=Paper=
{{Paper
|id=Vol-2638/paper24
|storemode=property
|title=Weighted networks in socio-technical systems: concepts and challenges
|pdfUrl=https://ceur-ws.org/Vol-2638/paper24.pdf
|volume=Vol-2638
|authors=Zeljko Stojanov,Jelena Stojanov,Gordana Jotanovic,Dalibor Dobrilovic
|dblpUrl=https://dblp.org/rec/conf/iccs-de/StojanovSJD20
}}
==Weighted networks in socio-technical systems: concepts and challenges==
https://ceur-ws.org/Vol-2638/paper24.pdf
Weighted networks in socio-technical systems:
Concepts and challenges
Z Stojanov1 , J Stojanov1 , G Jotanovic2 and D Dobrilovic1
1
Technical faculty ”Mihajlo Pupin”, University of Novi Sad, Serbia
2
Faculty of transport and traffic engineering, University of East Sarajevo, Republic of Srpska,
Bosnia and Herzegovina
E-mail: zeljko.stojanov@uns.ac.rs
Abstract. Socio-technical systems join together humans and technique. Basic concepts and
principles of socio-technical systems are outlined, as well as weighted networks as the appropriate
mathematical models. Particular examples of socio-technical systems with various usages of
weighted networks in domains such as airline connection networks, scientific collaboration,
social networks, software engineering and urban traffic are presented. Comparative analysis
of the selected examples is outlined with the focus on the nature and functions of nodes, links
and weights. The identified challenges, such as the creation of networks, evolving nature of
systems, and the need for multidisciplinary teams in the system design are discussed.
1. Introduction
The concept of socio-technical systems was established in the fifties of the last century
in order to envelope complex relation between humans and technical systems in industry.
The socio-technical concept has developed in terms of systems because it is concerned with
interdependencies within an organization, as well as in terms of open system theory since it is
also concerned with the environment in which the organization exists [1].
Socio-technical system is a collection of social and technical elements engaged in purposeful
goal-directed manner for achieving specific behavior [2]. Due to the intrinsic complexity
of contemporary socio-technical systems, their understanding requires a broad knowledge of
both technical and social disciplines. Therefore, perhaps nobody has enough knowledge and
understanding of socio-technical systems. Engineers and technicians usually ignore social aspects
of their work, while social scientists know very little about technology and technical systems.
Ropohl [3] suggested that it is important to shape both the technical and the social aspects
of work in a way that preserves both humanity and efficiency. The complexity of relationships
between the technical and social aspects of an organization reveals that it is not sufficient to
inquire them separately, requiring new approaches for researching socio-technical systems. It
has become clear that social and technical aspects should be considered interdependently, which
requires dual focus and joint optimization of these aspects within an organization [4]. However,
this optimization requires a wide range of knowledge and techniques because socio and technical
parts do not behave the same way, making non-linear relationships and behavior common in
Copyright c 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0
International (CC BY 4.0).
these systems. In this course of thinking, Fuenfschilling and Truffer [5] introduced the concept
of the socio-technical regime for addressing long-term changes in various sectors in industry,
which actually denotes co-evolution of institutional and technological aspects over time.
Business and overall organization of contemporary organizations depend on IT and software
systems that provide meaningful communication of humans and equipment, as well as data
processing. This situation emphases the importance of software systems in complex socio-
technical systems that exist in organizations, increasing the significance of software engineering
methods, tools and techniques for overall performance of socio-technical systems. Software
systems are used in all segments of human life, such as education, health, transport, industry,
communication, entertainment, etc. Therefore, activities related to the software life cycle cannot
be treated as isolated activities regarding more complicated systems that use software. Actually,
software engineering activities should be treated as an intrinsic part of systems engineering
processes. Sommerville [6] proposed the socio-technical systems stack that presents socio-
technical systems with layered structure, as it is shown in Figure 1. Software engineering part
in socio-technical systems includes operating system, communication and data management,
application system and business processes. System engineering envelopes equipment, software
engineering part and organization.
Figure 1. Layered structure of socio-technical systems
This model stresses the importance of having a holistic view of systems, which helps in
observing complex relationships between different parts, whether technical or social. This view
on systems assumes that, when designing one part, it is important to observe the influence on
other parts. For example, when designing a specific equipment in the industry, it is important
to carefully consider software elements that will assist in using and tracking equipment use, as
well as people that will operate both software and equipment. Since socio-technical systems
are situated within organizations, the following organizational factors affect their design and
operation [6]: organizational changes - changes in overall organization or changes in decision-
making structures, process changes - changes in internal process organization or interprocess
communication, and job changes - changes in description of work that may render some skills
obsolete and require some new skills.
