∈ P , and action

predicates are expected to be evaluated by external

actors, which can include human as well as compu-

tational agents.

4 THE LS

C PLATFORM

We are working on a robust implementation for the

C language, beneﬁtting from existing implemen-

tations of LCC and of JamSession, that shall be freely

deployed as the LS

C platform. In this software plat-

form, the graph of locations, the list of pairs hs, ai, s ∈

S,a ∈ A for each predicate indicating where and by

whom it can be evaluated, and the state of each loca-

tion are managed in a centralised cloud server.

Interaction protocols are stored in distributed

hosts. The processing of these protocols may require

human intervention, this way characterising the LS

Platform as a tool to support and manage social inter-

actions. The physical location where interaction pro-

tocols can be found is stored in the centralised cloud

server as an address catalog. This catalog can be re-

arranged locally according to private ranking crite-

ria, deﬁned by priority policies used in different sites

which can be used to rank interaction protocols.

LS2C-APlatformtoDesign,ImplementandExecuteSocialComputations

199

CLOUD

C client:

• Predicates

• Protocols

C client:

• Predicates

• Protocols

C client:

• Predicates

• Protocols

C server:

• Graph of sites

• State of each site

• Location of agents

• Messages

Figure 1: The architecture of the LS

C Platform.

The locations of agents are also managed in the

centralised server, characterising the notions of vir-

tual worlds as featured in the JamSession literature

(Correa da Silva, 2011) and mirror worlds as fea-

tured in the LSC literature (Murray-Rust and Robert-

son, 2014).

The deﬁnitions of predicates – including action

predicates, which can capture the input-output ex-

pected behaviour of human actions – are stored in the

distributed hosts. The locations of interaction proto-

cols and their corresponding predicates are stored in

the centralised server.

The architecture of the LS

C Platform is depicted

in Figure 1.

Protocols can be triggered concurrently and asyn-

chronously by several users. As a consequence, the

veriﬁcation of properties related to distribution and

concurrency is important to ensure an expected be-

haviour in a system whose interactions are speciﬁed

using LS

C. We are working on the characterisation of

C protocols using Nested Petri Nets, based on our

experience using the same formalism to characterise

JamSession protocols. Nested Petri Nets can be used

to formally verify properties such as fairness, liveness

and termination. Given that Nested Petri Nets can also

be translated as Promela programs to be veriﬁed using

the model checker SPIN (Fernandez Venero and Cor-

rea da Silva, 2013b), we will be able to verify such

properties also for LS

C social interaction protocols.

Similarly to LCC and to what can be observed

in business process modeling (Correa da Silva et al.,

2012; Robertson, 2004), social interaction protocols

can be considered at speciﬁcation time and at run

time. Speciﬁcation time refers to the design of so-

cial computers, while run time can refer to the a

posteriori analysis of the actual execution of social

computations, in which e.g. speciﬁc protocols are

used to enact concrete interactions. Such analysis

can reveal social network properties involving inter-

acting peers, such as centrality of a location, and co-

hesiveness and density of location-related interactions

(Jackson, 2008), whose interpretation can be relevant

to understand features of speciﬁc domains.

5 AN EXAMPLE – LS

C FOR S

In this section we mention some potential applications

for the LS

C platform, and sketch how some interac-

tion protocols can be encoded for one of these appli-

cations.

The LS

C platform has been conceived to design

and implement social computations in which situ-

ated interactions are most relevant. Social comput-

ers of this sort can be found in urban computing

and in intelligent city environments, which are ur-

ban landscapes augmented with digital communica-

tion and processing devices and applications (Jiang

et al., 2013; Komninos, 2006; Schaffers et al., 2011;

Zheng et al., 2011).

Urban computing refers to requesting citizens to

carry (most likely within their smartphones) software

applications that track their activities, interact with

them and provide information to service managers,

so that the quality of service provisioning can be im-

proved in issues such as trafﬁc monitoring, public

transportation and emergency relief.

Intelligent cities are urban settings which have

been augmented with ubiquitous computing devices,

in such way that existing services can become more

effective and novel services can be offered to citizens,

businesses and governments. A representative exam-

ple of what can be achieved under the concept of in-

telligent cities is the structuring of effective business

clusters supported by digital services. In the follow-

ing paragraphs we detail this possibility.

