ArticlePDF Available

What Is an Emerging Technology?

Authors:

Abstract

There is considerable and growing interest in the emergence of novel technologies, especially from the policy-making perspective. Yet as an area of study, emerging technologies lacks key foundational elements, namely a consensus on what classifies a technology as ’emergent’ and strong research designs that operationalize central theoretical concepts. The present paper aims to fill this gap by developing a definition of ’emerging technologies’ and linking this conceptual effort with the development of a framework for the operationalisation of technological emergence. The definition is developed by combining a basic understanding of the term and in particular the concept of ’emergence’ with a review of key innovation studies dealing with definitional issues of technological emergence. The resulting definition identifies five attributes that feature in the emergence of novel technologies. These are: (i) radical novelty, (ii) relatively fast growth, (iii) coherence, (iv) prominent impact, and (v) uncertainty and ambiguity. The framework for operationalising emerging technologies is then elaborated on the basis of the proposed attributes. To do so, we identify and review major empirical approaches (mainly in, although not limited to, the scientometric domain) for the detection and study of emerging technologies (these include indicators and trend analysis, citation analysis, co-word analysis, overlay mapping, and combinations thereof) and elaborate on how these can be used to operationalise the different attributes of emergence.
What Is an Emerging Technology?
Daniele Rotolo1,2,Diana Hicks2, and Ben R. Martin1,3
1SPRU Science Policy Research Unit, University of Sussex, Brighton, United Kingdom
2School of Public Policy, Georgia Institute of Technology, Atlanta, United States
3Centre for Science and Policy (CSAP) and Centre for Business Research, Judge Business School,
University of Cambridge, Cambridge, United Kingdom
Version: July 7, 2015
Accepted for publication in Research Policy§
Abstract
There is considerable and growing interest in the emergence of novel technologies, especially
from the policy-making perspective. Yet as an area of study, emerging technologies lacks
key foundational elements, namely a consensus on what classifies a technology as ’emergent’
and strong research designs that operationalize central theoretical concepts. The present
paper aims to fill this gap by developing a definition of ’emerging technologies’ and linking
this conceptual effort with the development of a framework for the operationalisation of
technological emergence. The definition is developed by combining a basic understanding
of the term and in particular the concept of ’emergence’ with a review of key innovation
studies dealing with definitional issues of technological emergence. The resulting definition
identifies five attributes that feature in the emergence of novel technologies. These are: (i)
radical novelty, (ii) relatively fast growth, (iii) coherence, (iv) prominent impact, and (v) un-
certainty and ambiguity. The framework for operationalising emerging technologies is then
elaborated on the basis of the proposed attributes. To do so, we identify and review ma jor
empirical approaches (mainly in, although not limited to, the scientometric domain) for the
detection and study of emerging technologies (these include indicators and trend analysis,
citation analysis, co-word analysis, overlay mapping, and combinations thereof) and elabo-
rate on how these can be used to operationalise the different attributes of emergence.
Keywords: emerging technologies; conceptualisation; definition; attributes of emergence;
operationalisation; detection and analysis; framework; scientometrics; indicators; Science
and Technology Studies.
Corresponding author: d.rotolo@sussex.ac.uk, Phone: +44 1273 872980
diana.hicks@pubpolicy.gatech.edu
b.martin@sussex.ac.uk
§DOI: 10.1016/j.respol.2015.06.006.c
2015 Rotolo, Hicks, Martin. Distributed under CC-BY-NC-ND.
1
1 Introduction
Emerging technologies have been the subject of much debate in academic research and a central
topic in policy discussions and initiatives. Evidence of the increasing attention being paid to
the phenomenon of emerging technologies can be found in the growing number of publications
dealing with the topic and news articles mentioning emerging technologies (in their headlines or
lead paragraphs), as depicted in Figure 1. Increasing policy interest in emerging technologies,
however, must be set against a literature where no consensus has emerged as to what qualifies a
technology to be emergent. Definitions proposed by a number of studies overlap, but also point
to different characteristics. For example, certain definitions emphasise the potential impact
emerging technologies are capable of exerting on the economy and society (e.g. Porter et al.,
2002), especially when they are of a more ’generic’ nature (Martin, 1995), while others give
great importance to the uncertainty associated with the emergence process (e.g. Boon and
Moors, 2008) or to the characteristics of novelty and growth (e.g. Small et al., 2014). The
understanding of emerging technologies also depends on the analyst’s perspective. An analyst
may consider a technology emergent because of its novelty and expected socio-economic impact,
while others may see the same technology as a natural extension of an existing technology. Also,
emerging technologies are often grouped together under ’general labels’ (e.g. nanotechnology,
synthetic biology), when they might be better treated separately given their different socio-
technical features (e.g. technical difficulties, involved actors, applications, uncertainties).
The lack of consensus over definitions is matched by an ’eclectic’ and ad hoc approach to
measurement. A wide variety of methodological approaches have been developed, especially
by the scientometric community, for the detection and analysis of emergence in science and
technology domains (e.g. Boyack et al., 2014; Gl¨anzel and Thijs, 2012; Porter and Detampel,
1995). These methods, favoured, because they take advantage of growing computational power
and large new datasets and allow one to work with more sophisticated indicators and models, lack
strong connections to well thought out concepts that one is attempting to measure, a basic tenet
of good research design. Often no definition of the central concept of an emerging technology is
provided. It is no surprise therefore that approaches to the detection and analysis of emergence
tend to differ greatly even with the use of the same or similar methods. The operationalisation
of emergence is also in a state of flux. It changes as new categorisations (e.g. new terms in
institutionalised vocabularies, new technological classes) are created within databases. This,
2
in turn, makes less clear the exact nature of the phenomena that these scientometric methods
enable us to examine.
These problems in the effort to understand emerging technologies limit the utility of the
research and so may hamper resource allocation and the development of regulations, which, in
turn, have a major role in supporting and shaping the directionality of technological emergence.
Year
Number of publications
1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
0 50 100 150 200 250 300
Year
0 500 1000 1500 2000 2500 3000 3500
Number of news articles
News articles
Publications in all disciplines
Publications in social sciences
Figure 1: Publications (left axis) and news articles (right axis) including the variations of the term ”emerg-
ing technologies”. Publications were retrieved by querying SCOPUS data: ”TITLE(”emerg* technol*”)
OR TITLE(”emergence of* technolog*”) OR TITLE(”techn* emergence”) OR TITLE(”emerg* scien*
technol*”)”. Publications in social sciences were defined as those assigned to the SCOPUS categories
”Business, Management and Accounting”, ”Decision Sciences”, ”Economics, Econometrics and Finance”,
”Multidisciplinary”, ”Psychology”, and ”Social Sciences”. News articles were identified by searching for
emerg* near2 technolog* in article headlines and lead paragraphs as reported in FACTIVA. From 1980
to 2013, the average yearly growth rates of the number of publications concerning emerging technologies
in all disciplines and in social sciences have been of 12.5% and 23.8%, respectively. The total number of
publications in SCOPUS has yearly grown on average by 4.9%.
Source: search performed by authors on SCOPUS and FACTIVA.
The present paper addresses both the conceptual and methodological gaps. We aim to
elaborate a framework that links what is conceptualised as ’emerging technologies’ with its
measurement, thus providing guidance to future research (e.g. development of novel methods
for the detection of emergence and analysis of its characteristics) and to policy-making (e.g.
resource allocation, regulation). To do so, we first attempt to clarify the conceptualisation
of emerging technologies by integrating different conceptual contributions on the topic into a
more precise and coherent definition of ’emerging technology’. We begin with the definition of
’emergence’ or ’emergent’, which is the process of coming into being, or of becoming important
3
and prominent. This is then enriched and contextualised with a review of major contributions to
innovation studies that have focused on technological emergence, highlighting both their common
and contradictory features. Conceptual attempts to grapple with emergence in complex systems
theory are also discussed where relevant to the idea of emergent technology.
The result is the delineation of five key attributes that qualify a technology as emerging.
These are: (i) radical novelty, (ii) relatively fast growth, (iii) coherence, (iv) prominent impact,
and (v) uncertainty and ambiguity. Specifically, we conceive of an emerging technology as a rad-
ically novel and relatively fast growing technology characterised by a certain degree of coherence
persisting over time and with the potential to exert a considerable impact on the socio-economic
domain(s) which is observed in terms of the composition of actors, institutions and patterns of
interactions among those, along with the associated knowledge production processes. Its most
prominent impact, however, lies in the future and so in the emergence phase is still somewhat
uncertain and ambiguous.
Second, the framework for operationalising emerging technologies is developed on the basis of
the attributes we identified. The scientometric literature forms the core of the methods discussed
because, as mentioned, this field has been remarkably active in developing methodologies for
the detection and analysis of emergence in science and technology. The reviewed methods
are grouped into five main categories: (i) indicators and trend analysis, (ii) citation analysis
(including direct citation and co-citation analysis, and bibliographic coupling), (iii) co-word
analysis, (iv) overlay mapping, and (v) hybrid approaches that combine two or more of the
above. Because scientometric techniques cannot address all the attributes comprehensively, we
also discuss approaches developed in other fields.