Complex design of socio-technical systems imposes an abstract and low level framework.
Graph theory, as a mathematical discipline, produces models for socio-technical systems that are
simple enough but analyzable [7]. Mathematical properties of a model are good representatives
of a socio-technical systems’ characteristics. Moreover, the model enables introduction of various
metrics useful for detailed determination of socio-technical systems.
This paper outlines the development of socio-technical concepts and principles in the next
section. The third section presents the basics of weighted networks and their use in socio-
technical systems. The fourth section presents a comparative analysis of the selected examples
and discussion of identified challenges. The last section contains conclusions and further research
directions.
2. Development of socio-technical concepts and principles
Review of literature dealing with socio-technical systems, from the early work of Trist and
Bamforth [8] to recent studies [2, 9], helps in refining the basic and enduring socio-technical
principles. The first identified basic principles are: responsible autonomy, adaptability, and
meaningfulness of tasks. These principles should ensure the best possible performance at both
individual and organizational levels.
Further research on socio-technical systems has been directed toward their implementation
in various fields of society and economy, which resulted in refinement of the basic principles.
Albert Cherns [10] distilled experiences on socio-technical system design principles from the
early period of research, and provided more detailed list of principles:
• Compatibility. The design process should be compatible with objectives (self-modified,
adaptive to changes), which assumes active participation of people in the design of their
jobs.
• Minimal critical specification. This principle assumes precise specification of what is
essential for the system, including minimal specification of tasks, allocation of tasks to
jobs and jobs to roles. Only essential objectives should be specified.
• The sociotechnical criterion. Identification and prevention of variance, or unpredictable
events, that affect system functioning should be planned and included in system design.
• The multifunctional principle. Design should consider option that system functions can be
performed in different ways, which requires people with different skills that can perform
highly specialized tasks.
• Boundary location. Boundaries are used for grouping people and activities based on
technology, territory and time, which requires careful management of boundaries between
groups, departments, and the whole organization.
• Information flow. All necessary information should be provided at the place where tasks
execution requires. The right information should be provided to all people and groups,
which support learning and better control of processes.
• Support congruence. The system should support all behaviors that an organization requires.
• Design and human values. Organization should enable high quality of work, which is usually
subjective judgment based on psychological assumptions.
• Incompletion. Design is iterative process, which means that all decisions may lead to new
design decisions and changes in design.
From its beginning devoted to productivity enhancement in various machine-driven industries,
socio-technical thinking has evolved towards contemporary systems based on knowledge work
and capabilities of workers. This evolution of socio-technical thinking affected design of
contemporary systems, with focus on agility in redesigning systems due to dynamic changes
in business and society [9]. Based on practical experience, Hirschhorn et al. [11] suggested that
in the time of mass customization in the economy, the concept of meaning should replace the
concept of autonomy (worker and task), which leads to more productive workers and learning
organizations that are more responsive to the challenges. This evolution of contemporary socio-
technical systems is primarily shaped with development and use of information technology,
Table 1. Principles of socio-technical systems design
Principle Description
Wholeness A set of activities in a work system should make a functioning
whole, rather than a collection of individual jobs.
Teams The work group should be considered a central unit in an
organization.
Process control All variances (problems or deviations from expectations) should
be identified and handled as soon as they occur.
Self-direction System should rely on internal regulation rather than on
external ones.
Multi-skilling System should be based on redundancy of functions rather than
on a redundancy of parts.
Discretion The discretionary component of the work is essential to the
success of the system.
Joint optimization System optimization should include both individuals and
technical parts.
Adaptation System should easily adapt to changes, which is based on
individual and organizational learning.
Meaning Individual jobs should be designed to support learning, some
level of decision making, and should be socially recognizable.
Incompletion Due to constant change of environment and organization, job
design cannot be treated as finished.
digitization and advanced technologies. The principles of socio-technical system design
summarized by Pasmore et al. [9] are presented in table 1
The next technology movement that significantly has affected the development and
implementation of socio-technical systems is the spread of IoT and smart technologies. Concept
of smart cities is widely researched from a technical perspective with focus on urbanization
problems, while the social component is usually neglected [12]. Shin and Jin Park [13]
offered a contextualized socio-technical analysis of IoT, providing insight into its challenges
and opportunities, which helps in understanding how IoT can be designed and situated within
human-centered contexts. The study reveals implications on conducting design according
to a combined technical and social views and how to structure design processes in multi-
disciplinary teams. A user-centered approach is required to map out the IoT design space, to
understand how to avoid negative individual and social consequences and to support acceptable
and useful information sharing. Since smart cities are based on technology which affect
people, organizations and society, they can be observed as socio-technical systems. Smart
cities include variety of networks and infrastructures, such as city services, business, transport,
communication, education, health-care, water and energy supply. One of the most challenging
aspects of smart cities is smart transportation aimed at solving traffic congestion problem or
pollution problems by providing the most relevant information to participants in traffic [14].
