Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Business of Knowledge and Modern Infrastructure Industry

Evolution of civil infrastructure from a technical artifact into an engineering system and a national asset over the past century has created a new discourse for development, construction, and management of infrastructure, which more and more emphasizes soft and subjective aspects of the system. Modern civil infrastructure is a complex system composed of the physical network of assets together with the social network of actors/users, and their interactions through the operational processes of the system (Lukszo & Bouwmas, 2005). This defines a sociotechnical system whose behavior cannot be studied without respect to the associated agents and the related social/institutional infrastructure. This system will be governed by organizational policies as well as social norms and standards. Such a definition for civil infrastructure has improved the role of the society from customers and end users of a service into stakeholders who may influence specifications of the system. This new role introduces new opportunities and challenges to domain decision makers. On one hand, it creates great opportunities for social engagement. Technical and professional decision makers can distill the distributed knowledge of public communities (referred to as non-expert or non-mainstream knowledge by Brabham & Sanchez, 2010) to reinforce the decision making procedure. On the other hand, given the diversity of interests and technical sophistications involved, an active participation of the public may result in a chaotic nature for the decision process.

If the ‘inter-organizational/networked-based arrangement’—as stated by Keast and Hampson (2007)—is the new measure to facilitate innovation development and diffusion within the construction engineering and management (CEM) industry, expanding such networks to include non-technical stakeholders and end users of the urban infrastructure in the future can expedite the process of innovation in CEM even more. Taking advantage of the ‘user innovation’ by involving lower layer nodes of the technical social network in construction projects (such as site workers and technology users) improves the process of innovation in CEM (Sanvido & Paulson, 1992; Slaughter, 1993). This can now be extended to include the knowledge-enabled social communities and end users of the built environment. Active involvement of communities is poised to move sociotechnical infrastructure into an innovative and socially-savvy environment for decision making. Apart from the contribution to the innovation, a mixed network of technical and nontechnical decision makers/decision participants is a response to the demand for more active role of NGOs, political and social groups and communities towards a sustainable development in global construction industry (as addressed by Levitt, 2007). Therefore, the role of Public Involvement (PI) agencies is no longer to promote ‘the’ best solution, but to empower communities to discover it through democratizing innovation (Von Hippel, 2005).

Public engagement has traditionally aimed at maintaining a balance of power between citizens and their government in the process of decision making. The main role of this mechanism is to minimize the impacts and risks of failure of decisions made in development and construction of infrastructure. Evolving community engagement processes from a passive process of ‘public relations’ to a process of ‘engaged partnership’ is one of the requirements for achieving a desired sociotechnical model of infrastructure decision making. In the past, community engagement was mostly limited to informing and educating the public with the aim of maintaining the required level of public support. The process is now evolving into other forms of consultations to establish a two-way communication. As Hansen and Jackson suggest (2001), the success of community engagement processes is closely tied into involving the community in a timely manner (from early stages of the project) and continuously (during different phases of the project lifecycle), as well as acknowledging the role of end users (as customers) in shaping decisions related to the infrastructure. At lower levels, static tools such as web pages and open houses are commonly used with objectives such as providing the community with the project specific information or making them aware of decision impacts. However, advances in Social Web (Web 2.0) have added a new dimension that includes multipurpose collaboration between community members for collective deliberation on complex topics.

1.1 Role of Web 2.0 in Public Involvement

As Brabham and Sanchez (2010) indicate, traditional methods of PI are problematic due to the lack of efficiency to engage a fully representative sample of the community, and power dynamics of face-to-face meetings: “citizens may feel their opinions are downed out or feel compelled to self-censor”. Moreover, it is difficult, if not impossible, to ‘educate’ participants of a public meeting with the required level of technical information and details within the limited time frame of a meeting. By analyzing the literature of public engagement best practices in transportation planning, Wagner (2013) suggested three main principles for performance of the engagement process: accessibility, interaction, and outcome–orientation. Web 2.0 can help the online community (e-society) to outperform the offline public in almost all of these dimensions. Epidemiology of knowledge through Social Web is upgrading the e-society into the k-society (knowledge-society). Acquiring the knowledge distributed among the online k-society can result in the development of more robust plans. Moreover, people claim to enjoy participating through online social media and it can increase the level of engagement (Evans-Cowley, 2012).

In short, a meaningful public engagement process must be “open, ongoing, and allow for two-way exchange of information” (Wagner, 2013), and Social Web can help to create all of these features. Many infrastructure planning organizations consider the infrastructure as a marketable product and approach the online public engagement in a similar way to online marketing. Engaging customers in the field of commerce and using reverse marketing mechanisms to collect user innovations are among the best motivating scenarios in this regard.