The paper is organised as follows. The next section introduces the concept of emergence and
its various components. In Section 3, these elements are integrated with key innovation studies
proposing definitions of technological emergence, and a definition of emerging technologies is
then elaborated. Section 4 reviews methods to both detect and analyse emergence, and then
examines the use of those approaches to operationalise the proposed definition and the various
attributes of emerging technologies. Section 5 discusses the limits of current methodologies for
the detection and analysis of emerging technologies and identifies directions for future research.
Section 6 summarises the main conclusions of the study.
4
Table 1: Dictionary definitions of the concept of emergence.
Dictionary definition of ’emerge’/’emergent’ Attributes
”the process of coming into being, or of becoming important and promi-
nent” (New Oxford American Dictionary)
come into being; important; promi-
nent
”to become manifest: become known [...]” (Merriam-Webster’s Colle-
giate Dictionary)
become manifest; become known
”to rise up or come forth [...] to become evident [...] to come into
existence” (The American Heritage Desk Dictionary and Thesaurus)
evident; come into existence
”move out of something and become visible [...] come into existence or
greater prominence [...] become known [...] in the process of coming
into being or prominence” (Concise Oxford English Dictionary)
visible; prominent; become known;
come into being
”starting to exist or to become known [...] to appear by coming out
of something or out from behind something (Cambridge Dictionaries
Online)
become known; to appear
Source: search performed by authors on major English dictionaries.
2 The concept of emergence
The word ’emerge’ or ’emergent’ means ”the process of coming into being, or of becoming
important and prominent” (New Oxford American Dictionary) or ”to rise up or come forth
[...] to become evident [...] to come into existence” (the American Heritage Desk Dictionary
and Thesaurus). Table 1 presents dictionary definitions of emergent. The primary attribute of
emergence is ’becoming’ that is, coming into existence. Emergent is not a static property; it
is a label for a process. The endpoint of the process is variously described as visible, evident,
important or prominent. Thus, among the dictionaries there is some disagreement as to whether
acknowledged existence is enough for emergence, or beyond that, a certain level of prominence
is needed in order to merit application of the term emergence.
There is a second definition of emergent given the by the New Oxford American Dictionary
as: a property arising as an effect of complex causes and not analysable simply as the sum
of their effects. An additional definition is: arising and existing only as a phenomenon of
independent parts working together, and not predictable on the basis of their properties. This
concept of emergence is used in the study of complex systems. It can be traced back to the 19th
Century in the proto-emergentism movement when Lewes (1875) referred to ’emergent effects’
in chemical reactions as those effects that cannot be reduced to the components of the system,
i.e. the effects for which it is not possible to trace all the steps of the processes that produced
them. Its application in the study of the dynamics of complex systems in physics, mathematics,
and computer science gave rise to other fundamental theories and schools of thought such as
5
complex adaptive system theory, non-linear dynamical system theory, the synergetics school,
and far-from-equilibrium thermodynamics (see Goldstein, 1999).
A number of studies focusing on the definitional issue of emergence were produced by schol-
ars in complex system theory see Table A1 in the Appendix for an overview of the definitions
of emergence proposed by major studies in complex system theory. Goldstein (1999), for exam-
ple, defined emergence as ”the arising of novel and coherent structures, patterns, and properties
during the process of self-organization in complex systems” (1999, p. 49). An ontological and
epistemological definition of emergence is instead developed by de Haan (2006). Ontological
emergence is ”about the properties of wholes compared to those of their parts, about systems
having properties that their objects in isolation do not have” (2006, p. 294), while epistemo-
logical emergence it is about ”the interactions between the objects that cause the coming into
being of those properties, in short the mechanisms producing novelty” (2006, p. 294).
Though research on complex systems may have a certain cachet (and perhaps for this reason
scholars of emerging technologies sometimes attempt to work with the meaning of emergent
as conceived by the complex system approach), we maintain that questions about emerging
technologies are not fundamentally about understanding the origins and the causal nature of
full system interaction; rather they are about uncertainty, novelty, identification at an early
stage, and visibility and prominence. It is true that some technologies in themselves may be
complex systems in the sense of exhibiting adaptation, self-organisation, and emergence, an
example being parts of materials science (Ivanova et al., 1998). However, other technologies
exhibit ’complicatedness’ rather than ’complexity’ as defined in complex system theory for
example, engineering systems. These systems are designed for specific purposes, but they do
not adapt and self-organise to changes in the environment (Ottino, 2004). It is also true that
emerging technologies may arise from complex innovation systems (Katz, 2006), but we would
contend that in the phrase ’emerging technology’, ’emerging’ is generally understood in the
standard sense, not the complex system usage.
3 Defining emerging technologies
To further clarify what is meant by emerging technology, we reviewed literature in innovation
studies dealing with definitional issues of emerging technologies. To identify relevant studies, we
searched for ”emerg* technolog*”,”tech* emergence”,”emergence of* technolog*”, or ”emerg*
6
scien* technol*” in publication titles by querying SCOPUS (see the left-hand column of Ta-
ble 2).1We restricted the search to the title field to limit results to publications primarily
focused on emerging technologies. The search identified a total of 2,201 publications from 1971
to mid 2014.2Within this sample we selected those publications in social science domains, thus
reducing the sample to 501 records (see Figure 1).
We then read the abstracts and accessed the full-text of these studies where necessary both
to identify additional documents from the list of cited references and to exclude studies that are
not relevant to the scope of this paper. We found that about 50% of the studies in the sample
refer to a specific industrial context (e.g. listing and discussing emerging technologies in a given
industry) or to the educational sector (e.g. emergence of novel technologies to improve education
and learning). These were deemed not relevant to our study. The remaining studies were further
examined to identify those that develop or provide definitions of emerging technologies we
searched for ’defining’ sentences within the publication full-text by using the keywords listed
above. This led to a core set of 12 studies from science and technology (S&T) policy studies,
evolutionary economics, management, and scientometrics that contributed to the conceptualisa-
tion of technological emergence. These are listed with their definitions of emerging technologies
in Table 3. We analysed the textual content of the definitions reported in Table 3 to extract all
the component concepts. These were grouped into the attributes discussed below and used to
construct our definition of emerging technologies. Extracted concepts excluded from our list of
attributes will also be discussed.
The first defining attribute of emerging technology, explicitly included in two of the 12 core
articles, is radical novelty : ”novelty (or newness)” (Small et al., 2014) may take the form of
”discontinuous innovations derived from radical innovations” (Day and Schoemaker, 2000) and
may appear either in the method or the function of the technology. To achieve a new or a
changed purpose/function, emerging technologies build on different basic principles (Arthur,
2007) (e.g. cars with an internal combustion engine vs. an electric engine, cytology-based tech-
niques vs. molecular biology technologies). Novelty is not only a characteristic of technologies
deriving from technical revolutions, i.e. technologies with relatively limited prior developments
1The terminology of ’emerging technologies’ has become central to a number of research traditions and especially
to the scientometric, bibliometric and tech-mining domains (cf. Avila-Robinson and Miyazaki, 2011), which,
as discussed, have been remarkably active in developing methods for the operationalisation of emergence. In
other words, ’emerging technologies’ have become a category of its own. For this reason, we do not include
epistemologically related terms, such as ’radical’, ’disruptive’, ’discontinuous’, ’nascent’ and ’breakthrough’.
2The search was performed on 13th May 2014.
7
Table 2: Search strategies used to identify the set of relevant publications for the conceptualisation and
operationalisation of emerging technologies.
Conceptualisation Operationalisation
Search terms ”emerg* technolog*” ”emerg* technolog*”
”tech* emergence” ”tech* emergence”
”emergence of* technolog*” ”emergence of* technolog*”
”emerg* scien* technol*” ”emerg* scien* technol*”
”emerg* topic*”
”emergence of* topic*”
Field(s) of search Title Title, abstract, keywords
Focus Social sciences Scientometric journals: Journal of the As-
sociation for Information Science & Tech-
nology (formerly the Journal of the Amer-
ican Society for Information Science &
Technology), Journal of Informetrics,Re-
search Evaluation,Research Policy,Scien-
tometrics,Technological Forecasting & So-
cial Change,Technology Analysis & Strate-
gic Management
Number of studies 501 155
Source: authors’ elaboration as based on SCOPUS data.
