Digital Democracy in Knowledge Society: A Proposed Architecture Based on Cloud and Complementary Technologies

This paper introduces some technologies that are fit for an architecture of digital democracy or E-democracy. It aims at proposing an architectural style emerged from tested and validated approaches, without relying on some radical innovation. Firstly, we propose an input-system-output model of E-democracy and knowledge society. This model is subject to permanent optimization following a trial and error paradigm similar to the artificial intelligence method of backpropagation. Secondly, we describe and advocate for some technologies and methodologies such as Cloud, Service-Oriented Architecture, Agile Development, Web-Oriented Architecture, Semantic Web and Linked Data. Finally, we assemble all these technologies and methodologies in an architectural style that follows several key concepts such as flexibility and adapability, citizen-oriented software development or abstract notions like participation, deliberation and inclusion.


Introduction
One of the salient matters of computing world is to expand its visions and horizons from a technical to a social dimension.Arguably, information and communication technology (ICT) may arrive with its new approaches and perspectives to build a better society.While ICT has evolved as a human-like universe transposed in mathematical and computational formalizations, it is now time for the public sphere to benefit from the scientific achievements of the virtual world.From a social perspective, these benefits may endeavor the development of E-society, which is either the information society (IS) or knowledge society (KS).Moreover, IS or KS are foundations for a better public sphere by supporting a democratic society.Digital democracy in knowledge society or Edemocracy (ED) is to many an extension of Egovernment, but we have already defined it as being more than this -a way of living [1].Based on Maier's research [2] and extending ED with new instruments like E-petition under the guard of Justice, Figure 1 illustrates our proposed model.While Figure 1 has a generic perspective, ED should focus, on a bottom-up approach, on solving contextual problems (CPs) based on participation, deliberation and inclusion (PDI).Advocated by participative democracy proponents from antiquity [3] to modern [4] and contemporary times [5,6], PDI is the key to ED and a metaphor for KS.Justice, seen as the backbone of democracy inspired by divinity [3,4], is herein subject to permanent transformation through PDI on medium-long term, while some stability is required on short-medium term [6].In addition, Figure 2 depicts the actors involved in the CP processes and their defining inter-relationships: committee of MPs (CMP), committee of citizens (CC), helping committee of citizens (HCC), nongovernmental organizations (NGO), social and professional associations (SPA) or political parties (PP) [1].All of them co-work on a platform of CP (PCP) or instantiate a CP (ICP) only under the surveillance of committee of justice representatives (CJR).

Fig. 2. Components of E-Democracy
We have already brought several important amendments to other (representative) democratic models [1] advocating for: the increased role of citizenry, the way of selecting representatives, the crucial part of justice and the significant aspect of PDI.In addition, we only want to address the issue of establishing a better framework for collaboration and cooperation based on ICT.While this is subject of future extended research, we mention that E-bureaucracy is an improved and more objective (non-Kafkian) form of bureaucracy that helps monitoring ED by CJR and supports the actors involved in PDI.E-bureaucracy incorporates techniques and methods of web semantics, neural language processing, text mining, artificial intelligence (AI) etc. that conceive a substratum for (E-) justice in particular and PDI in general.A short paperwork like this one could not thoroughly argue for ED or KS and we herein want to highlight an ICT perspective that designates some technologies (substantially visually illustrated, but also literally depicted), which loosely and naturally build a foundation for a democratic better society.Defining the architecture of ED (AED) by an individual or an organization is rather a totalistic approach.AED should build itself on a bottom-up strategy using validated common shared knowledge and technologies.Describing AED, we only want to prove that these technologies already exist as well as some propensity in achieving this goal.This article has the following structure: the next section discusses KS and its role for supporting ED, in section 3 we present the technologies of AED, section 4 assembles the previous sections framing AED and the final section presents the conclusions of this research.