1.2 Era of Prosumers

Informatics in its modern form does not deal with segregated producers and consumers of knowledge anymore. Today they are both morphed into “prosumers”; a portmanteau formed by contracting the word professional, or also producer, with the word consumer to emphasize the active role of consumers in producing the products they use. Circulation of knowledge between producers and prosumers is the hallmark of a big data movement and is a key factor of the evolving knowledge economy. By relying on human intelligence, this movement performs tasks which are impossible to accomplish otherwise. Wikipedia is an example of prosumer culture outcomes. With more than 26 million articles in 286 languages, it is one of the most (if not the most) popular encyclopedias in the world. It has at least 70,000 formal prosumers (editors) among its estimated 365 million readers. YouTube is another ultimate example of prosumerism. People on this website create (post), “consume” (watch), re-create (re-mix) and exchange the products (videos) produced by other prosumers.

In a more formal way, this mentality has given rise to crowdsourcing which expands beyond marketing and chattering, to enable “knowledge workers” to solve problems. InnoCentive, a problem-solving marketplace, has 250,000 “solvers” competing for more than $35 million in prizes. Still in its (perpetual) beta version, Amazon MTurk is utilizing hundreds of thousands of workers (500,000 as of Jan 2011) from all around the globe (over 190 countries as of January 2011) for a large number of human intelligence tasks of different types. Spigit in the corporate social innovation field, Covisint in the automobile manufacturing industry, and Salesforce in customer relationship management are among other examples of successful cases in this regard. In business administrations, this mindset paves the way for reverse marketing which in turn alleviates harvesting user innovation for new designs. As part of a program called “future by Airbus”, Airbus ran a 2-year global consultation with more than 10,000 of its future passengers, asking them for their requirements, demands, and innovative ideas for a 2050 aircraft. The idea behind this program was to involve people from various backgrounds to shape the future of aerospace industry in a more sustainable way. Since launched in 2008, they have been running a competition every two years with a prize of €30,000, asking students for their innovative solutions (Future by airbus, 2012).

Wisdom of the crowd can particularly be helpful in developing context-sensitive solutions for case-specific issues. “Online urban guides” (systems such as social recommenders, rating sites, and review services) are good examples of this type. Today, websites such as Tripadvisor and Yelp, may be among the most reliable sources of knowledge regarding local services. By adapting the user-generated content and aggregating prosumers’ micro knowledge, these websites help decision making in cases where not enough information is documented formally and officially, or where the documented information is in form of a ‘negotiable knowledge’.

1.3 Prosumerism in Infrastructure Industry

In the domain of infrastructure, prosumer culture can be helpful in different stages from detection and adjustment of demand to selection of design and construction alternatives. Knowledge-enabled communities discussing different aspects of their built-environment can reflect demands, interests, and (from time to time) innovative solutions for the infrastructure system. This is an invaluable opportunity for collecting the distributed micro-knowledge and shaping more pluralist solutions. Mining online discussions can result in capturing and formalizing the knowledge generated through communications among users who are a part of the socio-technical system and continuously interact with it. On the other hand, analyzing patterns of online social connectivity among such decision participants can foster creation of teams for public consultation in different stages of planning and construction. One major advantage of such a model is its self-organizing nature. On one hand, dynamics of discussions will be maintained by participants (which can eliminate concerns regarding outdated information on project websites), addressed by Wagner (2013) among others. Also, the wisdom of crowds can help to classify and prioritize the project-related content. On the other hand, the users unintentionally evaluate each other through activities such as liking, sharing, following, mentioning, etc. This establishes each individual’s influence level in a bottom-up manner.

Governments and other macro-level decision makers in the AEC industry (Architecture, Engineering, and Construction) have recently noticed such advantages and have started to benefit from prosumer culture in the process of engaging public partnerships for infrastructure projects. ‘New York city bike share’ program is an example of using prosumers’ knowledge in the process of demand detection. In 2011, New York City Department of Transportation (NYC DOT) asked local residents to suggest locations for bike stations through community workshops and meetings and then shared a draft of suggested locations with the users in form of an interactive map. The users were asked to help NYC DOT to refine stations’ locations by voting for or against locations suggested by others, offering new stations, starting new arguments, or attending existing discussions and submitting reasons in support or against other people’s suggestions. In another experiment, Kansas Department of Transportation (KDOC) launched a program called K-TOC (Kansas Transportation Online Community) in 2009, which was one of the first efforts in creating an infrastructure-specific online community. KTOC aimed to provide a forum for people and policy makers to communicate directly on transportation related issues. Encouraging online discussions, respecting customers by showing them they are being heard, showcasing opinion dynamics under social interactions, direct information exchange between all levels of the transportation industry, and unfiltered access to audience inputs were mentioned among the fundamental benefits of this program (KTOC, 2010).