Graph theory is commonly used for modeling and analyzing various types of networks that
models socio-technical issues in local and urban traffic systems [15].
The vast majority of modern systems can be classified as socio-technical systems. Research
and implementation issues go beyond general principles of socio-technical systems and they have
to rely on the realistic context where the systems are being introduced. Since these systems are
based on people and technological infrastructure, various networks may be identified that reflect
relations between people, people and technical objects, and between technical objects. Networks
containing people and technical objects have been researched in domains such as education
[16, 17], behavioral change for energy saving [18], sustainable urban water management [19, 20],
health care systems focused on e-Health interoperability [21] or personal health technology [22],
social networks [23], resource sharing in social communication [24], large-scale socio-technical
systems on the Web [25], air traffic management systems [26], transit service networks [27],
roadside air quality control [28], traffic forecasting [29], work of military staff in steady-state
and crisis threat scenarios [30], or software development projects [31, 32].
3. Weighted networks in socio-technical systems
3.1. Terminology
Weighted graph is a mathematical notion, abstract enough to model wide variety of real
phenomena. Structures of modeled phenomena imposed more popular name, weighted network.
It consists of objects of two kinds: vertices (nodes, actors) and edges (links, connections), and
a function that quantifies a feature of edges.
More precisely, weighted graph/network is a triple (V, E, f ), where:
• V = {vi | i ∈ I} is a set of vertices. Their amount |V | is graph’s order. The vertices
represent real objects (cities, persons, devices,...) and are usually labeled for practical
reasons by v1 , v2 , . . .. I is a set of indices that enable distinction of vertices.
• E = {ea | a ∈ A} is a set of edges. A is a set of indices that enable distinction of edges.
Amount of edges |E| is graph’s size. The edges represent a relation among nodes (roads,
collaborations, cable links,...). Each edge ea is a two-element set of vertices ea = {vi , vj }1 .
The edge ea is said to be incident with the vertices vi and vj . And more, the vertices vi
and vj are adjacent.
• f : E → [0, ∞] is the weight function, which assigns to each edge ea a number, its weight
that w(ea ) =: wij . The latter label contains information of the vertices connected by the
edge, so it is commonly used. In the case that all edges have the same weight, the weight
function is trivial, hence can be omitted and the weighted graph becomes the graph (V, E).
Vertices and edges are mathematical terms, but actors and connections, nodes and links, and
agents and links are commonly used in social and technical systems.
It could be of interest to consider only a part of the graph, some of the vertices V1 ⊂ V
and some of the edges, with overtaken their weights, incident with elements of V1 contained in
E1 ⊂ E. Obtained graph is named as a subgraph of the graph (V, E, f ).
Topological properties of weighted graph which are of the interest in modeling of socio-
technical systems will be presented.
The number of edges which are incident with a vertex vi is its degree, d(vi ). A set of vertices
that are adjacent to a vertex vi is the set of neighbors, N (vi ), and it contains exactly d(vi )
elements.
Path is a sequence v0 , e1 , v1 , e2 , . . . , vk , where each edge ea , a = 1, . . . , k is incident with
vertices va−1 and va . As the path is fully determined with the vertices, the edges in the sequence
are commonly omitted. The length of the path is the number of edges it contains. Exclusively
seen, the path is a subgraph.
1
Edge can also be related to couple of vertices ea = (vi , vj ). In that case, orientation of edges is emphasized and
the graph is named as the weighted directed graph.
Two vertices vi and vj are connected if there is a path with the first element vi and the last
element vj . The length of the shortest path between vertices vi and vj is their distance, d(vi , vj ).
In weighted graph, a path between any two vertices vi and vj has, besides the length, its weight,
obtained as a sum of the weights of all edges in the path. The shortest path in weighted graph
implies the minimum weight.
The graph is connected if each two vertices are connected.
Importance of a vertex in connected graph reflects its significance in the connectivity, and is
presented by the centrality of vertices. As the shortest path has different meaning relating to
the weights, there are two kinds of centralities:
• centrality is determined by the notion of eccentricity of a vertex vi , denoted by ecc(vi ), it
is the maximum distance between vi and any other vertex in the graph. The maximum
eccentricity is the graph diameter, and the vertices possessing it are central vertices. Their
collection is called the graph center. So, a vertex v is the central one if
v = argmin ecc(u).
u∈V
Analogously, the minimum eccentricity is the graph radius, vertices possessing it are
peripheral vertices and their collection is the graph periphery.