Knowledge Society (KS) and E-democracy (ED)
In the model herein, KS and ED overlap each other.While they share the same public sphere, they are different as the former deals mostly with the social life and the latter concentrates on political domains.Let us take them altogether as a unitary block for the moment and let us define them as the output in a joint model of ED and KS (MEDKS).For more than half a century, political sphere has already been seen [7] as an input -system (processes) -output model (ISOM).MEDKS has the same approach only that we compare it with an AI system that is self-adaptive and subject to optimization through the trial and error process called backpropagation (BPE).
Figure 3 illustrates the similarity of MEDKS and BPE, emphasizing the training process of CDJ through PDI, leading to a maximum output (i.e.KS and ED).While the so-called hidden layer of BPE concerns virtual computational intermediary results, MEKDS expects this layer (of PDI) to be transparent in the real public sphere.The first five steps of BPE are part of the feedforward process, which for MEDKS becomes an anticipation (of results) mechanism.Starting from the sixth step, the BPE itself trains the system to optimize the results, which for MEDKS means a trial and error PDI process to ameliorate and refine both outputs and inputs.The remaining of this section will discuss the role of KS in building ED, their common realm and some points that differentiate them based on several contexts.An acknowledged approach on describing KS belongs to UNESCO that defines three objectives to accomplish four desiderata [9]: cultural diversity, equal access to education, universal access to public information and freedom of speech.The objectives are: i) promoting digital opportunities and social inclusion by using ICT; ii) increasing capacity for scientific research, information and cultural propagation, performance and cooperation; iii) enhancing through ICT the opportunities for (E-) learn- From a different conceptual position and using an economic approach, Castells firmly asserts that technology is society, while making the analogy of contemporary world with a system of social connections axiomatically designated as network society [12].His sociological approach stresses out 'informationalism' that changes society (not quite in a revolutionary way) and the paradigm of knowledge acting over knowledge.More, network society is similar to Habermas' public sphere [5] on general level, but it extends to network of networks that lead to smaller public spheres fit for debate, information transfer and autonomy [11], an approach that supports the concept of CP (see Figure 2).Key elements like economic perspective and flexible labor force that helps integration of women, defines the new society (of flexible woman replacing man organization) that relies on the following concepts [12] Thanks to ICT, society has been experiencing many transformations in a more accelerated rate than decades ago, becoming quite an Esociety for some academics.Through innovation (not a radical but an incremental one at a fast rate) that resides on knowledge, an emergent KS seems to spring forth and this is the stage where PDI acts in order to achieve ED.KS and ED are related and a delimitation between the two takes into account the fact that KS may support a less democratic political system, while ED may perform even by lacking expertise and savvy.Nevertheless, when the two cooperates, they should manage different, inter-connected and complementary realms.KS deals with social, cultural, educational, professional dimensions leading to politics, while ED endures the political system yielding improved conceptualizations for the former.Metaphorically speaking, KS is practice while ED is theory, alternately modeling each other in a complementary way, like cast iron scaife (charged with diamond powder) polishes diamonds.One needs diamonds or similar composites to burnish other diamonds, in a more or less substantial manner and level of priority.Figure 4 undertakes the task of visually illustrating the theoretical balance of KS and ED, and the practical and undesirable situations that disturb this balance.The ideal situation is when KS and ED share a common ground and DOI: 10.12948/issn14531305/18.4.2014.07 the former encompasses a larger field of human interest (e.g.social and cultural life subdues politics).There are situations when one of the two becomes more important (e.g.KS is disproportional larger than ED in case of disaster, when rapid actions should prevail discussions).Unwanted instances are those when either avidity or apathy characterize political process (e.g. the economical flourishing life make people withdraw from social debates).

Fig. 4. Perspectives of Knowledge Society and E-democracy balance
ED and KS are complementary in MEKDS and, while ED springs from KS, the former reshapes the latter in permanent cyclical PDI processes.Although both have substantial idealized formation, the paper has shown some practices that proved a real tendency in envisioning them (especially KS).
After introducing elements of ED, discussing KS and ED in an ISOM approach and then delimiting the two, the next section will propose several technologies that accommodate AED.