2 Attempts and Shortcomings

Many studies in the literature such as Chinowsky, Diekmann, and O’Brien (2010) and Levitt (2011) have emphasized the role of managing construction projects through management of social networks involved. Since introducing the social network model of construction by Chinowsky, Diekmann, and Galotti (2008), some researchers have focused on social network analysis (SNA) of construction projects (Di Marco, Taylor, & Alin, 2010, and Wambeke, Liu, & Hsiang, 2012 among others). These studies however, primarily focused on a network of internal stakeholders involved in construction projects (traditionally called ‘actors network’). Moreover, scope of these models was mainly project management, rather than knowledge management. Parallel to these studies, given the diversity and complexity of stakeholders, in infrastructure management, network decision making has been emphasized as a process to reflect interests of all stakeholders in the final solution (Bruijn & Heuvelhof, 2000). In this respect, using online communication channels through the web was suggested as a more direct way to involve participation from external stakeholders of projects. However, scope of many practices in this regard (such as Lorenz, 2011) has been limited to channels run and owned by formal decision makers. Popularity of Social Web has recently drawn the attention of researchers in domain of civil infrastructure to online social media.

Evans-Cowley (2012) refer to public engagement through online social media (and particularly micro-blogging) as ‘Micro-participation’. Studies on micro-participation have been mostly centered on the content and sentiment of public inputs and have more or less ignored the social network formed in the background of discussions. Although the domain has generally agreed on micro-participation as a demand for PI, several barriers challenge its efficiency. The following paragraphs briefly discuss these challenges.

First and foremost, involving the wisdom of prosumers is normally associated with some levels of risks. When several parties, with diverse (and in many cases, conflicting) interests are involved in the problem solving, the information provided by different parties takes the form of negotiated knowledge (Bruijn & Heuvelhof, 2000). Information covering different angles of the problem and representing different decision makers’ perspectives may contradict each other, although all being true and valid. Such an issue is more severe when dealing with problems having a soft and subjective nature. Dichotomy of online and offline identities and attitudes is another challenge in involving online communities. This may result in incorrect or even fake opinion expressions among other issues. Usually forums where opinions are posted publicly suffer an obvious polarization of opinions, indicating that more moderate ideas are either not expressed or are dimmed under more extremist arguments. Moreover, involving public communities will change the nature of decision making from a structured project into a chaotic process (Bruijn & Heuvelhof, 2000). For example, in the case of NYC bike share, in period between September 2011 and April 2012, more than 10,000 locations were suggested and more than 60,000 supporting votes on the suggested locations were collected (NYC Bike Share, 2012).

On the other hand, the rapid growth of online social media and social networks (in form of websites such as Facebook and Twitter) has enabled citizens to express their opinions on various topics in full transparency, using a variety of devices. This helps governments to become independent from their proprietary social communication channels. It also creates a unique opportunity for a pro-active engagement system. As a result, in the beginning of 2012, Only 3 years after launching KTOC, KDOT suspended this website and transferred all activities hosted there to its headquarter Facebook page and Twitter account. In the announcement of this decision, KDOT’s social media manager stated:

“Given the rapid public embrace of the agency’s Facebook presence, the fact that Facebook hosts an online audience far larger than any that can be reached by a proprietary online community, and that Facebook is free, it is no longer possible to justify the annual expense[s] associated with the K-TOC software lease.” (Quinn, 2012)

In North America alone, 82 out of the 100 strategic infrastructure projects announced by North America strategic infrastructure leadership in 2011, have now active Facebook or Twitter accounts to post news, host public discussions, and collect the feedback. Based on TRB (Transportation Research Board) transit cooperative research program—synthesis 99, major transportation providers who use micro-participation to involve the public in USA and Canada find Twitter and Facebook in many aspects the most convenient communication tools (Bregman & Watkins, 2013).

As an example, the network of Twitter followers of a Light Rail Transit (LRT) project in Toronto is shown in Fig. 1. This network was collected in May 2014, when the project was in the early construction phase. The network has 2,078 nodes and 46,852 edges in total, where node is a twitter ID and an edge connecting node A to node B implies that A is following B on Twitter. In order to detect the order beneath the chaos of such a network, we have filtered nodes with in-degree centralities lower than certain thresholds (to remove nodes with low number of followers). If a threshold of 50 followers is used, only 8.5 % of nodes and 17.1 % of edges will remain in the network (Fig. 1b). When we raise the threshold up to 300, as shown in Fig. 1c, only 1.5 % of nodes and 1.3 % of edges (34 nodes and 614 edges) will be kept. Although the lower-sized core can be interpreted easier, it is not yet a meaningful input to the decision making procedure.