• betweenness centrality measures appearance of each vertex in all shortest paths in the graph,
X σst (v)
g(v) = ,
σst
s6=v6=t
where σst is the number of shortest paths from vertex s to vertex t and σst (v) is the number
of those paths that pass through the vertex v. Betweenness orders vertices, and the central
ones are those with the maximum betweennes.
Weighted networks can be directed, which means that each edge has a starting and target
vertex, i.e. the edge is oriented. In directed weighted networks vertex has a degree, but also has
input degree and output degree. These basic differences produce variety of new possibilities in
their use.
More details on graph theory can be found in [33].
3.2. Examples
Barthélemy et al. [34] presented two case studies, the airline connection network and the
scientific collaboration network, which are representatives of critical infrastructure and social
system respectively. Characterizations of corresponding weighted networks are determined by
use of statistical tools. The empirical results confirm the importance of the distributions of
the various network’s quantities and the existence of weight-topology correlations. The weights
relevance is emphasized and the network modeling of complex systems has to go beyond the
topology. Produced weighted networks support large heterogeneity in the capacity and intensity
of the connections that exists in real networks. Considered model relies on dynamical coupling
of topology and weight through the particular local properties: strength of node, weighted
clustering, weighted assortativity and disparity. The model supports the weight dynamical
evolution occurring when new vertices and edges are introduced in the system. Hence, the
evolution of the network can be inspected analytically.
The hidden metric space weighted model introduced in [35] relies with network’s topology,
its weights and underlying metric space. Based on the metric space, the model produces the
weighted architectures of real complex networks. Therefore, the weights of the links contain
information of the interest in finding more accurate embeddings of real networks. The underlying
metric space governing the network topology determines the weighted structure of the network.
Framework of the model can be seen as a novel generalization of the gravity concept, and
it could reproduce the existence and the intensity of interactions within the network. The
results are widely applicable to very different domains: biology, information and communication
technologies and social systems. They can also be useful for navigation and searching protocols
which take into account both the existence and the intensity of a connection. The evidence
examples are taken from empirical data sets.
Software development process is considered as a socio-technical system with four types of
objects [36]: agents are team members and developers, knowledge represent skills possessed by
the team members or required for task solving, tasks are problems to be solved, and resources
relates equipment used by agents and required to complete tasks. A network of a socio-technical
system represents actual communications and dependencies in software development process
(task dependency on other tasks, resources and skills; developer connections with other agents,
their skills and equipment they can use). Socio-technical congruence measures interactions
of agents, but indirectly, which could be used for proposing weights of agents’ interactions.
For example, by considering the following relations: agent-task (AT), knowledge-task (KT),
and agent-knowledge (AK). Particular relations within the network are presented as adjacency
matrices of subgraphs containing specific partitions of objects denoted by AT , KT and AK,
respectively. Further, knowledge-dependent congruence requirement matrix CR is determined,
CR = AT × (KT )T × (AK)T . Socio technical congruence indicates how the obtained matrix
differs from adjacency matrix AA,
dif f (CR, AA)
Congruence(CR, AA) = 1 −
|CR|
where dif f (CR, AA) and |CR| are numbers of nonzero elements in the matrices CR − AA and
CR, respectively. Data set for the experiment contains data from version control system for
tracking changes in project realization, survey data of team members knowledge and skills, and
data from students’ discussion forum. Observation of socio-technical congruences on the time
base (during some weeks) has strong positive impact on the software development process.
Mavromoustakis and Karatza [24] used weighted graphs for modeling resource sharing in
the opportunistic wireless environment by considering social parameters. Communication was
presented by two socio-technical layers: Social Connectivity Layer (SCL) represents social
interactions of users that use a certain platform, and Physical Connectivity Reflection Layer
(PCRL) represents physical connectivity in wired or wireless infrastructure. Resource location
is determined with social associations and interactions among individuals. Weighted undirected
connected graph is used for modeling random movement of devices. The weights are determined
by the intensity of social interaction between two users (nodes in network). Probability of
movement from one node to the next node (one of the neighboring nodes) is calculated by using
the weights of all edges associated to the current node. Since mobility is unrestricted in a
wireless opportunistic network, the network topology dynamically change over time. Proposed
resource sharing schema was evaluated through experimental simulation, which revealed that
the reliability of successful packet delivery ratio is not aggravated by the mobility factor.