Cloud and complementary technologies (CCT)
The previous section emphasized the interconnectivity (e.g.network of networks) and dynamics of KS (e.g.PDI) which require a technological approach that relies on flexibility, adaptability, stability etc. Considering these, on a large scale and with wide perspectives, the paradigm of Cloud Computing (or Cloud) seems appropriate for AED.only some areas (see example of item b).The most comprehensive definition of Cloud, according to the National Institute of Standards and Technology Laboratory, points out several characteristics [13]: on demand self-service, ubiquitous network access, location independence, resource pooling, rapid elasticity and measured service (pay per use).In network of networks society with large and small public spheres (e.g.KS), Cloud provides room for everybody, offering four deployment models: private (enterprise owned or leased), community (shared infrastructure), public (megainfrastructure), hybrid (two or more clouds).On a joint approach of IBM, Google and Intel, Cloud provides the ability of end-user to benefit from technology without managing its complexity [14] and this is a key aspect for inclusion in KS.
There are three basic services model for Cloud: i) Software as a Service (SaaS), providing user applications over a network; ii) Platform as a Service (PaaS), deploying custom applications and iii) Infrastructure as a Service (IaaS), supplying computing resources.
Figure 6 illustrates cloud levels in a more detailed approach and from two perspectives: provider [13] and user [15].The former is a bottom-up view with IaaS as foundation and the latter is a top-down view emphasizing the role of SaaS in commanding the other services.

Fig. 6. Perspectives of Cloud: a) provider and b) user
Cloud is not a miraculous solution that solves all the problems and it still needs complementary technologies (CT) in order to address issues of AED (and other complex systems).Table 1 illustrates most of the pros and cons of Cloud [13] and it is obvious that it is adequate for AED.CT will fix some of the issues that make Cloud unsuitable for AED: dependent and old applications, security level, control and cost.   1 and Table 2 we notice that security, control, cost and openness are some important issues to address.ED relies on transparency and citizenry control (see E-control in Figure 1) so the first two issues are negotiable on conceptual level.Normally, every cost should diminish with shared resources and even high investments, subject to permanent analysis and improvement, will bring a decent return on a medium and long term (see subsection 3.2).All elements of PDI need openness in a manner that overtakes transparency and requires direct access, a problem that this paper will address, through presentation of CT, partially in subsection 3.1 and highly in subsection 3.3.

Service-oriented architecture (SOA)
SOA is not a new concept and, although Web Services (WS) are its key factor nowadays, it has its roots in technologies like Common Object Request Broker Architecture (CORBA), Message-Oriented Middleware or Java Messaging Server (JMS).In a simplified approach, it is mostly a business architecture relying on well linked (i.e.defined processes), though loosely coupled (i.e.simplicity and autonomy in a wide range of particular services), black boxes (i.e.hidden complexity) components, assessing four key concepts [16].They are: i) reusability (keeping old technology); ii) superior quality through safety, accuracy, predictability and regularity; iii) non-professional DOI: 10.12948/issn14531305/18.4.2014.07user access and iv) compliance and adaptability to external rules.Using the same simplified approach, SOA has two major components: (Web) Service Client or Consumer (SC) and (Web) Service Provider (SP).However, due to transferred message complexity, scope or environment, new components (NC), belonging to SP or externalized in Cloud, may emerge.NC and their purposes are [16]: i) broker: mediator that keeps track of all WS on logical and conceptual level (e.g.identification, role, semantic, links of WS); ii) registry: a (physical) database of WS; iii) manager: registry or NC administrator; iv) security: rules and monitoring of WS transfer, protection of messages and NC and v) monitoring (useful for broker, security and manager): traffic, statistics, correlations and analysis.

Fig. 7. SOA approach
Figure 7 illustrates SOA with NC, depending on three important functionalities or tiers: i) demand or request (from SC); ii) abstracting (blend data from different mutual unaware interfaces at middle tier) and iii) answer or response (from SP).Table 3 succinctly presents several deployment models [17], through different methodologies that (auspiciously) prove that there is a multitude of approaches for communication in AED.In the beginnings, SOA used to rely on CORBA or JMS, but nowadays WS absorb most of the market and this is the reason many take the former and the latter as analogous.Yet, they only share some common realms (not all WS are part of SOA and SOA may use other services) and the novelty WS has brought to SOA consists in decoupling from operating systems or specific (programming) languages with benefits on flexibility and adaptability to re-conceptualization. Three concepts are important on a WS [15]: i) atomicity or finer granularity means that a service is self-contained, independent from the state of other services; ii) composability is ability to (re)compose services that become compound, thus having a large granularity and iii) loosely coupling is a design approach where complexity is hidden and external behavior is available for other services.Subsection 3.3 elaborates more on WS, but for the moment, we wrap up Cloud and SOA in a joint approach.SOA is an assembly of architectural methodologies that can provide Cloud instances of heavy business to external endusers, outside the perimeter of firewalled enterprise.From AED point of view, Cloud provides (formatted) data for simple requests that stay in delimited area and SOA may mediate interconnectivity between clouds.In order to leverage Cloud through services, five concepts of SOA incremental analysis (SIA) are needed Next subsection proposes a way of mediating user involvement as well as acquiescing SIA.