Fig. 1
figure 1

Network of followers of Eglinton Crosstown LRT project on Twitter (collected on May 2014). (a) Complete network of project followers on Twitter. (b) Filtering out nodes with under 50 followers in the network. (c) Filtering out nodes with under 300 followers in the network. (d) Sub-network of mentioning. (e) Sub-network of Re-tweeting

Full size image

Taking advantage of listening to what citizens say about their built environment on online social media can provide planners with an opportunity to engage the public in a very different way (Evans-Cowley & Griffin, 2011). However, harvesting relevant items from the corpus of user generated content on the social media and analyzing them to achieve meaningful results require tools and methods which do not formally exist in the field at this time. As the survey by TRB synthesis 99 indicates, public relation agencies in North America use online social media with goals such as ‘communicating with current customers’, ‘improving customer satisfaction’, and ‘improving agency image’. Online social media plays the role of a communication channel to connect with the customers/community for a real-time communication, advocacy, and feedback collection. Moreover, it can put the customers in power by creating opportunities for user innovation (Bregman & Watkins, 2013).

Evans-Cowley and Griffin focused on project-related ideas discussed over social media and analyzed public discussions from perspectives of content (type and theme) and sentiment. By starting a program called SNAPPatx (Social Networking and Planning Project), they investigated if micro-participation can be analyzed to help understand the public's views on transportation issues. They used linguistic analysis and word count to assess emotional cognitive and structural components of more than 8,300 relevant tweets collected around transportation related issues in Austin, Texas. Results of this study approved that aggregation of microblogs can create meaning and help to understand perspective of the public community. However, from the aspect of providing decision makers with meaningful inputs, results of this study could not satisfy expectations. Official decision makers expected real-time data analysis with more meaningful results. More importantly, the public officials were generally interested in the identity of the users on top of aggregation of their opinion (Evans-Cowley & Griffin, 2011). The research was using off-the-shelf software for linguistic analysis; while as the researchers admitted, context-specific tools are required for this purpose. The SNAPPatx report ends with an emphasis on the demand for further empirical investigations to find ways in which information extracted through microblogging can be processed and weighted. It also insists on the need for developing a model to use micro-participation in planning effective engagement practices.

Therefore, most of studies in infrastructure PI have tried to evaluate online social media as a communication channel. They originally focus on how to build an online relationship with users to engage them in a dialogue, and more or less ignore the social network formed in the background of micro-participation. The conversational nature of social media however, has a ‘networked’ structure in which expressed ideas are linked to the followers of a project, including end users and public officials. These followers are nodes of the same layer of a network. El-Diraby (2011) refers to the network of people and network of ideas for AEC projects and their interdependency. Analyzing anonymous comments would focus on the ideas and ignore the people. As Evans-Cowley and Griffin (2011) mention, it is only a little more than “a finger in the wind”! Web 2.0 plays the role of a platform which brings people together and connects them to each other in the form of a heterogeneous network. It also documents/showcases their ideas and keeps the flow of project-related discussions among them. Networks which are formed around infrastructure projects are called Infrastructure Discussion Networks or IDN for short (Nik Bakht & El-diraby, 2013a, 2013b). The common needs and shortcomings in studying such networks can be summarized as follows:

  • Understanding the contents of ‘chattering’: While there exist many online IDNs, not much work has been done in applying formal socio-semantic analyses to understand the contents of the discussions and the relative importance and connections between ideas exchanged. The few attempts in this regard have used off-the-shelf software tools and were limited to finding keywords and sentiment of discussions.

  • Profiling the Stakeholders: with the increasing role of society in decision making, we need to create means to profile community members. Understanding citizen attitude and what impacts it is very important to customizing both the project and its communication policy to their needs. No systematic process or protocol has been offered to date for analyzing IDN layout (connections between members) to help in profiling and/or analyzing stakeholders.

  • Distilling crowd knowledge for long-term us: Context-sensitive solutions for urban infrastructure require local knowledge which is not exclusively owned by internal stakeholders (aka decision makers) anymore. Such knowledge is distributed among users (or future users) of the service and the social organization who interacts (or will interact) with the system. One prerequisite to such a heterogeneous and networked decision making scenario is distilling micro-knowledge of prosumers and aggregating it in a bottom-up fashion. It is important, then, to analyze the links between ideas, stakeholder profiles, project features to extract the basic rules/facts that makes a project more acceptable to the “new” decision makers. Through tracking IDN evolution over time and comparing different IDNs we can create a more suitable and customizable models of what makes a project a successful one. In other words, we should feedback the results of IDN analysis and comparison to reshape our ontologies of infrastructure projects.

In the light of these observations, we argue that effective analysis of the IDN would be the intersection of semantic and social analyses; while the former determines what has been uttered, the latter reveals who the utterer is. Text mining and natural language processing can handle semantics of discussions over the IDN (similar to Evans-Cowley, 2012, Lorenz, 2011, and Nik Bakht & El-Diraby, 2013b), and tools from SNA can help to uncover the composition of the social network which drives the discussions (similar to Nik Bakht & El-Diraby, 2013b).