(e.g. DNA sequencing technologies, molecular biology, nano-materials), but it may also be gen-
erated by putting an existing technology to a new use. The evolutionary theory of technological
change views this as the speciation process of technology, that is the process of applying an exist-
ing technology from one domain to another domain or ’niche’ (Adner and Levinthal, 2002). The
niche is characterised by a selection process that is different from the one where the technology
was initially applied. The niche specifically may differ in terms of adaptation (the needs of the
niche) and abundance of resources. The technology applied in the niche may adapt and then
emerge as well as potentially invading other domains including the initial domain (giving rise to
a ’revolution’ or a process of ’creative destruction’). This implies that ’evolutionary’ technology
(those not characterised by revolutionary technical developments) can also be radically novel in
domains of application different from those where the technology was initially developed. Adner
and Levinthal (2002) provided a compelling example of the speciation process by reporting on
the evolution of wireless communication technology. This technology was created for laboratory
purposes, and specifically for the measurement of electromagnetic waves. Yet, it found numerous
subsequent applications. Wireless communication technology first enabled communication with
locations (e.g. lighthouses) otherwise not reachable with wired telegraphy. Then, applications
expanded to the transmission of voice (radiotelephony and broadcasting), and, more recently, to
8
Table 3: Definitions of emerging technologies (studies are chronologically ordered).
Study Domain Definition (elaborated or adopted)
Martin (1995) S&T policy ”A ’generic emerging technology’ is defined [...] as a technology the
exploitation of which will yield benefits for a wide range of sectors of
the economy and/or society” (p. 165)
Day and
Schoemaker
(2000)
Management ”[...] emerging technologies as science-based innovation that have the
potential to create a new industry or transform an existing ones. They
include discontinuous innovations derived from radical innovations [...]
as well as more evolutionary technologies formed by the convergence of
previously separate research streams” (p. 30)
Porter et al.
(2002)
S&T policy ”Emerging technologies are defined [...] as those that could exert much
enhanced economic influence in the coming (roughly) 15-year horizon.”
(p. 189)
Corrocher
et al. (2003)
Evolutionary
economics
”The emergence of a new technology is conceptualised [...] as an evo-
lutionary process of technical, institutional and social change, which
occurs simultaneously at three levels: the level of individual firms or
research laboratories, the level of social and institutional context, and
the level of the nature and evolution of knowledge and the related tech-
nological regime.” (p. 4)
Hung and
Chu (2006)
S&T policy ”Emerging technologies are the core technologies, which have not yet
demonstrated potential for changing the basis of competition” (p. 104)
Boon and
Moors (2008)
S&T policy ”Emerging technologies are technologies in an early phase of develop-
ment. This implies that several aspects, such as the characteristics of
the technology and its context of use or the configuration of the actor
network and their related roles are still uncertain and non-specific” (p.
1915)
Srinivasan
(2008)
Management ”I conceptualize emerging technologies in terms of three broad sub-
heads: their sources [...], their characteristics [...] and their effects [...]
Specifically, I consider two aspects of the sources of emerging technolo-
gies the ’relay race evolution’ of emerging technologies, and ’revolu-
tion by application’ four characteristics of emerging technologies
the clockspeed nature of emerging technologies, convergence, dominant
designs, and network effects and three effects of emerging technolo-
gies shifting value chains, digitization of goods, and the shifting locus
of innovation (from within the firm to outside the firm).” (pp. 633-634)
Cozzens et al.
(2010)
S&T policy ”Emerging technology a technology that shows high potential but
hasn’t demonstrated its value or settled down into any kind of consen-
sus.” (p. 364) ”The concepts reflected in the definitions of emerging
technologies, however, can be summarised four-fold as follows: (1) fast
recent growth; (2) in the process of transition and/or change; (3) mar-
ket or economic potential that is not exploited fully yet; (4) increasingly
science-based.” (pp. 365-366)
Stahl (2011) S&T policy ”[...] emerging technologies are defined as those technologies that have
the potential to gain social relevance within the next 10 to 15 years.
This means that they are currently at an early stage of their develop-
ment process. At the same time, they have already moved beyond the
purely conceptual stage. [...] Despite this, these emerging technologies
are not yet clearly defined. Their exact forms, capabilities, constraints,
and uses are still in flux” (pp. 3-4)
Alexander
et al. (2012)
S&T policy ”Technical emergence is the phase during which a concept or construct
is adopted and iterated by [...] members of an expert community of
practice, resulting in a fundamental change in (or significant extension
of) human understanding or capability.” (p. 1289)
Halaweh
(2013)
Management Characteristics of (IT) emerging technologies ”are uncertainty, network
effect, unseen social and ethical concerns, cost, limitation to particular
countries, and a lack of investigation and research.” (p. 108)
Small et al.
(2014)
Scientometrics ”[...] there is nearly universal agreement on two properties associated
with emergence novelty (or newness) and growth.” (p. 2)
Source: search performed by authors on SCOPUS and extended to cited references.
9
data transmission (Wi-Fi). With each shift, wireless communication technology appeared radi-
cally novel in its new domain of application, although the technology itself had existed since the
early laboratory and telegraphy applications. The evolutionary theory of technological change
teaches us that radical novelty may characterise innovations based on both revolutionary and
evolutionary inventions resulting from the speciation process. However, the term ’evolutionary’
is also used to refer to incremental technological advances. To avoid ambiguity, we opted to
use the term ’radical novelty’ rather than ’revolutionary/evolutionary’ and to contextualise it
in relation to the domain(s) in which the technology is arising.3
The second defining attribute of emerging technologies, identified by three of the 12 core
articles is ”clockspeed nature” (Srinivasan, 2008) or ”fast growth” (Cozzens et al., 2010), or at
least ”growth” (Small et al., 2014). Growth may be observed across a number of dimensions
such as the number of actors involved (e.g. scientists, universities, firms, users), public and
private funding, knowledge outputs produced (e.g. publications, patents), prototypes, products
and services, etc. As with the radical novelty attribute, the fast growth of a technology needs
to be contextualised. A technology may grow rapidly in comparison with other technologies in
the same domain(s), therefore relatively fast growth may be a better term.
The third attribute of emerging technologies, identified by four of the 12 core articles is
coherence that persists over time. The core articles variously describe this attribute as ”conver-
gence of previously separated research streams” (Day and Schoemaker, 2000), ”convergence in
technologies” (Srinivasan, 2008), and technologies that ”have already moved beyond the purely
conceptual stage” (Stahl, 2011). Alexander et al. (2012) point instead to the role of ”an expert
community of practice”, which adopts and iterates the concepts or constructs underlying the
particular emerging technology. The concept of a community of practice suggests that both
a number of people and a professional connection between those people are necessary. Com-
ing together, intertwining and staying together are all entailed in coherence. Coherence refers
to internal characteristics of a group such as ’sticking together’, ’being united’, ’logical inter-
connection’ and ’congruity’. The status of external relations is also important. The emerging
technology must detach itself from its technological ’parents’ to some degree to merit a separate
identity. Furthermore, it must stay detached for some period of time to be seen as self-sustaining
3The word ’novelty’ alone may also create ambiguity with regard to the types of technologies we aim to include
in our conceptualisation of emerging technologies. Technologies of a more incremental nature, as derived from
the improvement of existing technologies, are somewhat novel. For the sake of conceptual clarity, we therefore
prefer to add the attribute ’radical’ to the word ’novelty’.
10
(Gl¨anzel and Thijs, 2012). As we stated above, emergence is a process and coherence, detach-
ment and identity do not characterise a final state, but are always in the process of realisation,
presenting challenging issues of boundary delineation and classification. Perspective matters
since an analyst may see an exciting emerging technology about to make a major economic
impact in something a scientist sees as long past the exciting emerging phase.
The fourth defining attribute of emerging technologies, identified by nine of the 12 core
articles is to yield ”benefits for a wide range of sectors” (Martin, 1995), ”create new indus-
try or transform existing ones” (Day and Schoemaker, 2000), ”exert much enhanced economic
influence” (Porter et al., 2002), or change ”the basis of competition” (Hung and Chu, 2006).
Corrocher et al. (2003) also point to the pervasiveness of the impact that the emerging technol-
ogy may exert by crosscutting multiple levels of the socio-economic system, i.e. organisations and
institutions, as well as knowledge production processes and technological regimes. Accordingly,
we identify prominent impact as another key attribute of emerging technologies. Most of the
core articles conceived the prominent impact of emerging technologies as exerted on the entire
socio-economic system. In this usage the concept of emerging technologies becomes very close to
that of ’general purpose technologies’ and so excludes technologies prominent within a specific
domain. We wish to include relatively smaller scale prominence in our definition. For example,
a diagnostic technology may emerge and significantly reshape the clinical practices associated
with a given disease, profoundly affecting one disease domain but not others. In other words, our
definition allows for prominent impact with narrow scope (emergence in one or a few domains),
as well as wide-ranging impact across domains and potentially the entire socio-economic system
(e.g. ICT and molecular biology). Such a perspective suggests, as with the attributes of radical
novelty and relatively fast growth, the importance of contextualising the prominent impact of
the observed technology within the domain(s) from which the technology emerges.