Agile Development
Before discussing Agile, let us introduce the concept of Lean architecture that will also help configuring AED by departing from computational information concepts towards wider informational and knowledge-oriented perspectives.Cloud and SOA are solid techniques that mostly deal with computers for the benefit of users, while Lean and Agile are ways of integrating human outsider approaches to system building, a necessity for AED that must not be a grant of programmers or computer scientists.Having its roots in Japanese culture and auto industry, Lean brings to (Agile) software development seven key concepts.They are [18]: i) deferring (classical) engineering concerns and paying attention to form; ii) empowerment of developers to make decisions regarding architecture; iii) using light application programming interfaces (APIs) and protocols; iv) simple documentation; v) people driven model; vi) collective planning and collaboration and vii) enduser mental model.The last concept is probably the most important for AED as a citizenoriented endeavor.Taking into account Table 4  While the relationship between elements of Table 4 is at row level in its original form, the herein approach stresses both the author and manifesto aggregated preference for the items of the left side, although there is value in the items on the right.There are 12 principles described in Agile Manifesto and key concepts for AED refer to frequently or adaptable to change continuous delivery of software, (faceto-face) collaboration between customers and developers, simplicity, self-improvement or sustainable development.Although the word semantic implies flexibility and speed, Agile, matching the pattern of MEDKS, is about self-organizing and feedback [18] or change through feedback by using four aspects of a shared mental model [20]: knowing, learning, understanding and executing.More, paradigm of Agile is a fit for KS requiring flexible strategies of testing, verification and validation [21], which also match the processes of the herein introduced MEDKS.Figure 8 illustrates a comparison of Agile and two classical development methodologies, using a simplified ADCT model.It also describes a more detailed structure of Agile development which somehow proves that is not a radical innovation but an improvement.

Fig. 8. Agile development
Actually, Agile is an umbrella of methodologies like Scrum, Extreme Programming (XP), Crystal, Lean etc., with the first two as the most prominent and mixed sometimes in practice [22] because Scrum exposes risks to improve project management and XP is about team activities practices [20].More, practice prove to proponents of Agile-and-architecture incompatibility that Agile develops i) components such as communication among team members, inputs to subsequent design decisions, documenting design assumptions, evaluating design alternatives and even architectural documentation related to extended geographical distribution, multiple demands and beneficiaries [23] or ii) aspects like planning, From SOA view and complying with new approach of SIA, Agile comes as an improvement based on adaptability, flexibility and empowerment by reasonably and dynamically modifying, on a temporal axe, the static robustness and resilience of the former, though keeping an architectonic perspective [23].SOA (mocked at as Same Old Architecture), based on SIA, benefits from Agile that focuses on client and direct action.More, Lean adds new value to SOA and Agile by taking into consideration the end-user and focusing on the thinking process over complicated, but predictable, aspects [18].We conclude this subsection by stating that, from ED and KS views, PDI may benefit not only from AED implementation using the herein discussed technologies, but at conceptual level it may import their patterns: participation of all stakeholders in a service-oriented cloud, agile deliberation between CDJ and lean inclusion of marginal citizens.