In order for micro-participation to be integrated into a comprehensive engagement plan, infrastructure industry needs domain-specific tools to deal with IDNs from the two aspects mentioned above. This from one side requires benchmarking best practices from other domains and evaluating their applicability to the domain of infrastructure, and from the other side involves understanding infrastructure-specific content and trends of community inputs. The lack of tools to analyze seemingly chaotic public discussions and dialogues results in wasting opportunities for collecting prosumers’ knowledge and user innovations. This is becoming frustrating to both communities and official decision makers (Evans-Cowley, 2012). Off-the-shelf tools cannot be the perfect answer to this need, unless being validated and customized through adequate empirical analysis and investigation in nature of existing IDNs. Social network analytics must be combined with text mining to classify followers of the project based on their affiliations and interests (stakeholders’ typology), rank them based on their influence level (stakeholders’ power level), and detect subject and sentiment of their discussions (stakeholders’ vested interests and position). Outputs of such a detailed analysis may then provide decision makers with insights regarding stakeholders, which can help the PI program target the right people to be engaged and the right interests to be addressed by the plan.

3 From Big Data to Knowledge

As mentioned, IDNs are networks of ‘people’ and ‘ideas’. Formalizing analysis of online discussions in the process of micro-participation requires evaluating the two aspects and integrating them to create meaningful inputs for decision makers.

Managing distributed knowledge of prosumers through Social Web requires adhering to the social business model. Such a business model is founded based on helping customers to collaborate with stakeholders by sharing and organizing information via Web2.0 technologies. This is a deviation in traditional PI mindset; from a re-active approach which pushes the public to generate inputs, into a pro-active method which pulls information required for decision making from users’ discussions.Footnote 1 An example in using collective intelligence of crowds to detect small magnitude earthquakes can help to highlight the difference between the two approaches. “Did You Feel It?” was the name of a project using tools of Web 2.0 in a re-active fashion to crowdsource detection of small magnitude earthquakes. In this project, participants were asked to fill in online surveys about their experiences during an earthquake (Atkinson & Wald, 2007). Crooks, Croitoru, Stefanidis, and Radzikowski (2013) tried a pro-active alternative for reaching the same goal. They detected responses to the earthquake over Twitter by following the hashtag #earthquake right after the event of US East Coast earthquake in August 2011. They used social media feeds as a sensor system to detect and locate the 5.8 magnitude earthquake, without pushing users to generate any inputs.

In the pro-active school of thought, prosumers routinely express their experiences and opinions in full transparency and in a free format through multiple channels available by online social media. On top of administering these channels, the role of PI officials will be to distill the knowledge through detecting relevant comments, classifying, and evaluating them. As emphasized earlier, penetration of agencies such as Twitter and Facebook leaves no room for official decision makers’ propriety social networks to involve the community in a truly active manner. Tools and techniques are required to monitor the general social media and automatically collect the relevant data. The relevance would refer to the semantics of discussions and social value of the supporters. Such data must be then processed into the meaningful knowledge. In more specific words, the relevant content generated by the public over Social Web must be processed to answer the following questions (which are essential to the stakeholder analysis in many domains including civil infrastructure):

  • Who are the project followers? Answering this question is essential to understanding the typology of stakeholders;

  • What is the relative influence of stakeholders? Finding the answer to this question can provide decision makers with insights regarding community leaders and level of impact that each project follower can have on others. This can be known as ‘network value’ of project followers and can be used to evaluate the impact level of ideas discussed;

  • Which topics are being discussed? Detecting, understanding, and clustering topics discussed over IDNs can help decision makers with identifying needs, vested interests, feedback, and user innovations;

  • What is the sentiment of discussions? Stakeholders’ position in terms of being proponent or opponent to the project and/or decisions is normally reflected in the sentiment of their discussions.

It is noted that answering these questions must be an ongoing process during different phases of project lifecycle. Dynamics of answers to these questions can help to explore patterns of order existing beneath the chaos of public participation.

Analysis of microblogs (mainly, Twitter networks) centered on infrastructure projects can be vital in answering the questions above. On one hand, many infrastructure projects in North America have active Twitter accounts, and on the other hand, the open API (Application Programming Interface) of Twitter provides a great opportunity for researchers as well as field practitioners to pull data from this micro-blogging website. Moreover, Twitter not only archives social opinion in form of short statements, but also keeps the record of connectivity among individuals in the form of following, mentioning, and re-tweeting. Therefore, the ‘networked-ness’ of ideas can be tracked by studying Twitter. In the following, we address some of the results from studying IDNs formed in Twitter. Most of these analyses can be directly repeated for similar websites such as Facebook, online forums, and blogs.