An important well known factor for regional eco-

nomic development is innovation. Innovative en-

trepreneurship is frequently associated with start-up

companies, which in most cases are small companies

which hold deep knowledge and skills over a narrow

and specialised domain. One factor that has proven

to be inﬂuential for the survival of these companies is

their ability to cooperate with other companies, pos-

sibly forming or entering a network of cooperating

organisations. Local and regional governments have

taken notice of that and have created programmes

to support and incentive the blooming of such net-

works, as well as studied how these networks should

be structured in order to minimise the risk of failure of

participating companies and maximise the economic

efﬁciency of the networks (Feldman and Audretsch,

1999; Lazzarini et al., 2001; Mesquita and Lazzarini,

ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence

200

2007; Pedrozo and Pereira, 2006).

Business clusters are emergent agglomerations of

companies which beneﬁt from the proximity of each

other to grow. Smart Specialisation Strategies (S

)

have been proposed recently as means for policy mak-

ers to build “smart clusters”, in which the use of

knowledge and resources is optimised and coopera-

tion among companies is brought to be most effective

(EU, 2012; EU, 2013). S

can be seen as an effort to

design business clusters, instead of simply providing

appropriate means for them to emerge.

Business relations are partially constrained by ra-

tionality rules (such as proﬁt maximisation and risk

minimisation). On the other hand, companies are

human-powered and human controlled institutions,

and therefore – especially small companies – are in-

ﬂuenced by human decision making, which takes into

account rules beyond those that can be captured by

simpliﬁed models of pure rationality (e.g. brand ﬁ-

delity, intuition driven trust relations, aesthetic con-

siderations and cultural afﬁnity). Company relations

are diversiﬁed and include customer-supplier rela-

tions as well as cooperative relations involving sim-

ilar companies (Lazzarini et al., 2001; Pedrozo and

Pereira, 2006; Mesquita and Lazzarini, 2007). Hence,

we suggest that business networks can be treated as

social computers, and that the LS

C Platform can be

a useful tool to design, implement, run, monitor and

iteratively reﬁne Smart Specialisation Strategies.

In order to illustrate how this can be done, we

show a simpliﬁed version of interaction rules that

could be relevant in a customer-supplier relationship

involving two companies. In this example, company

A asks company B to provide a service that is required

to carry on production activities within company A

Company A may wish to minimise risk in its op-

erations by limiting the number of open requests sent

to company B to a ﬁxed value N: once A has sent N

requests to B, it will only send a new request after B

has fulﬁlled at least one of the queued requests.

In order to model this small example, the graph

consists of two locations s

and s

, and edges con-

necting these two locations in both directions (Figure

2). Agents, in this example, represent orders: when

company A places an order, an agent is sent from s

to s

, and when this order is delivered by company B

the agent is sent back from s

to s

. In Figure 2 we

depict agents as black dots. In that ﬁgure, company

A accepts to have seven simultaneous open orders at

This example is borrowed and adapted from (Correa da

Silva et al., 2012). Evidently, we are exhibiting only a very

small fraction of a model for S

using this example. Our

goal is simply to illustrate how rules that could be used to

model a S

would look like.

Figure 2: The two locations and corresponding agents for

the customer-supplier example.

most (i.e. N = 7), as shown by the seven agents that

are inside s

The following three small interaction protocols

implement this interaction

• Protocol 1:

1. h[i]

req1(X,Order),

2. [i]

→s

mv u[e]

write(message(Order),X) u

3. [i]

supply(X ,Order)

4. i.

• Protocol 2:

1. h[i]

supply(Y,W ),

2. [e]

message(W ) u [e]

prOrder(W ) u

3. [i]

req2(Y,W )

4. i.

• Protocol 3:

1. h[i]

req2(Z,U),

2. [e]

del(message(U),Z) u [i]

→s

mv u

3. [k]

orderEnd(U) u [e]

contProd(U)

4. i.

Interaction protocols 1 and 3 reside in a host

managed by company A, and interaction proto-

col 2 resides in a host managed by company B.

[i]

req1(X,Order) triggers the interactions, by ask-

ing an agent X in location s

to start interaction req1,

in which order Order will be requested to company B.

This is performed by moving the agent to location s

where it registers the order and triggers protocol 2.

By pattern matching on the right hand side of pro-

tocol 2, the message stored in location s

containing

the speciﬁcation of the order is veriﬁed as being part

of the knowledge of Z while in s

, based on which the

We adopt the Prolog convention that variables begin

with capital letters, and all other terms begin with small let-

ters.

LS2C-APlatformtoDesign,ImplementandExecuteSocialComputations

201

order is processed (by triggering the action predicate

prOrder) and ﬁnally protocol 3 is triggered.

By pattern matching on the right hand side of pro-

tocol 3, the message is deleted and the agent is moved

back to location s

, then it is checked whether the or-

der has been properly delivered (using the knowledge

predicate orderEnd) and the protocol returns the con-

trol back to company A internal actions (using the ac-

tion predicate contProd).