Web-Oriented Architecture (WOA), Semantic Web (SW) and Linked Data (LD)
WOA, an extension of and relying on SOA, refers the same concepts (e.g.reusability, interoperability, loose coupling, abstraction etc.), being an emergent global modular software architecture, but not defined by any standard body.It employs only WS and it has its roots in the architectural style of Representational Stateless Transfer (REST), in Hyper-Text Transfer Protocol (HTTP) and Internationalized Resource Identifier (IRI) and in communication dependent on client state [24].REST is a guide that provides some semantics to communication (using the verbs GET, POST, PUT and DELETE) based on four principles [25]: i) using IRI; ii) resources through representations; iii) self-descriptive messages and iv) hypermedia as the engine of application state (HATEOAS).In the beginnings of SOA, the standard of almost all WS heavily relied on Simple Object Access Protocol (SOAP), Web Service Description Language (WSDL) and Universal Description, Discovery and Integration (UDDI), but the latter, an intended WS registry, has been lately abandoned.While REST / HTTP WS (RHWS) may accept any protocol and is datadriven, using syntax light JSON as preferred data format (XML and YAML also work fine), SOAP and WSDL (SOWS) use only (heavy syntax) XML and become a burden because the latter obliterates the former's flexibility [26].Figure 9 illustrates WOA in a simplified way with emphasis on three tiers: web server, WS and client browser, all of them complying with REST and HTTP 1.1 principles.Although RHWS may rely on Web Application Description Language (WADL) that is optional and not very generous, it has an important feature that SOWS does not have: caching mechanism.Thus, not only that is more flexible, but also RHWS supports openness (a key concept for AED), while SOWS are suitable for APIs that require high security using a well defined standard (RHWS is hardly fit for this).RHWS has most of its consumers, if not all, bound to web browsers and this assures a high portability that leads to wide-open access.
REST provides some meaning on operations (using the verbs), but what is more important for end-user and computer is the semantic of data transferred on the internet.SW or Web 3.0 is not a substitute for Web 2.0, but a facility for knowledge management of content that is supposed to be machine-accessible through some meaning [27].Figure 10 illustrates SW (item a) and two types of databases that SW could rely on: regular (item b) and LD (item c).While the former has a predefined structure that assures easy management of data and it benefits from established expertise, the latter is irregular, more difficult to manage, but it provides meaning of data for all systems that accede to a model such as WOA.Regular databases are locked-in and accessible through Cloud and SOA and sometimes WOA, while, on the contrary, LD is subject to transparency over WOA.It is practical to use different IRIs for objects themselves and for documents that describe them and this coherence depends on three types of LD links: i) relational; ii) identity and iii) self-descriptive [28].Cloud APIs may rely on SOA and provide distribution of data and information for other APIs or they may follow SaaS approach.SaaS is a great opportunity as it supports ubiquitous devices with low processing power (smart phones, tablets etc.).Mobile devices seem to keep their ascending trend and, in response, more big companies move to Cloud some of their technologies [30].
Figure 12 illustrates CCT applied on KS, in a paradigm with one end-user.KS is all about openness without neglecting privacy.Thus, the end-user may have different statuses and roles that imply utilization of web application (e.g.news, feeds etc.), client application (e.g.financial statistics and analysis of some organization under EU umbrella) or SaaS for public information or comparative analysis of health systems at local level, for example.There are two classes of users: citizen and Evoter, with the latter as an extension of the former.Citizens participate actively through PDI (which may follow the paradigm of Agile) and has the characteristics of the end-user in KS, see Figure 12.Although it is not compulsory, citizens are political non-apathetic and they constantly debate not only at sociopolitical conceptual level but also, involving software developers, about the API functionalities that serve them as ED's stakeholders (e.g. through Agile).Moreover, when a citizen has also the assumed role of voter (again, not compulsory), it has the facility (a euphemism for obligation) of getting informed.Cloud and one or several SaaS provide info for E-voters and citizens, by casting data into PaaS that subsequently integrates them.In addition, SaaS facilitates the process of online voting, which requires a higher level of security (using SOWS, while RHWS would be the choice for other processes).At conceptual point of view, Cloud encompasses the whole and, in addition, the notion of Web of Data gives a new flavor by disseminating the former to the end-user level through SOA and, especially, WOA.While CCT follow a less collaborative development model of APIs at KS level (though marketing strategies may function to identifying client needs), ED requires a feedback-oriented approach.Thus, through Agile (a concrete reliable and established model mapping the more abstract PDI) citizens cooperate with owners (private enterprises or public organizations) and developers to enhance or build new APIs.While for private APIs, cooperation between citizen and owner is not compulsory, if they are not sanctioned by public organizations, public APIs are definitely subject to citizenoriented approach.E-voting is an instrument that deserve a research on its own and we only point out some aspects that are important for ED at architectural level.There are many actors that participate in an electoral process (candidates, voters, governmental organizations, NGOs, PPs etc.) and the herein E-voting is more than just the activity of one-day voting process.Firstly, there are two main types of voting: E-referendum and electing representatives.Secondly, E-vote is part of a bigger picture whose name is E-control (see Figure 1), and this involves several APIs regarding analysis, candidate descriptions, archives etc. Thirdly, there are security and compliance issues and these two demand APIs for validations, testing, identity checking, E-voter training etc.While many info aspects of E-voting could be subject to private organization applications, the herein proposal is a multi-SaaS approach, which means that a PaaS/IaaS is the solution to implement applications, repositories, management systems etc. that support E-voting.Evoter is an extension of (E-)citizen that have one particular duty: to participate to the decision-making process directly (referendum) or indirectly (through electing representatives).Concluding the short review on E-voting, Cloud solutions (presumably IaaS, PaaS and SaaS) extended with SOA and SOWS are the probable approaches to implementing this. Figure 14 is the final illustration of CCT that supports AED, integrating the previous two approaches of CCT for KS (see Figure 12) and CCT backing ED (see Figure 13).