The final defining attribute of emerging technologies, identified in seven of the 12 core articles
is that the prominent impact of emerging technologies lies somewhere in the future the
technology is not finished. Thus, uncertainty features in the emergence process. The non-linear
and multi-factor nature of emergence provides emergence with a certain degree of autonomy,
which in turn makes predicting a difficult task (de Haan, 2006; Mitchel, 2007). As a consequence,
knowledge of the probabilities associated with each possible outcome (e.g. potential applications
of the technology, financial support for its development, standards, production costs) may be
particularly problematic (Stirling, 2007). Core articles expressed this attribute in terms of the
11
’potential’ that emerging technologies have for changing the existing ’ways of doing things’ (e.g.
Boon and Moors, 2008; Hung and Chu, 2006; Stahl, 2011).
However, these definitions seem not to disentangle explicitly another important aspect of
emergence from the concept of uncertainty. This is ambiguity. Ambiguity arises because pro-
posed applications are still malleable, fluid and in some cases contradictory, i.e. even the knowl-
edge of possible outcomes of emergence is incomplete. A variety of possible outcomes may occur
because social groups encountered during emergence hold diverging values and ascribe different
meanings to the technology (Mitchel, 2007). It is worth noting that uncertainty and ambiguity
are, however, not mutually exclusive (Stirling, 2007). These are not discrete conditions. A con-
tinuum exists as defined by the extent to which knowledge of possible outcomes and likelihood
for each outcome is incomplete. For example, it may be problematic evaluating the probabili-
ties associated with known possible outcomes, but at the same time there may also be a lack
of knowledge of other possible outcomes such as unintended/undesirable consequences deriving
from the (potentially uncontrolled) use of the technology. Uncertainty and ambiguity are key
starting concepts for a wide variety of science and technology studies (STS) focusing on the role
of the expectations in technological emergence (e.g. van Lente and Rip, 1998).
The studies reviewed here introduced various additional concepts such as the science-based-
ness, network effects, and early-stage development of emerging technologies. While the last of
these seems to be implicit in the definition of emergence and the key role of networks (of users
adopting the technology) is certainly not a unique feature of emerging technologies, the associ-
ation with science-based-ness is less clear. The importance of science (especially public science)
for the development of industrial technologies is widely accepted on the basis of substantial
evidence (e.g. Narin et al., 1997). However, even today not all technological revolutions may
depend on breakthrough advances in science. In certain domains, a technology can be developed
without the need for deep scientific understanding of how the phenomenon underlying it works
”it is possible to know how to produce an effect without knowing how an effect is produced”
(Nightingale, 2014, p. 4). For example, Vincenti (1984) provided evidence of this in the case
of the construction of airplanes in the 1930s. The different parts of an airplane were initially
joined using rivets with dome-shaped heads. These types of rivets, however, caused resistance
to the air, thus reducing the aerodynamic efficiency of the plane. As other dimensions of air-
plane performance were improving (e.g. speed), the aerodynamic efficiency became increasingly
relevant. The dome-shaped rivets were therefore replaced with rivets flush with the surface of
12
Table 4: Attributes of emergence and reviewed key innovation studies.
Innovation studies defining emerging technologies
Attribute of emergence
Martin (1995)
Day and Schoemaker (2000)
Porter et al. (2002)
Corrocher et al. (2003)
Hung and Chu (2006)
Boon and Moors (2008)
Srinivasan (2008)
Cozzens et al. (2010)
Stahl (2011)
Alexander et al. (2012)
Halaweh (2013)
Small et al. (2014)
Radical novelty x x
Relatively fast growth x x x
Coherence x x x x
Prominent impact x x x x x x x x x
Uncertainty and ambiguity x x x x x x x
Source: authors’ elaboration.
the airplane. This was a major improvement for the aerodynamics of airplanes in 1930s, but it
required no major scientific breakthrough.4A more recent example is the development of smart-
phones which did not require major advancements in science since most of the technologies used
already existed the integration of these technologies, and advances in design for the creation
of novel user interfaces instead provided the foundation of the innovation.5For these reasons,
’science-based-ness’ does not feature in our definition of emerging technologies.
In summary, as reported in Table 4, our review of innovation studies identified five main
defining characteristics or attributes of emerging technologies: (i) radical novelty, (ii) relatively
fast growth, (iii) coherence, (iv) prominent impact, and (v) uncertainty and ambiguity. Com-
bining these attributes, we define an emerging technology as a radically novel and relatively fast
growing technology characterised by a certain degree of coherence persisting over time and with
the potential to exert a considerable impact on the socio-economic domain(s) which is observed
in terms of the composition of actors, institutions and patterns of interactions among those,
along with the associated knowledge production processes. Its most prominent impact, however,
lies in the future and so in the emergence phase is still somewhat uncertain and ambiguous.
4Other classical examples include prehistoric cave dwellers using fire for cooking without any scientific under-
standing of it, the development of steam engines that predated the development of thermodynamics, or the
Wright brothers testing flying devices before the field of aerodynamics was established.
5The innovation was architectural rather than modular according to the distinction proposed by Henderson
and Clark (1990). Also, smartphone technology can be considered as an example of emerging technology of
an evolutionary nature. As discussed above, the radical novelty of this technology is the result of existing
technologies converging in new domains of applications.
13
It is reasonable to assume that the attributes of emergence range from ’low’ to ’high’ levels.
Nonetheless, to try and pin them down to some absolute level is rather meaningless. As discussed,
the attributes of emergence (especially radical novelty and relatively fast growth) provide an
indication of emergence when they are considered in the domain in which the given technology
is arising and therefore in relation to other technologies that may exist in that domain. Most
importantly, these attributes are likely to co-evolve and assume very different levels over different
periods of emergence. In the early stage of emergence (’pre-emergence’), a technology is likely
to be characterised by high levels of radical novelty as compared to other technologies in the
domain in which it is arising. However, the impact the technology can exert on that domain
is still relatively low. The technology has not yet gone beyond the purely conceptual stage,
multiple communities are involved in its development, and the delineation of the boundary of the
technology is particularly problematic (i.e. low levels of coherence). As a consequence, its growth
is relatively slow or not yet begun, and high levels of uncertainty and ambiguity are associated
with the future developments of the technology the technology may not even emerge. The
technology may then acquire a certain momentum. Some trajectories of development may
have been selected out and certain dimensions of performance prioritised and improved. A
community of practice may have also emerged. The technology thus becomes more coherent.
Its impact is also relatively less uncertain and ambiguous, and the technology starts to take
off in terms of publications, patents, researchers, firms, prototypes/products, etc. However,
at the same time, it is likely that the radical novelty of the technology will diminish other
technologies that exploit different basic principles may be emerging as well in the domain in
which the considered technology is emerging. We conceived ’emergence’ as this phase where the
attributes of emergence are subject to dramatic change. Finally, impact and growth may enter
a stable or declining phase, the technology loses its radical novelty, knowledge of the possible
outcomes of the technology becomes more complete (probabilities can be perhaps assigned to
outcomes), and the community of practice may become well-established (e.g. regular conferences,
dedicated journals). The technology enters in a ’post-emergence’ period. In line with the S-
shaped patterns highlighted in early studies on the growth of science (e.g De Solla Price, 1963)
and in technological adoption literature (e.g. Mansfield, 1961; Rogers, 1962), we ’stylised’ the
change in the levels of the attributes of emergence as following an S curve (or more strictly, a
reversed S curve in two of the five cases). This is qualitatively depicted in Figure 2.
Defining ’emerging technology’ is, however, only half the battle. If the definition is to be
14
Time
Attribute
Pre-emergence Emergence Post-emergence
Relatively fast growth
Coherence
Prominent impact
Radical novelty
Uncertainty and ambiguity
Figure 2: Pre-emergence, emergence, and post-emergence: attributes and ’stylised’ trends.
Source: authors’ elaboration.
useful, we must show how the attributes can be measured and thus how technologies can be
classified as emerging or not. In the next section, we link our definition to the the operational-
isation of our definition of emerging technologies. We rely mainly on scientometric techniques,
bringing in other approaches to fill certain gaps.
4 A framework for the operationalisation of emergence
Scientometric research has developed methods to detect emergence in science and technology
and is therefore central to operationalising our definition. From the vast literature that touches
on emerging technologies, we drew upon studies that offer ideas on operationalising our five
attributes. We identified relevant scientometric studies by including the term ’topic’ in the search
string we used to select research works dealing with definitional issues of emerging technologies
’topic’ is often used in scientometrics to refer to the emergence of a new set of research activities
in science and technology (e.g. Gl¨anzel and Thijs, 2012; Small et al., 2014). The search was also
extended to publication titles, abstracts and keywords, but narrowed to journals mainly or to
a significant extent oriented toward the publication of novel scientometric techniques (see the
right-hand column of Table 2). The search in SCOPUS returned 155 publications.