Goggins et al. [37] presented an analytic framework, based on Group Informatics Model [38],
which improves awareness of virtual group dynamics by analyzing electronic trace of individuals’
interactions and enables identification of emergent groups. The framework is based on the
following items, which can be customized for specific cases: domain of study is a description of an
application domain; contextualized interactions represents the nature of the collected interaction
traces with their contextual attributes; weighting procedure is determination of weighting that is
applied to the contextual attributes; collaboration opportunities network represents the nature of
the network resulting from weighting for the selected domain; aggregation procedure represents
the aggregation algorithms and parameters used for making groups visible; emergent groups
are the group structures emerging from the aggregation; and group context is description of
contextual information associated to each identified group. Nodes in the network represent
people and artifacts, while weights represent intensity of interactions between the nodes. The
framework is applied in two domains: (1) asynchronous online learning for discourse around
ideas in discussion forums, and (2) open source software development for software developer
interactions with knowledge-intensive, technical artifacts of the software product.
Ferrara et al. [39] used a directed weighted graph, named Relational Instagram Network
(RIN) for modeling and analyzing social interactions at Instagram, focusing on follower-followee
relations and user communication by means of posts/comments. The analysis focuses on the
structural characteristics of the Instagram network, the dynamics of production and consumption
of content, and the social tagging for presenting users’ interests in the content. An Instagram
data set was collected by querying the Instagram API (e.g. features such as the users API, the
relationships API, the media API, the comments and the likes APIs, and the tags API,), which
is publicly available developer library. Nodes in RIN present users, edges present relationships
of the form follower-followee, while edge weights were calculated proportionally to the number
of likes and comments.
Islam et al. [40] proposed a new heuristic algorithm based on underlying dependency structure
in procedural programs for migrating software design using hierarchical clustering. The proposed
approach produces candidate classes for an object-oriented design based on given procedural
code. A procedural program is presented by a new type of graph, Weighted Data Call Graph
(WDCG), in which a similarity measure called Weighted Distance Matrix (WDM) is used for
computing similarity between two nodes in WDCG. Weights are presented with data structure
Entity Map, in which all data and function nodes have calculated weights for each relationship
in the graph. WDCG contains two types of nodes: Function Nodes for representing functions,
and Data Nodes for representing data. WDCG contains four types of edges: Self-Edge connects
a node to itself, Call Edge connects a function that calls another function, Read Edge connects a
function that reads data, and Write Edge connects a function that writes data. The weight of an
edge in WDCG depends on the relationship represented by the edge. The weight of a self-edge is
always 1, while weights of call edge, read edge and write edge are from a predefined set of weights.
Migration approach contains the following steps: generation of a WDCG from procedural code,
hierarchical cluster analysis of the WDCG, identification of desired clusters based on migration
objectives, and transformation of clusters to classes. The proposed algorithm was tested on a 5
procedural programs written in C language.
Traffic regulation system based on measured air quality data is presented by El Fazziki et al.
[28]. The system uses big data and intelligent systems concepts for traffic regulation based on
real-time and predicted air pollution indexes. The system helps in reducing vehicle emissions in
the most polluted sections of the road. Data collected from sensors, contextual data, and road
network available data are used for constructing a weighted graph that presents a road network
in which intersections characteristics are associated with the nodes and roads characteristics are
associated with the edges. Edges weights evolve based on the pollution indexes. The system is
tested in Marrakech city with real-time data.
Kalloniatis et al. [30] proposed empirically derived Situation Awareness Weighted Network
(SAWN) model, which is based on Situation Awareness model proposed by Endsley [41] involving
perception, comprehension and projection, and Distributed Situation Awareness model proposed
by Stanton et al. [42] indicating that situation awareness exists in a social and semantic
network of people and information objects. The SAWN model represents a target socio-technical
system as a weighted semi-bipartite network of interactions between people (human nodes) and
information artefacts (product nodes). Links in the network are relations between people and
between people and artefacts. The weights of the links in the network represent the levels
of social awareness that individuals acquire, which is rendered in the network by three levels
proposed by Endsley through the color scheme: Green for Perception, Blue for Comprehension,
and Red for Projection. The SAWN method is illustrated with empirical data from a case study
related to work of Australian military staff in steady-state and crisis threat scenarios.
Binzagr and Medjahed [43] proposed a crowdsourcing approach for recommending mashup
teams, which is based on analysis of online developer community and API directories at
StackOverflow. The approach uses three data structures based on data from StackOverflow:
interest table (developers interests in specific APIs), reputation table (developers reputation
based on comments and replies) and sociometric graph (connections among developers).
Interactions between developers are comments and replies, which are modeled as a weighted
graph (sociometric graph). Developers are nodes in the graph, interactions represent the edges
between the developes, while the number of interactions between the developers represents the
weights of the edges. The authors implemented a CrowdMashup prototype software in Java, in
which mashup administrators define queries for forming a team by defining the number of teams,
the number of members in each team and a list of required APIs. The inputs for software are the
sociometric graph, the interests table and the reputation table. The implementation software is
evaluated in an experiment with real data from StackOverflow and programmableWeb.