Fig. 14. E-democracy's architecture
There are two large conceptual-overlapped units of AED relying on CCT: KS and ED.On the other hand, when there is abstract PDI involved, this may also require Agile methods in software development and maintenance.
While the herein proposed model of AED is also subject to change, every API or technology should follow a collaborative citizen-oriented approach such as Agile.AED or any other model of E-democracy may have one certainty (at least for the short and medium term up to new findings): it must follow a service-oriented model that provides room for reconfiguration, flexibility and adaptability.

Conclusions
This paper has undertaken the difficult task of conceiving an architectural model of digital democracy in knowledge society.While an architecture of such entity may not be subject of an independent research by an individual or an organization, we have identified some technology that better fits E-democracy and that are components of a trend described by human society.This endeavor only assembled validated methodologies and technologies belonging to digital world.Moreover, when we succinctly introduced an input-system-output model of E-democracy (i.e.MEDKS), we borrowed an artificial intelligence approach to describe the evolution of the model.Based on backpropagation analogy, MEDKS is subject to permanent optimization seeking for improvement through participation, deliberation and inclusion, which find an equivalent in a software development methodology (i.e.Agile).
Tested and validated business models (e.g.Cloud, SOA) and methodologies (e.g.Agile) or architecture styles (e.g.REST) borrowed from digital world serve the purpose of building an architectural model of E-democracy.
Our model relies on two large components: knowledge society and E-democracy itself and we introduce an architectural pattern for each of them in order to subsequently combining them in one piece.There are some compulsory aspects for E-democracy: transparency for politics and openness for public information systems, citizenoriented application development based on feedback or flexibility and adaptability at architectural level.Giving the complexity of this endeavor, an implementation of a system similar to the herein proposed one should start from local community (laboratory-oriented) level, which will provide useful insights of technological nature.Moreover, we will observe social and political behavior of the community and we will measure the short and medium term outcomes (defined at community level) of such system in order to improve (as a permanent required task) E-democracy.

Table 1 .
Opportunity and inopportunity for Cloud

Table 2 .
Cloud: benefits and drawbacks

Table 4 .
Agile declaration DOI: 10.12948/issn14531305/18.4.2014.07 Actors involved in PDI in non-political activities (e.g.NGO, SPA) communicate through WOA and RHWS in an open-access paradigm.The political activities of ED refer three actors HCC, CMP and CJR (at conceptual level they are similar to CDJ), but the latter uses the roles of the former two (being the supervisor of the political process, see Figure1and Figure2).Political actors' services rely on SOA and SOWS as already explained, see Figure13, but there are two other types of relationships.One regards PDI with the non-direct decision makers (i.e.SPA, PP) through open-access RHWS (for transparency).The other takes into consideration APIs that help PDI between the decision makers (i.e.HCC and CMP) and follow SaaS model.In addition, these SaaS APIs use facilities of IaaS (e.g.Testing as a Service (TaaS), Database as Service (DaaS), Management as a Service (MaaS), see Figure6.a) derogating software development and maintenance to Cloud.Using PaaS, citizenry services access info by RHWS and vote (in a referendum) through SOWS.