The examination of cited references of these publications enabled us to retrieve additional
studies that were not captured with the search string, but are potentially relevant to for our
15
analysis. This increased the initial sample to 183 studies. We then analysed these publications
to identify studies that were relevant to the operationalisation of the attributes of emergence.
This process led to a final set of 55 publications,6which were then classified in terms of the
methodological approach adopted to detect or analyse emergence (e.g. indicators, citation pat-
terns between documents, co-occurrence of words in text), data sources used (e.g. publications,
patents, news articles), and proposed operationalisation of emergence. This information is sum-
marised in Table 5 where studies are grouped into five groups: (i) indicators and trend analysis
studies that are mainly based on document counts; (ii) citation analysis studies which focus on
examining citation patterns between documents; (iii) co-word analysis studies that build on the
co-occurrence of words across document text; (iv) overlay mapping technique studies, which use
projections to position a given set of documents within a wider or more global structure (e.g.
a map of science); and (v) hybrid studies that combine two or more of the above approaches.
Table 5 shows how definitions of emergence varied, even within the same group of techniques,
thus providing further evidence of the low level of consensus on what constitutes emergence.
Given the definitional weaknesses in the original studies, our use of a particular study often
varies from that of its authors. We will briefly introduce the major techniques and our interpre-
tation of the contribution they make to measuring attributes of emerging technologies. For each
attribute, we will first describe how it can be operationalised for contemporary and then for
retrospective cases of emerging technologies. When data scarcity or the nature of the attribute
of emergence limit the applicability of scientometrics, we will discuss qualitative approaches.
The role of experts remains crucial for the validation of the results obtained with the use of
the techniques discussed below, especially for qualitative approaches to the operationalisation
6We excluded 76 studies that did not operationalise emergence (e.g. use of emerging technologies as empirical
context for various analyses, examination of ethical issues associated with emerging technologies), three studies
focused on the review of scientometric methods for the analysis of emerging technologies, and two studies elab-
orating document search strategies based on a modular lexical approach. 33 studies that were concerned with
Future-oriented Technology Analysis (FTA) techniques (e.g. foresight, forecasting, roadmapping, Constructive
Technology Assessment (CTA)) were also not included in the review. While about 67% of these do not rely
on scientometrics, the remaining FTA studies in the sample propose frameworks for selecting, rather than
identifying, emerging technologies, or adopt conventional scientometric/bibliometric approaches, which will be
instead discussed with the review of the selected scientometric studies. FTA methods, however, remain crucial
for more prospective analyses of emerging technologies and decision-making on possible future scenarios (e.g.
Ciarli et al., 2013; Irvine and Martin, 1984; Porter et al., 2004). 14 STS studies included in the sample will be
instead referenced in our review and discussion when the operationalisation of the attributes of emergence with
the use of scientometric approaches is limited by a lack of data or by the nature of the considered attribute. It
is worth noting that our search did not capture ’technometric’ studies (e.g. Grupp, 1994; Sahal, 1985; Saviotti
and Metcalfe, 1984). This research stream has been particularly important for the measurement of technology
and technological change. Nonetheless, technometric models tend to rely on a variety of assumptions and often
require data, the collection of which can be particularly labour-intensive (e.g. extraction and coding of data on
the features of the considered technologies) (e.g. Coccia, 2005).
16
Table 5: Methods for the detection and analysis of emergence in science and technology (studies are ordered by technique and publication year).
Method/Study Data Operationalisation of emergence
Indicators and trends
Porter and Detampel (1995) Publications/patents Count of keywords in publication abstracts and trend analysis based on Fisher-Pry curves
Kleinberg (2002) Publications/e-mails ’Burst of activity’ detected as state transitions of an infinite-state automaton
Bengisu (2003) Publications Positive slope of the line derived by regressing the number of publications on time and no decrease
of more than 10% or stability (no increase) in the last period or continuos decline in the last three
periods of observation
Watts and Porter (2003) Publications Indicators of emergence: cohesion (based on cosine similarity between documents), entropy, and
F-measure
Bettencourt et al. (2008) Publications Epidemic model to describe the increasing number of authors involved in an emerging field
Bettencourt et al. (2009) Publications Increasing densification (average number of edges per node), stable/decreasing diameter (average
path length between nodes), and increasing fractional count of edges in the largest component of the
co-authorship network
Moed (2010) Publications Journals characterised by high values of Source Normalised Impact per Paper (SNIP) indicator
Roche et al. (2010); Schiebel et al. (2010) Publications Publication keywords initially labelled as ”unusual terms”, by using tf-idf and Gini coefficient, that
subsequently become ”cross section terms”, i.e. they diffuse in several research domains
Guo et al. (2011) Publications Indicators of emergence: frequency of keywords (ISI keywords, authors’ keywords, and MeSH terms),
growing number of authors, and interdisciplinarity (based year-average Rao-Stirling diversity index)
of cited references
arvenp¨a et al. (2011) Mixed Absolute and cumulative count of the number of basic and applied research publications, patents,
and news
Abercrombie et al. (2012) Mixed Normalised number of publications and citations, patents, and web news fitted to a polynomial
function
Jun (2012); Jun et al. (2014) News Normalised searching traffic (Google trends)
Avila-Robinson and Miyazaki (2013a,b) Publications/patents Overview of indicators to analyse emergence
de Rassenfosse et al. (2013) Patents Count of the priority patent applications filed by a country’s inventor, regardless of the patent office
in which the application is filed
Ho et al. (2014) Publications Cumulative number of publications fitted to a logistic curve
17
Table 5: Methods for the detection and analysis of emergence in science and technology (studies are ordered by technique and publication year) (continued).
Method/Study Data Operationalisation of emergence
Citations analysis
Direct citation
Seminal paper: Garfield et al. (1964) Publications -
Kajikawa and Takeda (2008); Kajikawa
et al. (2008); Takeda and Ka jikawa (2008)
Publications Clusters of publications with the highest average publication year
Scharnhorst and Garfield (2010) Publications Historiographic approach combined with ’field mobility’ of publications
Shibata et al. (2011) Publications Clusters of publications with the highest values of betweenness centrality
Iwami et al. (2014) Publications Publications (’leading papers’) with high values of in-degree (’height’), large variation of in-degree
between one year and the next year (’slope’), or large cumulative in-degree (’area’) as defined on the
basis of the yearly direct citation network
Co-citation
Seminal paper: Small (1973) Publications -
Small (2006) Publications Clusters with no continuing publications from the prior period
Cho and Shih (2011) Patents Technological patent classes (IPC) that span structural holes in the co-citation network
´
Erdi et al. (2012) Patents Clusters of patents present in a given time period and not in the previous period
Boyack et al. (2014) Publications Yearly clustered publications of which references overlap less than 30% with references cited by
previous clusters
Bibliographic coupling
Seminal paper: Kessler (1963) Publications -
Morris et al. (2003) Publications Clusters of publications that cite more recent clusters of publications, namely emerging research
fronts
Kuusi and Meyer (2007) Patents Clusters of patents as source to identify guiding images (’leitbild’) of technological development
Co-word analysis
Seminal paper: Callon et al. (1983) Publications -
Lee (2008) Publications Clusters in the co-word network that show low values of degree, high betweenness, and low closeness,
i.e. those clusters that are more likely to turn into hub in the future.
Ohniwa et al. (2010) Publications MeSH terms (clustered with co-word analysis) that are included in the top-5% by incremental rate
in a given year the increment rate for a MeSH term is defined as the number of time the terms
occurred at the time t,t+ 1, and t+ 2 out the number of times the term occurred at t1, t,t+ 1,
and t+ 2
Yoon et al. (2011) Patents Small and dense sub-networks in the ’invention property-function’ network
Furukawa et al. (2015) Publications Sessions of conferences in which previous sessions converge according to the average cosine similarity
(based on tf-idf -identified keywords) between the papers included in the sessions
Zhang et al. (2014) Publications Combination of cluster analysis with term clumping and principal component analysis
18
Table 5: Methods for the detection and analysis of emergence in science and technology (studies are ordered by technique and publication year) (continued).