3.1 IDN as Network of People

Studies show that IDN as a collection of social entities connected through social relational ties more or less shares the general behavior of general social networks. As an example, in the network of followers of a project over Twitter, different levels of social connections can be defined among the nodes. They range from weak ties such as “following” to stronger ties such as “mentioning” or “direct messaging”. Figure 1c and d respectively show the network of mentioning and re-tweeting among followers of Eglinton Crosstown LRT project in the period of May and June of 2014. This is another way to filter the chaos of IDN and reach into the core of a network. The results in these cases are weighted directed graphs in which node A is connected to B with an edge of weight n if A has mentioned B n times/has re-tweeted n tweets by B. Although as Huberman, Romero, and Fang (2008) suggest, such networks reflect stronger ties and can reveal more meaningful interactions among the nodes; as indicated by Easley and Kleinberg (2010) among many other authors, strength of weak ties cannot be ignored. Focusing on any of these types of connections uncovers a social network at a different level. The network built in this way portrays a layout of social connectivity among followers who have vested interest in the project. Studying IDN as a graph of social identities connected through social linkages reveals topographical and geometrical properties which have roots in formation and evolution of such networks.

It is shown that the geometry of IDN depends on the project’s nature and behavior of the community. Like many other social networks, IDNs follow power law degree distribution. This is a result of ‘popularity’ mechanism among project followers; while there are a few popular nodes with a high number of followers, majority have a low number of followers. As it is shown by Nik Bakht and El-diraby (2014), this behavior becomes more dominant (the rich get richer) as the project proceeds in its lifecycle. Networks with this behavior are called ‘scale-free’; this is not only due to the mathematical scale-free nature of the power function, but also to emphasize that networks related to fundamentally different systems exhibit the same characteristics. Therefore, models and algorithms from other domains, including reverse marketing and the social business model can be benchmarked and directly applied in infrastructure planning to improve the efficiency of the public consultation process.

Moreover, it is shown that IDNs are local and issue-centered ‘small worlds’ with a relatively high clustering, comparatively small diameters, and short average path lengths. Small world phenomenon, which refers to the rich nature of social networks in terms of short paths, facilitates information diffusion and provides a good opportunity for viral marketing around the project. On the other hand, measures and algorithms to evaluate the influence level of nodes can assist to find the high influential nodes among project followers and involve them in the process of consultation with the public. As projects proceed in their life cycle, their IDNs mature and demonstrate a behavior which is closer to more established online social networks. IDN of projects in later phases of their lifecycle engage more followers and grow in size. Moreover, their growth is involved in triadic closure; i.e., more connections are formed among friends of friends. The fact that evolution of IDN does not stop as the project progresses into later phases of its lifecycle can be a significant opportunity to extend the online public engagement into the whole project lifecycle as a continuous and self-organizing process.

As suggested in (Nik Bakht & El-diraby, 2014), Topological parameters of the IDN can provide decision makers with indicators of maturity and performance measures for the public outreach programs. Moreover, project followers in the IDN group together and form limited numbers of communities based on different forms of similarities (Nik Bakht & El-diraby, 2013a, 2013b). Project funding sources, various levels of decision makers, geographic similarity and project impacts are among the main criteria which segment the followers into communities. Analysis of influence patterns can detect the leaders of those communities. Detecting communities and their leaders can help the PI practitioners with the team-building process (by engaging community leaders), and classifying cores of public interests. This is particularly value-adding when hidden nodes from the public community are detected. For example, finding influential figures (such as a prominent journalist or an urban activist), who influence public’s mind the most is normally not an easy task if ever possible, in traditional PI practices.

Therefore, answering the first two questions in the set of four questions above requires detecting communities in the social graph of the IDN and uncovering patterns of influence among its nodes. Figure 2 illustrates some applications of topology analysis for IDNs.

Fig. 2
figure 2

Analysis of IDN as network of ‘people’; social network analytics

Full size image

Detecting patterns of influence—Real world dependencies among people in many cases are reflected in their online social relations. As mentioned, social connections over the web have various types and levels: from loose concurrencies such as subscription to the same group, to more direct relations like ‘following’ and ‘friendship’, and stronger ties such as direct messaging, re-tweeting, and mentioning. Although none of these can be a perfect representative of the real/offline relations, some of them can be adequately taken into account as indicators of the social influence. For example, from a ‘following’ relationship on Twitter or Facebook, it can be inferred that the followee may have a level of influence on the follower (Huberman, Romero, & Wu, 2009) (at least all of the followee’s posts can be seen by the follower). After taking such an assumption, detecting top influential nodes of the IDN can help the PI processes to find the right people from the public to be involved. There are several measures and methods for finding influential nodes in a network. Different types of centrality (including degree, betweenness, and eigenvector) are among classical tools for this purpose. In particular, researchers in the domain of construction have widely used these metrics to analyze project networks. In a previous study, we tested the methods normally used in ranking webpages (such as Hyperlink-Induced Topic Search: HITS, and PageRank) for this purpose (Nik Bakht & El-diraby, 2013a). Such methods typically consider not only the number of followers (quantity) for an individual, but also their level of importance (quality). It was shown that given the size and level of complexity of the IDN, such methods provide more precise measures to analyze influence patterns in IDNs. Having a full list of top influential nodes, the team-building processes can target those who represent different vested interest in the project and at the same time have higher levels of influence on other followers of the project.