This is a simpliﬁed example, in which suc-

cess/failure veriﬁcations of performed operations and

security issues are not taken into account. Additional

features can be implemented by extending these pro-

tocols and/or by adding special purpose protocols, to-

wards the design of interaction rules that can specify

and characterise successful relations between compa-

nies in a supply chain.

Other protocols can be designed to compete with

these protocols, and protocols can also be designed

to characterise cooperative behaviour of suppliers

to provide combined services to customers, towards

the design of interaction rules that can specify and

characterise relations between companies in a net-

work. Hence, the relations involving companies in a

netchain, i.e. a network of relations mixing supply

chains and cooperative/competitive relations (Laz-

zarini et al., 2001) can be designed.

Based on theoretical analysis of properties of the

interaction protocols, as well as empirical analysis of

actual relations that can result from the use of these

protocols, iterative reﬁnements and adjustments can

be made.

6 CONCLUSION AND FUTURE

WORK

In this article we have introduced the LS

C platform to

design, implement and execute social computations,

and sketched how it can be used to model and sup-

port a complex system of economic relevance, namely

the organisation of companies in a business cluster ac-

cording to S

A platform for social computations should present

features such as:

• The possibility to empower domain experts and

end users to build speciﬁcations and execute them,

• Technology-agnosticism, meaning that imple-

mentations can be built based on various and di-

verse software platforms, operating systems and

programming languages,

• Explicit account of participants in social interac-

tions and their possible behaviours,

• Resources for the design of interaction protocols

as well as for the analysis of existing protocols,

including formal analysis based on algebraic and

logical concepts.

The design and implementation of the LS

C plat-

form is work in progress. Since it inherits features and

properties of LCC/LSC as well as of JamSession, we

claim that this platform addresses all these features.

Our immediate future work concerns the imple-

mentation of the LS

C platform, and its ﬁeld test in

the development of realistic social computers.

The speciﬁcation of social interactions as charac-

terised in the LS

C platform can be used at least in

three different ways:

1. As a design tool to specify desired features of in-

teraction protocols in a decentralised way,

2. As a platform for the execution of social compu-

tations, and

3. As a tool to reason about speciﬁcations, including

strategic reasoning (e.g., given alternative proto-

cols that can be built, what is best for me/my com-

pany?), whereby participants may try out certain

behaviours “in vitro” before these can be actually

enacted.

LSC has been combined with an existing so-

cial network platform (Murray-Rust and Robertson,

2014), and JamSession has been combined with an ex-

isting workﬂow management platform (David, 2012).

We envisage that a full LS

C Platform can be imple-

mented as the combination of a novel implementation

of the LS

C language, a workﬂow management sys-

tem (e.g. Bonita

) and a social network platform (e.g.

elgg

or eXo

). The implementation of the LS

C lan-

guage shall beneﬁt from previous experience imple-

menting LSC and JamSession.

We are particularly interested in the characterisa-

tion of Smart Specialisation Strategies (S

) as a disci-

pline to steer the emergency of networks of social in-

teractions involving human-powered agencies aiming

at regional economic efﬁcacy. We believe that this ap-

proach can be appropriate to implement S

effectively,

and that the LS

C platform can be useful to support

the design and operation of business clusters follow-

ing S

. In future work, we shall explore these views,

hopefully through the analysis of empirical data re-

sulting from the actual structuring of clusters of inno-

vation as goal-oriented social interaction networks.

http://www.bonitasoft.com/

http://elgg.org/

http://www.exoplatform.com/

ICAART2015-InternationalConferenceonAgentsandArtificialIntelligence

202

ACKNOWLEDGEMENTS

This work has been partially supported by FAPESP

and CNPq.

REFERENCES

Bai, X., Klein, E., and Robertson, D. (2012). Choreograph-

ing web services with semantically enhanced script-

ing, pages 583–587. Web Intelligence and Intelligent

Agent Technology.

Christoff, Z. and Hansen, J. U. (2013). A two-tiered formal-

ization of social inﬂuence. In Logic, Rationality, and

Interaction, volume 8196 of Lecture Notes in Com-

puter Science – Springer LNCS 8196, pages 68–81.

Correa da Silva, F. S. (2011). Knowledge-based interac-

tion protocols for intelligent interactive environments.

Knowledge and Information Systems, 30:1–24.

Correa da Silva, F. S., Fernandez Venero, M. L., David,

D. M., Saleem, M., and Chung, P. W. H. (2012). Inter-

action protocols for cross-organisational workﬂows.