Method/Study Data Operationalisation of emergence
Overlay mapping
Rafols et al. (2010) Publications Overlays of publications projected on a basemap of ISI WoS subject categories linked by cosine
similarity of co-citations patterns between journals
Bornmann and Leydesdorff (2011) Publications Overlays of publications on Google maps to identify cities publishing more than expected
Leydesdorff and Rafols (2011) Publications Overlays of publications and co-authorship networks on Google maps to trace collaboration activity
Leydesdorff et al. (2012) Publications Overlays of publications projected on a basemap of MeSH terms linked by cosine similarity (based
on the co-occurrence of MeSH terms at the publication level)
Leydesdorff and Bornmann (2012) Patents Overlays of patents on Google maps to identify cities patenting more than expected
Leydesdorff et al. (2013) Publications Overlays of publications projected on the basemap of journals linked by cosine similarity of co-
citations patterns between journals
Kay et al. (2014) Patents Overlays of patents projected on the basemap of 466 IPC classes linked by cosine similarity of citing-
to-cited relationships between classes the basemap is built by using patents included in 2011
PATSTAT
Leydesdorff et al. (2014) Patents Overlays of patents projected on the basemap of 124 3-digit or 630 4-digit IPC classes linked by
cosine similarity based on co-citations between classes the basemap is built by using patents
granted at the United States Patent and Trademark Office (USPTO) from 1976 to 2011
Hybrid
Chen (2006): co-citation analysis and burst
detection Publications Trends in the bipartite network of research-front terms (burst detection) and intellectual base articles
the network includes three types of links: co-occurring research front terms, co-cited intellectual
base articles, and a research-front term citing an intellectual base article
Leydesdorff et al. (1994): co-citation analy-
sis and bibliographic coupling Publications New journals that build on multiple existing areas, i.e. they load on multiple factors obtained by
the factor analysis of the matrix of the cited references, and have unique ’being cited’ patterns,
i.e. they are ’central tendency journals’ reporting highest load on a given factor as obtained by the
factor-analysis of the matrix of received citations
Gl¨anzel and Thijs (2012): co-word, direct
citation analyses and bibliographic coupling Publications Existing clusters with exceptional growth, completely new clusters with roots in other clusters, and
existing clusters with a topic shift
Gustafsson et al. (2015): co-occurrence of
IPC classes Patents Technological co-classification to identify clusters of patents and detect guiding images or ’leitbild’
from patent full-text
Small et al. (2014): direct and co-citation
analyses Publications Clusters of publications that show high growth and are new both to the direct citation and co-citation
models
Yan (2014): co-word analysis and topic
modelling Publications Topics that are not a close variation of other topics, i.e. a topic iin the year tis emerging if no
predecessors are found and no other topics are transformed into topic iat t+ 1
Breitzman and Thomas (2015); Chang and
Breitzman (2009): direct citation and co-
citation analyses
Patents Clusters of patents (co-citation clustering) that form around ’hot’ patents defined as those patents
that are highly cited (top 5%-10%) by patents issued in the last two years and the citations of which
mostly come from patents issued in the last two years
Source: search performed by authors on SCOPUS and extended to publication cited references.
19
of emergence.
4.1 Radical novelty
Emerging technologies are radically novel, i.e. they fulfill a given function by using a different
basic principle as compared to what was used before to achieve a similar purpose. Publications
and patents are of limited use in assessing radical novelty in contemporary technology. In
contrast, news articles, editorials, review and perspective articles in professional as well as
academic journals represent valuable sources, providing participant perspectives on if and why
a technology is viewed as radically novel. These documents may also provide an understanding
of the basic principles underpinning the examined technology.
In contrast, in retrospective analyses citation and co-word analyses can be particularly ef-
fective for identifying radical novelty. Relatively large amounts of data can be exploited to map
the cognitive networks of a knowledge domain over time. Citation analysis builds on citation
patterns among documents to generate a network in which nodes are documents and links be-
tween nodes represent (i) a direct citation between two documents (direct citation analysis)
(Garfield et al., 1964), (ii) the extent to which two documents are cited by the same documents
(co-citation analysis) (Small, 1973), or (iii) to what extent two documents cite the same set of
documents (bibliographic coupling) (Kessler, 1963). Co-word analysis instead exploits the text
of documents to create a network of keywords (or key phrases) that are linked according to the
text to which they co-occur across the set of selected documents (Callon et al., 1983).
On the premise that clusters of documents or words in these networks represent different
knowledge areas of a domain or different literatures on which the domain builds, several studies
have considered the appearance of clusters not previously present in the network as a signal of
novelty (e.g. ´
Erdi et al., 2012; Kajikawa and Takeda, 2008). Others dispute this interpretation.
Given the continuous evolution of science and technology, one is unlikely to find a cluster again in
subsequent annual networks so the percentage of clusters that would qualify as newly appearing
tends to be relatively high. For this reason, additional criteria have been suggested such as
the appearance of new clusters that also link otherwise weakly connected (e.g. betweenness
centrality) clusters (e.g. Furukawa et al., 2015; Shibata et al., 2011), that form around documents
that are highly cited by recent documents and the citations of which also are mostly from recent
documents (Breitzman and Thomas, 2015), or that cite more recent clusters as identified by the
(Salton) similarity of their references (Morris et al., 2003).
20
Small et al. (2014) have recently proposed a hybrid approach based on a combination of direct
citation and co-citation models as applied to publication data. This approach is particularly
focused on the detection of novelty, which is defined in terms of clusters that are new to the
co-citation model that is, clusters with limited overlap with the cited documents included in
clusters in previous years (Boyack et al., 2014) as well as to a parallel direct citation model.
By combining bibliographic coupling, co-word analysis, and direct citation analysis, Gl¨anzel
and Thijs (2012) instead defined novelty (namely emerging topics) as three cases of clusters:
those that show exceptional growth, those that are completely new but with their roots in
other clusters, or already existing ones that exhibit a topic shift. Yan (2014) combined co-word
analysis with Natural Language Process (NLP) approaches (topic modelling). Emergence, as
reflected in novelty, is then associated with the appearance of topics that are not a close variation
of other topics calculated on the basis of the Jenson-Shannon Divergence.7Specifically, a topic
iappearing at time tis considered to be emerging if it has no predecessors and none of the
identified topics transforms into topic iat t+1. A different perspective is provided by Scharnhorst
and Garfield (2010) who extended the analysis of historiographs (based on direct citations) to
trace the extent to which publications move across fields as they receive citations from new
fields (namely ’field mobility’). Assuming that these publications are associated with a basic
principle used for technological applications, this approach enables one to identify which fields
may be using a different knowledge base and thus in which fields radically novel technologies
are potentially emerging. However, this requires a priori knowledge of the basic principle and
the set of documents associated with it.
Research in scientometrics has also focused on the development of techniques to expand the
’local’ (domain) perspective that citation or text-based approaches may provide. This effort has
generated a number of overlay mapping techniques (for an overview see Rotolo et al., 2014),
which in turn may be particularly well suited to detecting radical novelty. The basic idea is
to project a given set of documents (e.g. publications associated with a research domain) on a
basemap through the use of an overlay. The basemap can represent the ’global’ science struc-
ture at the level of the scientific discipline (ISI Web of Science (WoS) subject categories) (e.g.
Rafols et al., 2010), journal (e.g. Leydesdorff et al., 2013), Medical Subject Headings (MeSH)
(Leydesdorff et al., 2012), or the technological structure at the level of patent classes (e.g. Kay
7The Jenson-Shannon Divergence is a measure of similarity between empirically-determined distributions (e.g.
co-occurrence of words in documents) based on Shannon entropy measures (for more details see Lin, 1991).
21
et al., 2014; Leydesdorff et al., 2014).8Once the set of documents (publications or patents) asso-
ciated with a given domain has been identified, the projection of these documents over different
time slices on the global map of science or technology may reveal the increasing involvement
of new scientific or technological areas. This may suggest that new knowledge areas are being
accessed to conduct research, and thus that potentially different basic principles are drawn upon
to achieve a given purpose.
Among the studies within the ’indicators and trends’ group of techniques, Moed (2010)
proposed the source normalised impact per paper (SNIP) indicator for the evaluation of journals’
impact and claims it is relevant for identifying emerging technologies. This indicator is defined as
the ratio between the journal’s raw impact per paper (number of citations in the year of analysis
to the journal’s papers published in the three previous years, divided by the number of the
journal’s papers in these three years) and the relative database citation potential in the subject
field covered by the journal (mean number of 1-3-year-old references per paper citing the journal
and published in journals included in the considered database divided by that for the median
journal in the database). Moed (2010) argued that the SNIP indicator, and specifically high
values of this indicator, also provides information on the extent to which a considered journal
covers emerging topics. Given the focus on recent citations and database coverage, the SNIP
indicator is clearly associated with the radical novelty attribute of emergence. This indicator is,
however, evaluated at the aggregate level of the journal and journal-by-journal. It is therefore
less clear whether signals of radical novelty (i.e. relatively high values of SNIP) are associated
with one or multiple emerging topics the considered journal may cover. In addition, the SNIP
may not capture signals of radical novelty in those instances of journals that cover few emerging
topics and therefore characterised by low values of SNIP.
All these techniques have various advantages and limitations. The qualitative analysis of
news articles, editorials, review and perspective articles, for example, may be effective for con-
temporary analyses. However, the technical language used in these documents may be an im-
portant barrier to a non-expert’s efforts to independently assess radical novelty. The application
of citation and co-word analyses is strongly dependent on time. Data need to be longitudinal
in order to permit the tracing of cognitive dynamics and associated changes in the knowledge
8The elements of the basemap are linked according to similarity based on the co-occurrence of citations or, in the
case of MeSH, the co-occurrence of terms. The same approach can be used to project a sample of publications
and patents onto geographical maps (e.g. Google maps) to reveal the most active cities and collaborative
activities (seeTable 5).