Pattanaik et al. [44] presented a smart technique for congestion avoidance by estimating
real-time traffic congestion on urban road networks, and based on that the technique predicts
an alternate shortest route to the destination. The congestion on different roads is estimated by
K-Means Clustering Algorithm, while Dijkstra’s Algorithm is used for predicting the shortest
route. The system identifies a road network from Google Maps and convert it to a weighted
graph. Nodes represent intersections, while edges represent roads. Initial weights on all roads
are set to 1, but based on real-time data on the number of vehicles on a particular road, the
weights are updated periodically. Updated weights are used for estimating the travel time and
for calculating the shortest path to the destination based on real-time congestion of roads in the
network. The approach was tested over several road maps of New Delhi and simulation results
indicate reduction in travel time if road congestion is used for updating the weights of the edges
in the network (the roads in the city).
Cepulis and Niu [32] conducted a software requirements traceability case study aimed at
creating socio-technical patches for information foraging. Software developers usually conduct
a search task based on a predefined patch of information, which in requirements traceability
relates to traversing a combination of technical artifacts and social structures. The authors
proposed requirements socio-technical graph for analyzing human-human and human-artifact
relationships that connect a traceability question to an answer. The graphs contains three
node types: people, artefacts and questions/answers. The weights of the relations in the graph
depend on the knowledge that possesses a person providing an answer to the posed question.
The method is tested on four open source projects from the Apache software foundation and
the JBoss.
Ikram et al. [45] presented an approach supported with DaDiDroid, an Android malware
app detection tool for examining graph features of mobile apps, by building weighted directed
graphs of the API calls. The approach detects the presence of malware code in Android apps by
analyzing graph features. DaDiDroid performs static analysis of apps and extracts call graph
among APIs. Nodes in the graph present API libraries, a directed link present call of a method
from another API, and weights represent the numbers of times a methods from other APIs are
called. Effectiveness of DaDiDroid is examined in an experiment with six publicly available data
sets.
4. Discussions
This research on the use of weighted network in socio-technical systems revealed several questions
and challenges related to network creation, types and nature of nodes, edges and weights, network
evolution and multidisciplinary nature of research of socio-technical systems that need discussion.
4.1. Comparative analysis of presented examples
Due to the complexity and diversity of real socio-technical systems, they can be presented with
homogeneous and heterogeneous networks [46]. In homogeneous network, all elements in the set
of vertices are uniform, while in heterogeneous network a clustering or hierarchy exist in the set
of vertices or in the set of edges.
In homogeneous networks, nodes are commonly of the same type and have the same function,
such as intersections in road traffic networks [28, 44], airports in airline connection network and
scientists in scientific collaboration networks [34], API libraries in software architecture network
[45], or people in Instagram social network [39]. In heterogeneous networks, nodes and links
can be of variety types depending on the domain modeled with the network. In many cases,
vertices can present entities of different types in the same network, such as persons, resources
and knowledge [32, 36], data and functions in software migration project [40], or humans and
information product artefacts [30].
Relations that are represented by the edges (links, relations) can be given explicitly or
implicitly, and they present actual or virtual links among vertices (nodes). Links can connect
the nodes of the same type in the network, or can connect nodes of different types, depending on
the network nature (the nature of domain problem analyzed by the network). The common case
are links representing social interactions between people in network [24, 39, 43]. In considering
social aspects of socio-technical systems, it could be necessary to additionally examine people or
their communications and collaborations for more accurate modeling of links in created networks
[34, 37]. In complex socio-technical systems with different types of nodes, links can also represent
interactions among people and artefacts [32, 37]. However, in some cases links connect only
technical artefacts, such as links representing roads or their characteristics [28, 44], links between
functions and data in software systems [40], or calls between different API libraries in mobile
software systems [45].
A weight function of a network models differentiation among the edges. They can be obtained
by using statistical observations and analysis [34, 35, 39, 43, 45], by measuring real-time data
[28, 44], by analyzing and quantifying triangulated data from the interaction context [37], by
examining knowledge level of people in the network [32], or by using models such Situation
Awareness model proposed by Endsley [41] that is used in [30]. In addition, weights can be
predefined [40].
4.2. Challenges
The first challenge relates to creation of a network for a particular system. The creation of a
network as mathematical model for real phenomena imposes exact and distinct determination
of objects which would be represented as vertices. Statistical tools produce distribution of
vertex degrees and distribution of edge weights [34, 43], which are particularly important in
large networks. Further, creation of weighted networks for analyzing soci-technical networking
systems can be done by using publicly available developer libraries such as Instagram API [39],
or by analyzing existing video materials such as Google Maps for extracting urban road networks
[44]. In software development, systems networks can be created by extracting API libraries and
calls from source codes of applications [45], by analyzing traceability of questions and answers in
specifying software requirements [32], or by analyzing connections among developers at software
development forums such as StackOverflow [43].