Studying communities of interest—Social networks are typically composed of clusters of nodes called communities. Nodes within each community are densely connected to each other, and are sparsely connected to nodes from other clusters. Existence of communities has roots in the social behavior of community members. During formation and evolution of a social network, people are interested in joining groups in which not only they have more friends, but also their friends are more closely connected to each other. Therefore, communities of a social network typically form around commonalities and shared interests. Detection of communities, therefore, not only can classify those who have interest in the project, but also can classify vested interests with respect to the project. On top of community leaders, nodes with strategic and inter-disciplinary positions can be great sources of feedback and/or innovation. These nodes can also assist in regulating cross-community relations. Pivot nodes which are at the intersection of multiple communities, and bridges that connect two communities to each other are examples of such nodes (Fig. 3).

Fig. 3
figure 3

Eglinton Crosstown LRT project discussion profiles in the four dimensions of sustainability

Full size image

3.2 IDN as Network of Ideas

As shown above, an IDN can be modeled and studied as a graph of social identities, connected to each other through direct social linkages. On the other hand, connectivity among people in an IDN can be defined through the ideas they support (or oppose!). Monitoring ideas discussed and analyzing them can also help add meaning to network clusters, detect vested interests, and monitor opinion dynamics among projects’ followers. Computational Linguistics (CL) can support these aims. In order to ‘understand’ the content of discussions and to follow their dynamics, domain-specific tools may be required. As mentioned before, complete analysis of the big data over IDN will be the intersection of SNA and CL; the latter can help to interpret results of the former, and also to answer the last two out of the four questions we introduced earlier.

Profiling people and labelling communities—As it was addressed in the previous part, detecting communities of followers in the graph of IDN can provide decision makers with a good insight for team-building and problematization process. To make sure the PI program has involved the right people and to guarantee that all different groups of project followers have their voices in process of decision making; different groups must not only be detected (through community detection), but also they should be ‘labeled’ based on their backgrounds and roles with respect to the project. The labeling requires a better understanding on the typology of followers. Users normally describe themselves in their online profiles, referred to as ‘bio’ or ‘description’. Text mining can help to classify users’ descriptions in each community of the IDN. In our previous study, we used an information retrieval measure called Term Frequency-Inverse Document Frequency (TF-IDF) which is usually used in topic detection to achieve this goal (Nik Bakht & El-Diraby, 2013b). This method scores up terms with high frequency in the set of descriptions for one community, and at the same time scores down terms which are common across multiple communities. By detecting discriminating terms for each community, this method crystallizes the sets of buzzwords for descriptions of each community’s members. Semantic clustering of these terms can then automatically label each community and provide decision makers with a layout of the social composition around the project.

Detection of core interests through profiling communities of interests—If labeling communities is clustering ‘people’, detecting core interests can be known as clustering ‘ideas’ discussed. As mentioned, shared interests and common values bring people with different backgrounds together in form of a community. Detecting cores of interest among project followers is an important insight that analysis of IDN can provide decision makers with. Any solution offered by decision makers must try to address as many core interests as possible. Shared interests can be detected either by semantic clustering of ideas discussed as suggested by Steinhaeuser and Chawla (2008), or in a reverse format by labeling common opinions expressed within each community, similar to Nik Bakht and El-diraby (2013a, 2013b). In order to achieve this goal, text mining must target discussions supported by members of each community. The result can also detect the topics discussed within and across various communities and portray the ‘social dialogue’ around the project (Nik Bakht & El-Diraby 2013b).

Detection and evaluation of opinions—As mentioned, dynamics in structural properties of the IDN can be an important indicator for monitoring the public outreach. Dynamics of the IDN however, is not limited to the followers’ interconnectivity, but also at a more sophisticated level, it includes dynamics of opinions they support and express. By defining ‘opinion’ as a combination of the ‘subject’ discussed, and ‘sentiment’ of the discussion, each discussion over social media can be modeled as an instance of opinion expression. Detection of the opinion will consequently require automatic classification of the subject, and the sentiment of the discussion. Such classifiers can be trained using NLP and machine learning techniques among other methods. Detected opinions for one project can then be aggregated and analyzed statistically over the time, to identify patterns of dynamics in the social opinion. Monitoring formation and evolution of opinions enables decision makers to detect bottlenecks in communication with the public. It also allows detection of social alarms from analysis of the opinion dynamics. Since offline social opposition most of the time lags the online declaration of dissatisfaction, detecting online alarms will give the official decision makers enough time to change the decisions appropriately, or to apply timely policies to prevent formation of snowballing social opposition and to reduce the risk of failure in such projects before it is too late.