22
structure. Co-word analysis and bibliographic coupling are, however, less sensitive to time than
direct citation and co-citation analyses and can be applied as documents become available (e.g.
Breitzman and Thomas, 2015). Finally, overlay mapping provides a global perspective on emer-
gence for the assessment of radical novelty, but interpretation of the resulting maps is mainly
based on visual inspection.
4.2 Relatively fast growth
Emerging technologies show relatively fast growth rates compared to non-emerging technologies.
The assessment of this attribute is particularly problematic for contemporary analyses. Growth
is not yet observed in terms of publications and patents, for example, so scientometric indicators
cannot be used. Early indications of growth may be revealed from the analysis of funding data,
big data, and altmetrics. This is an important research direction for future studies on the
operationalisation of the relatively fast growth attribute, as we will discuss later in the paper.
In the case of retrospective analyses, ’relatively fast growth’ is perhaps the most frequently
measured attribute of emergence in scientometrics. Most studies assume rapid growth as a sine
qua non condition of emergence, and so a number of operationalisation approaches have been
proposed. Indicators and trend analyses based on the yearly or cumulative count of documents
publications, patents or news articles, according to the nature of the examined technology
and the availability of data over a given observation period are widely used. Documents
are generally identified over time by using expert-defined keywords appearing in the publication
titles and abstracts (e.g. Porter and Detampel, 1995) or by exploiting more institutionalised
vocabularies such as the MeSH classification in the case of publication counts in the biomedical
domain (e.g. Guo et al., 2011). With a focus on patent data, de Rassenfosse et al. (2013)
proposed counting the priority patent applications filed by a country’s inventor, regardless of
the patent office in which the application is filed, as an indicator to identify fast growth and
therefore potential emerging technologies. However, yearly publication or patent counts are
always dynamic, so the problem becomes one of setting a criterion by which to distinguish the
signal from the noise, that is differentiating emerging technology from other increasing trends.
Some theoretical foundations are needed to do this.
Rapid growth is also detected by fitting the document count to a function (e.g. forms of
23
logistic function such as Fisher-Pry curves).9Bengisu (2003), for example, regressed the number
of publications over publication year and defined emerging technologies as those technologies
showing a positive slope and a decrease of less than 10% or stability (no increase) in the last
period compared to the previous one, or no continuous decline in the last three periods of
observation. Ho et al. (2014) instead fitted the cumulative number of publications to a logistic
curve, whereas Abercrombie et al. (2012) extended the count of publications to patents, web
news, and commercial applications. Data were then normalised and fitted to a polynomial
function for comparison a similar approach is employed by arvenp¨a et al. (2011) and Jun
et al. (2014).
The number of documents is also used to detect ’bursts of activity’, i.e. the appearance of
a topic in a document stream. This relies on the approach of Kleinberg (2002), who modelled
the number of publications and e-mails containing a given set of keywords as an infinite-state
automaton, i.e. a self-operating virtual machine that may assume a non-finite number of states
and where the transition from one state to another is regulated by a ’transition function’ (sim-
ilarly to Markov models). The frequency of state transitions with certain features identifies
bursts of activity, which are used as a proxy for fast growth. The burst detection approach is
combined with co-citation analysis by Chen (2006) to build a bipartite network10 of research-
fronts linked with intellectual base articles. This network is then analysed in order to identify
emerging trends.
Schiebel et al. (2010) and Roche et al. (2010) proposed instead an approach to emergence
that is based on a diffusion model (and diachronic cluster analysis to identify topics) that com-
bines a modified tf-idf 11 with the Gini coefficient to identify three stages: ”unusual terms”,
”established terms”, and ”cross section terms”. Unusual terms are those that are rare in publi-
cations since they describe a research discovery at a very early stage. When research intensifies,
terms first become more established in the original domain and subsequently they may diffuse
9Fisher-Pry curves were developed to model technological substitution between two competing technologies
(Fisher and Pry, 1971). This family of curves is built on the basis of three assumptions: (i) technological
advancements are the results of competitive substitutions of one method (technology) used to satisfy a given
need for another; (ii) the new technology completely replaces the old technology; and (iii) the market share
follows Pearl’s Law, i.e. ”the fractional rate of fractional substitution of new for old is proportional to the
remaining amount of the old left to be substituted” (Fisher and Pry, 1971, p. 75).
10 A bipartite network is a network in which nodes can be partitioned into two distinct groups, N1and N2, and
all the links connect one node from N1with a node from N2, or vice versa (Wassermann and Faust, 1994).
11 The tf-idf (term frequency-inverse document frequency) is an indicator that reflects the importance of a word
to a document in relation to a corpus. Specifically, the tf-idf is the result of the product between two indicators:
the term frequency and inverse document frequency.
24
into other domains, thus becoming cross section terms. Terms that change their classification
(i.e. that show pathways) from unusual to cross section terms from one period to another are
characterised by rapid diffusion and therefore relatively fast growth. This approach, however, is
highly dependent on the thresholds of the tf-idf and Gini coefficient selected to classify terms
as well as on the duration of the periods used to trace changes in the classification of terms.
Citation and co-word analyses can also be used to identify the relatively rapid growth of a
potential emerging technology. Longitudinal analysis of the size of the clusters of documents or
words obtained with the application of these techniques can detect knowledge areas that show
rapid growth. For example, Ohniwa et al. (2010) used co-word analysis to cluster MeSH terms.
For each MeSH term an increment rate was calculated in year tas the number of times the term
occurred at time t+ 1 and t+ 2 out of the number of times the term occurred at t1, t,t+ 1,
and t+ 2. Fast growing topics are those in the top 5% of the increment rate in a given year.
Gl¨anzel and Thijs (2012) combined bibliographic coupling, co-word analysis, and a direct
citation model. First, documents are clustered in time slices according to their cosine similarity
resulting from bibliographic coupling and textual similarity. The core clusters identified through
this process are next linked across different time slices via direct citations. Emergence is then
detected by identifying clusters with exceptional growth the study also considers emerging
clusters to be those that are completely new with roots in other clusters or existing clusters
exhibiting a topic shift, but this clearly refers to the radical novelty attribute of emergence.
Overlay mapping techniques can visually reveal knowledge areas characterised by a rapid
increase in the number of documents (publications or patents) in the ’global’ maps of science
or technology and which therefore, in comparison with other areas, may be growing at a faster
pace. (Overlay mapping can also reveal diffusion across disciplines and technological areas.)
Other studies instead operationalised relatively fast growth by examining the growing number
of authors involved in an emerging field over time (e.g. Bettencourt et al., 2008; Guo et al., 2011).
For example, (Bettencourt et al., 2008) found that the growth of the population of authors in a
given field tends to be relatively well described with epidemic models that consider novel ideas
as spreading by ’infecting’ authors.
4.3 Coherence
Coherence and its persistence over time distinguish technologies that have acquired a certain
identity and momentum from those still in a state of flux and therefore not yet emerging.
25
Coherence in contemporary technologies may be detected by examining the scientific discourse
around a given emerging technology. Initially, a variety of terms may be in use and reduction
in the number of terms may signal increasing coherence. Abbreviations or acronyms take time
to appear and, when they do, signal persistence; they also indicate shared interpretations and
thus coherence (Reardon, 2014). Additional signals of coherence may come from the creation of
conference sessions, tracks, dedicated conferences and subsequently from journal special issues
and new specialist journals (Leydesdorff et al., 1994). New categories in established classification
systems may also be created (Cozzens et al., 2010).
In retrospective analyses, entropy measures can be used (Watts and Porter, 2003) as well
as clustering and factor analysis of citation and text networks. The coherence of clusters of
documents or terms can be assessed in comparison to the overall network by applying, for
example, local network density measures as well as by examining cluster persistence over time.
Furukawa et al. (2015) propose using year-to-year coherence of conference sessions to indicate
emergence. They applied co-word analysis to generate ’chronological’ networks of conference
sessions (nodes) linked by their (cosine) similarity as based on the keywords included in the
sessions’ papers keywords were selected using the tf-idf indicator. Within these networks,
emerging topics are defined as sessions where previous conferences’ sessions converge according
to similarity.
In a similar vein, Yoon et al. (2011) developed a NLP algorithm capable of identifying prop-
erties and functions in the sentences of patent abstracts.12 The method generates an ’invention
property-function’ network (IPFN). Nodes in this network represent properties and functions.
A property is what a system is or has and is expressed by using ’adjectives+nouns’, whereas a
function is what a system does and is expressed by using ’verbs+nouns’. Links between nodes
are defined by the co-occurrence of properties and functions in patents. Emerging properties
and functions are those clustered in small and highly dense sub-networks i.e. de facto showing
a certain degree of coherence.