As socio-technical systems are evolving systems it is of interest to embed dynamics into the
weighting network structure or into the network topology. Parameters and characteristics of
artifacts in systems can change over time, and this should be modeled in weighted graphs with
weights that can change [28], or with topology that can change [24]. For example, dynamical
evolution occurs when new vertexes and edges are introduced in the existing system [34], when
network topology change is natural like in wireless opportunistic network [24], or when weights
are periodically updated based on real-time data measured in the real system [44]. According
to Rouse and Serban [47] it is essential to understand how causality and complexity influence
the nature of change in complex socio-technical systems. Management of network evolution can
be used for long-term analysis and control of socio-technical systems, as well as for prediction
of their further behaviour and states.
This review of examples and socio-technical systems development reveals that weighted
networks can be used in many domains of human life and work. Due to the complex
nature of real socio-technical systems, systems’ analysis and creation of models require forming
multidisciplinary teams with experts from technical, natural and social disciplines, who can
observe all relevant aspects of the system [5, 48], and that will have deep knowledge on the
context to be modeled as a socio-technical system [38]. Further, it will require the use and
mixing of multiple quantitative and qualitative research methods [49] for acquiring a more
comprehensive understanding of socio-technical systems.
5. Conclusions
Presented concepts, examples and challenges related to socio-technical systems reveal their
importance for understanding, modeling and analyzing complex behavior in real systems with
humans and technology. The contribution of this paper is a comprehensive analysis of the global
aspects of weighted networks in highly diverse socio-technical systems, as well as the discussion
of the challenges in tackling with systems’ design and dynamics.
Systematic literature review is planned as one of the next directions for further work. This
systematic review will provide more comprehensive review of challenges in the selected areas,
such as urban traffic or software development, and identify issues for further research.
Acknowledgment
Ministry for scientific and technological development, higher education and information society
Republic of Srpska supports this research under the project ”Smart system based on IoT
technology designed for monitoring of traffic air pollution”, contract number 19.030/3-2-25-2/19.
References
[1] Trist E 1981 Occasional paper No. 2, Ontario Ministry of Labour ISSN: 0227-4426
[2] Walker G H, Stanton N A, Salmon P M and Jenkins D P 2008 Theoretical Issues in Ergonomics Science 9
479–499
[3] Ropohl G 1999 Techné: Research in Philosophy and Technology 4 186–194
[4] Fox W M 1995 The Journal of Applied Behavioral Science 31 91–105
[5] Fuenfschilling L and Truffer B 2014 Research Policy 43 772–791
[6] Sommerville I 2011 Software Engineering 9th ed (Boston, MA, USA: Addison Wesley)
[7] Hu F, Mostashari A and Xie J (eds) 2011 Socio-Technical Networks: Science and Engineering Design (Boca
Raton, FL, USA: CRC Press, Taylor & Francis Group)
[8] Trist E L and Bamforth K W 1951 Human Relations 4 3–38
[9] Pasmore W, Winby S, Mohrman S A and Vanasse R 2019 Journal of Change Management 19 67–85
[10] Cherns A 1976 Human Relations 29 783–792
[11] Hirschhorn L, Noble P and Rankin T 2001 Journal of Engineering and Technology Management 18 241–252
ISSN 0923-4748
[12] Kopackova H and Libalova P 2017 2017 International Conference on Information and Digital Technologies
(IDT) (Zilina, Slovakia) pp 198–205
[13] Shin D H and Jin Park Y 2017 Digital Policy, Regulation and Governance 19 77–100
[14] Jotanovic G, Brtka V, Curguz Z, Stojcic M and Eremija M 2018 (Bitola, North Macedonia) pp 94–98
[15] Stojanov J, Jausevac G, Stojanov Z and Hristoski I 2019 VII International Symposium NEW HORIZONS
of Transport and Communications (Doboj, Bosnia and Hercegovina) pp 535–541