Knowledge Based Systems, 37:1–16.

Correa da Silva, F. S., Robertson, D., and Vasconcelos, W.

(2013). Experimental interaction science. Artiﬁcial

Intelligence and Simulation of Behaviour - Annual

Convention 2013: Workshop on Social Coordination

– Principles, Artifacts and Theories.

David, D. M. (2012). Protocolos de interacao baseados em

conhecimento : implementacao da plataforma Jam-

Session (in Portuguese). MSc dissertation, University

of Sao Paulo, Brazil, Brazil.

EU, E. C. (2012). Guide to Research and Innovation Strate-

gies for Smart Specialisation. European Union, Brus-

sels.

EU, E. C. (2013). The role of clusters in smart specialisa-

tion strategies. European Union, Brussels.

Feldman, M. P. and Audretsch, D. B. (1999). Innovation

in cities: science-based diversity, specialization and

localized competition. European Economic Review,

43:409–429.

Fernandez Venero, M. L. and Correa da Silva, F. S. (2013a).

Modelling and simulating interaction protocols using

Nested Petri Nets. Workshop on Formal Methods in

the Development of Software.

Fernandez Venero, M. L. and Correa da Silva, F. S. (2013b).

On the use of SPIN for studying the behavior of nested

Petri nets. 16th Brazilian Symposium on Formal

Methods.

Gelernter, D. (1985). Generative communication in linda.

ACM Transactions on Programming Languages and

Systems, 7:80–112.

Giunchiglia, F. and Robertson, D. S. (2010). The social

computer - combining machine and human computa-

tion. University of Trento Technical Report, DISI-10-

036.

Hansen, J. U. (2014). Reasoning about opinion dynamics in

social networks. In Proceedings of the eleventh con-

ference on logic and the foundations of game and de-

cision theory (LOFT 11).

Jackson, M. O. (2008). Social and Economic Networks.

Princeton University Press, USA.

Jiang, S., Fiore, G. A., Yang, Y., Ferreira Jr, J., E., F., and

Gonzalez, M. C. (2013). A review of urban computing

for mobile phone traces: current methods, challenges

and opportunities. In Urban Computing 2013.

Komninos, N. (2006). The architecture of intelligent cities:

integrating human, collective, and artiﬁcial intelli-

gence to enhance knowledge and innovation. In In-

telligent Environments 2006.

Lazzarini, S. G., Haddad, F. R., and Cook, M. (2001). Inte-

grating supply chain and network analyses: the study

of netchains. Journal of Chain and Network Science,

1(1):7–22.

Mesquita, L. M. and Lazzarini, S. G. (2007). Horizontal and

vertical relationships in developing economies: impli-

cations for smes access to global markets. Academy

of Management Journal, 51(2):359–380.

Murray-Rust, D. and Robertson, D. (2014). LSCitter: build-

ing social machines by augmenting existing social net-

works with interaction models. International World

Wide Web Conference Committee.

Pedrozo, E. A. and Pereira, B. A. D. (2006). Empreende-

dorismo coletivo e possivel? uma analise do pro-

cesso de constituicao de relacionamentos cooperativos

em rede. Revista de Administracao (in Portuguese),

12(4).

Robertson, D. (2004). Multi-agent coordination as dis-

tributed logic programming, pages 416–430. Proceed-

ings 20th International Conference on Logic Program-

ming – Springer LNCS 3132.

Schaffers, H., Komninos, N., Pallot, M., Trousse, B., Nils-

son, M., and Oliveira, A. (2011). Smart cities and the

future internet: towards cooperation frameworks for

open innovation, pages 431–446. Future Internet As-

sembly – Springer LNCS 6656.

Seligman, J., Liu, F., and Girard, P. (2011). Logic in the

Community, pages 178–188. Logic and Its Applica-

tions – Springer LNCS 6521.

Sindhu, J., Perreau De Pinninck, A., Robertson, D., Sierra,

C., and Walton, C. (2006). Interaction model lan-

guage deﬁnition. Open Knowledge Project – technical

reports, UK.

Zhen, L. and Seligman, J. (2011). A logical model of the

dynamics of peer pressure. Electronic Notes in The-

oretical Computer Science, 278:275–288. Proceed-

ings of the 7th Workshop on Methods for Modalities

(M4M2011) and the 4th Workshop on Logical As-

pects of Multi-Agent Systems (LAMAS2011).

Zheng, Y., Liu, Y., Yuan, J., and Xie, X. (2011). Urban

computing with taxicabs. In UbiComp 2011.

LS2C-APlatformtoDesign,ImplementandExecuteSocialComputations

203