4 Project Discussion Profile

By selecting a particular context for analysis of discussions over IDN, results of SNA and lexical analysis can be aggregated to form the profile of online discussions for a particular project. This not only classifies major vested interests and highlights followers’ positions (in terms of being proponent or opponent) over the time, but also synthesizes them according to the network value of the followers. The context is scoped as a semantic space with a certain number of semantic clusters as dimensions of the space. In this space, each data point (e.g., each tweet in the case of Twitter, or each comment in case of Facebook or forums) is represented as a vector. Every entry of such a vector corresponds to one semantic cluster (topic) and takes numerical values only if that topic is covered by the discussion. Sign of entries highlight sentiment of the discussion; for this purpose, either a binary method (similar to Sousa, 2005: opponent −1, proponent +1, or neutral 0), or a fuzzy system (similar to Olander, 2007 to highlight intensity of support or disapproval) can be used. Signed binary vectors resulted in this way represent opinions as subject-sentiment dyads.

In order to model the network value of discussions, opinion vectors must be connected to the influence level of people who express them. This can be numerically represented as the product of the binary vector of opinion and impact factor of the utterer. The latter can be evaluated as a normalized factor from analysis of influence in IDN through measures such as centrality or PageRank. Weighted vectors in each space can then be combined and the resultant will aggregate all data points for a project in form of the project discussion profile.

As an example at the project level, Fig. 3 shows the discussion profile for Eglinton Crosstown LRT project in within a time-span of 18 months. Sustainability was selected as the context of analysis and its components (environmental, social, economic, and engineering) were forming dimensions of the semantic space. As it is seen, public satisfaction in online media reaches its minimum in December of 2012; this is when decision makers revised the environmental assessment study and removed two major stops from the plan. However, after public consultations and listening to community’s feedback, the plan was modified and the stops were returned back in May 2013. At this point, the social dimension hits its maximum level. Also, as it is seen, while in many cases engineering, environmental, and social sustainability have similar trends (with different amplitudes), economic-related opinions move in an opposite direction in most of the cases. By the beginning of construction, in summer 2013, negative sentiment with respect to the economic aspect is at its maximum. Figure 3 also gives a range for opponent and proponent discussions at each snapshot.

Project discussion profiles can be known as a formal distillate of online media discussions. This can be used as a measure to evaluate outcomes of different communication strategies and control the public consultation procedure among other applications.

5 Concluding Remarks

Trends of change in profile of decision makers for construction and development of the urban infrastructure improved them from individual decision makers to a hierarchy of technical decision makers, and recently to a network of technical and official decision makers. Such an evolution, along with modern trends of prosumerism and knowledge epidemiology over Web2.0 is now calling for another shift to a heterogeneous network of official decision makers and public decision contributors. Public involvement programs should be upgraded to pave the road for such a shift. Social web can help decision makers of a project to engage important sectors of the community and develop bidirectional communication strategies to not only educate the public regarding the project and related decisions, but also build trust and promote a culture of collaboration through formation of IDNs. At the same time, similar to many other domains, an IDN can be a great source of prosumers’ knowledge and innovative ideas. In fact, the advantages of using Web 2.0 in community engagement go beyond improving the quality of communication with local communities; it can build the foundation for reaching solutions which are more robust and more innovative. In addition to harnessing innovative ideas, the core interests extracted from IDN discussions can be embedded in the ultimate solution offered by the decision making process. This can create a true sense of ownership of an urban project by local communities, which is a key factor to the project success from different aspects including stakeholder management and social sustainability.

Some of core contributions of IDNs to the public consultation process were addressed in this chapter. On one hand, structure of the IDN as a social network can provide decision makers with some insight about the performance of their public engagement practices. This structure also includes important information regarding influential nodes among the internal and external stakeholders. On the other hand, semantic analysis of the user-generated content can lead decision makers towards a better understanding of vested interests and social concerns/values. In general, IDNs can help to direct the decision making towards a more pluralistic process, rather than a pre-planned project.

In order to take advantage of IDNs, more research is required in at least two major streams: logistics of the IDN, and behavioral issues. From the logistics point of view, developing context-aware tools and context sensitive mechanisms is necessary to specifically support the realm of infrastructure. This ranges from creating topic/sentiment classifiers for infrastructure-specific discussions, to defining performance measures for assessment of public involvement in a project through its IDN. The latter can facilitate monitoring of bottlenecks in communication with the public, and detection of public dissatisfaction alarms with respect to a project or its certain aspects. At a behavioral level, different phases in the lifecycle of projects and their IDNs must be studied to develop models for predicting behavior of IDNs in different conditions. Such models will be helpful in creating online (and maybe offline) communication strategies with the public in infrastructure projects. There are issues and barriers which may postpone the applicability of IDNs in practice. The difference between offline and online attitudes of many users as well as issues such as multiple identities, security and trust are among other examples of this type. Collaboration between practitioners and researchers will be required to address such issues and solve them in the future.