[16] Chu K H, Suthers D D and Rosen D 2012 2012 45th Hawaii International Conference on System Sciences
(Maui, HI, USA) pp 3560–3569
[17] Land S M and Zimmerman H T 2015 Educational Technology Research and Development 63 229–255
[18] Fraternali P, Cellina F, Herrera S, Krinidis S, Pasini C, Rizzoli A E, Rottondi C and Tzovaras D 2018
2018 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom
Workshops) (Athens, Greece) pp 59–64
[19] Baki S, Rozos E and Makropoulos C 2018 Science of The Total Environment 622-623 1590–1602
[20] Britto A L, Maiello A and Quintslr S 2019 Journal of Hydrology 573 1007–1020
[21] Balka E, Whitehouse S, Coates S T and Andrusiek D 2012 Information Systems Frontiers 14 1942
[22] Bardram J E and Frost M M 2018 Designing Healthcare That Works ed Ackerman M S, Goggins S P,
Herrmann T, Prilla M and Stary C (Academic Press) chap 10, pp 167–186 ISBN 978-0-12-812583-0
[23] Suthers D 2017 50th Hawaii International Conference on System Sciences (Hawaii, USA) pp 2128–2137
[24] Mavromoustakis C X and Karatza H D 2013 The Computer Journal 56 184–197
[25] Walton R and De Roure D 2019 Proceedings of the 10th ACM Conference on Web Science WebSci ’19
(Boston, MA, USA) pp 301–310 ISBN 9781450362023
[26] Zafar N A 2016 Complex Adaptive Systems Modeling 4 1–26
[27] Vodopivec N and Miller-Hooks E 2019 Reliability Engineering & System Safety 191 106561
[28] El Fazziki A, Benslimane D, Sadiq A, Ouarzazi J and Sadgal M 2017 IEEE Access 5 13192–13201
[29] Zhang Q, Jin Q, Chang J, Xiang S and Pan C 2018 2018 24th International Conference on Pattern Recognition
(ICPR) (Beijing, China) pp 1018–1023 ISSN 1051-4651
[30] Kalloniatis A, Ali I, Neville T, La P, Macleod I, Zuparic M and Kohn E 2017 Applied Ergonomics 61 178–196
[31] Masci P, Curzon P, Furniss D and Blandford A 2015 Innovations in Systems and Software Engineering 11
113130
[32] Cepulis D and Niu N 2018 2018 IEEE Symposium on Visual Languages and Human-Centric Computing
(VL/HCC) (Lisbon, Portugal) pp 17–21
[33] Gross J L and Yellen J (eds) 2003 Handbook of graph theory Discrete mathematics and its applications (Boca
Raton, FL, USA: CRC Press, Taylor & Francis Group)
[34] Barthélemy M, Barrat A, Pastor-Satorras R and Vespignani A 2005 Physica A: Statistical Mechanics and its
Applications 346 34–43
[35] Allard A, Ángeles Serrano M, Garcı́a-Pérez G and Boguná M 2017 Nature Communications 8
[36] Jiang L, Carley K M and Eberlein A 2012 2012 International Conference on Software and System Process
(ICSSP 2012) (Zurich, Switzerland) pp 160–169
[37] Goggins S P, Valetto G, Mascaro C and Blincoe K 2013 User Modeling and User-Adapted Interaction 23
345379
[38] Goggins S P, Mascaro C and Valetto G 2013 Journal of the American Society for Information Science and
Technology 64 516–539
[39] Ferrara E, Interdonato R and Tagarelli A 2014 Proceedings of the 25th ACM Conference on Hypertext and
Social Media HT 14 (Santiago, Chile) pp 24–34
[40] Islam M, Toma T R, Selim M, Gias A U and Khaled S M 2016 International Journal of Information
Engineering and Electronic Business(IJIEEB) 8 1–13
[41] Endsley M R 1995 Human Factors 37 32–64
[42] Stanton N A, Stewart R, Harris D, Houghton R J, Baber C, McMaster R, Salmon P, Hoyle G, Walker G,
Young M S, Linsell M, Dymott R and Green D 2006 Ergonomics 49 1288–1311
[43] Binzagr F and Medjahed B 2018 International Conference on Service-Oriented Computing (ICSOC 2018)
(Lecture Notes in Computer Science vol 11236) pp 679–693
[44] Pattanaik V, Singh M, Gupta P K and Singh S K 2016 2016 IEEE Region 10 Conference (TENCON)
(Singapore, Singapore) pp 3420–3423
[45] Ikram M, Beaume P and Kaafar M 2019 Proceedings of the 16th International Joint Conference on e-Business
and Telecommunications (ICETE 2019) vol 2 (SECRYPT) (Prague, Czech Republic) pp 211–219
[46] Liu Z, Lai Y C and Ye N 2003 Physical Review E 67(3) 031911
[47] Rouse W B and Serban N 2011 Information Knowledge Systems Management 10 25–49
[48] Baxter G and Sommerville I 2011 Interacting with Computers 23 4–17
[49] Creswell J W 2009 Research design: Qualitative, quantitative, and mixed methods approaches 3rd ed
(Thousand Oaks, CA, USA: SAGE Publications)