The approaches discussed above examine cognitive dynamics. However, coherence can also be
assessed on the basis of changes in the social structure. In this regard, Bettencourt et al. (2009)
examined the evolution of co-authorship networks at the level of the scientist to identify network
patterns associated with the emergence of new scientific fields. Increasing average number of
12 This enables one to overcome the main limitation of co-word analysis techniques, that is the need to define an
initial set of keywords before the analysis can be performed.
26
edges per nodes (densification), stable or decreasing average path length between two nodes
(diameter), and increasing fractional count of edges in the largest component of the considered
network were suggested as signals of emergence and specifically of the topical transition of a
field. These indicators clearly refer to increasing connectedness of the co-authorship network,
identifying emerging communities as an indicator of emerging technology.
4.4 Prominent impact
Emerging technologies exert a prominent impact on specific domains or more broadly on the
socio-economic system by changing the composition of actors, institutions, patterns of interac-
tions among those, and the associated knowledge production processes. Scientometric methods
cannot identify contemporary prominent impact due to a lack of data and the difficulty in delin-
eating the technology in its very early stages (e.g. keywords may still be used by groups of actors
with different meanings and in different contexts). Mixed qualitative-quantitative approaches
used by Science and Technology Studies (STS) scholars on the role of expectations in driving
technological change are of a particular relevance.13 The main argument of the sociology of
expectations is that ”novel technologies and fundamental changes in scientific principle do not
substantively pre-exist themselves, except and only in terms of the imaginings, expectations
and visions that have shaped their potential” (Borup et al., 2006, p. 285). These expectations
are ”real-time representations of future technological situations and capabilities [...] wishful
enactments of a desired future” (Borup et al., 2006, p. 285) and they play a generative role
by stimulating and steering as well as coordinating actions. Evidence of this has been found
in a number of emerging fields such as gene therapy, pharmacogenomics, and nanotechnology
(e.g. Hedgecoe and Martin, 2003; Martin, 1999; Selin, 2007). Expectations of the performance
of novel technologies or, more generally, the ability of novel technologies to address societal
problems are both important.
News articles, editorials, review and perspective articles in professional and academic jour-
nals, vision reports and technological roadmaps have all been used to identify statements rep-
resenting multiple and potentially competing expectations surrounding a technology (e.g. Alke-
13 Scientometrics can be considered as the more quantitative end of STS work. For this reason, the distinction
we make between the two traditions is not intended to be a particularly strong one. However, it also true that
there has been relatively little interaction between scientometrics and STS since the late1980s. Each of these
tradition has its own conferences and journals, and only a handful of researchers operate at the interface
most individuals would identify themselves as either ’scientometricians’ or ’STS’ scholars.
27
made and Suurs, 2012; Bakker et al., 2011; van Lente and Bakker, 2010). STS work has also
illuminated the central role played by hype in technology emergence. Actors who understand
the constitutive role of expectations have an incentive to raise expectations in order to moti-
vate the funding and activity needed to realise their preferred technological future. Hype, or
over-claimed expectation, is often the result. This over-claiming can touch most attributes of
emergence and especially prominent impact. For example, press releases prior to the launch
of the Segway claimed it would ’change walking’. Similarly, in the case of coherence, for the
government to fund nanotechnology research, they must ’believe’ nanotechnology is a ’thing’,
as opposed to a name applied by some to a rather miscellaneous selection of materials science
research activities. Therefore, proponents have an incentive to claim coherence where others
might disagree.
These studies have been retrospective, but their data sources are contemporaneous with
technology emergence so the method could be extended to contemporary analyses. Moreover,
mapping of expectations can be combined with scientometrics when suitable data become avail-
able. Gustafsson et al. (2015), for example, used technological co-classification to identify clus-
ters of patents, the full-text of which is subsequently analysed qualitatively to detect guiding
images or leitbild, which are generalisations shared by several actors which guide actors towards
similar objectives. Guiding images are used to explain the dynamics of expectations.
Retrospective analyses can rely more extensively on scientometrics, although this has not
been done very often. Scientometricians have mostly focused on the detection and analysis of
growth and novelty, whereas impact seems to be taken for granted. Nonetheless, scientometrics
can greatly contribute to evaluating the impact of a potentially emerging technology. A number
of techniques can be used to produce intelligence on the emergence process. These include
the analysis of highly-cited documents, of authorship data to generate intelligence about the
actors drawn into knowledge creation processes over time (e.g. private vs. public organisations
and incumbents vs. newcomers), and of changes in the collaboration structure as mapped with
co-authorship data (e.g. Hicks et al., 1986; Melin and Persson, 1996). Impact on knowledge
production processes can instead be assessed by examining the dynamics of cognitive networks
obtained from the study of the citations or the co-occurrence of terms across a particular set of
documents.
28
4.5 Uncertainty and ambiguity
Emerging technologies are characterised by uncertainty in their possible outcomes and uses,
which may be unintended and undesirable, as well as by ambiguity in the meanings different
social groups associate with the given technology (Mitchel, 2007; Stirling, 2007). For analyses
of contemporary emerging technologies, news articles, editorials, review and perspective articles
on professional and academic journals can be examined to qualitatively assess the degree of
uncertainty and ambiguity associated with an emerging technology as well as to identify possible
multiple visions of the future associated with the technology. As for the evaluation of how
prominent the impact of an emerging technology will be, an STS approach to the mapping of
expectations can be used for the assessment of uncertainty and ambiguity.
For retrospective analyses, the evaluation of uncertainty and ambiguity remains largely un-
explored in scientometric studies, however. The few attempts made along these lines tend to
overlap with those already discussed for the evaluation of the coherence attribute, since the main
focus has been on the measurement of the reduction of uncertainty in scientific communication
rather than on uncertainty and ambiguity associated with the potential impact or uses of emerg-
ing technologies. For example, the creation of a novel category (such as a new subject category
in the classification of ISI WoS), in which subsequent journals associated with the emerging
technology under examination may fall, is conceived as an indicator of increasing redundancy
in the communication process as new journals are established and achieve a critical mass to
justify the creation of a new category, the redundancy of the communication process associated
with the considered emerging technology has also increased. Increasing redundancy, in turn,
may indicate diminishing uncertainty.14 In a similar vein, Lucio-Arias and Leydesdorff (2009)
considered words in publication titles (which are selected by authors to position knowledge
claims at a given time), cited references (which enable authors to position knowledge claims in
the existing socio-cognitive domain), and time as key dimensions describing the scientific dis-
course at the research front of a specialty. The mutual information exchanged between these
dimensions (measured in terms of Shannon entropy) is suggested as an indicator of uncertainty
reduction. The gap in the assessment of uncertainty and ambiguity represents, however, an
important arena for future research, as we will discuss in the next section.
14 Personal communication with Loet Leydesdorff on 2 October 2014.
29
5 Discussion
We characterised emerging technologies on the basis of five attributes (i) radical novelty, (ii)
relatively fast growth, (iii) coherence, (iv) prominent impact, and (v) uncertainty and ambiguity
and used these to develop a framework for a coherent and systematic operationalisation of
emerging technologies. A wide variety of scientometric methods are available to operationalise
the various attributes of emergence. Nonetheless, these are strongly dependent on time, on the
nature of the attribute, and on the data used.
Scientometric techniques are intrinsically more effective for retrospective analyses than con-
temporary examinations. Time is required before documents such as publications and patents
can be observed and techniques can be applied longitudinally. For example, measuring growth
is particularly problematic for more contemporary analyses. Techniques using future citations
are more sensitive to this issue than methods that rely on data available when documents are
published (e.g. co-word analysis and bibliographic coupling). Lags in database indexing may
also contribute to the time limitations of scientometric approaches.
Scientometrics is also of little use in the operationalisation of uncertainty and ambiguity.
The focus of scientometrics has been mainly on the detection of what is emerging, rather than
on characterising the potential of what is detected to be emerging. To our knowledge, this area
is largely unexplored. Likewise, the methods reviewed in this paper show no explicit focus on
how the societal aspect of prominent impact can be assessed. This is somewhat surprising when
one considers the extensive scientometric work carried out for research evaluation purposes.
Furthermore, most studies have focused on publications and patents data that are not
only sensitive to time, but also provide limited perspectives on the multifaceted phenomenon
of emerging technologies. A few studies have focused on the use of news articles and big data
sources (e.g. Google Trends). These are clearly emerging streams in scientometric and data-
mining research, but so far little attention has been paid to the use of these novel data sources
in the context of emerging technologies.
The risk that detected technological emergence may be merely an artefact of the method
used adds to these limitations. The reviewed methodologies rely on different models, data,
thresholds, clustering algorithms and parameters, the selection of which may bias the detection
of emergence towards certain patterns. For example, technological emergence is often detected
with comparatively static analyses rather than with dynamic examinations. Data for a given
30
observation period are divided into time windows and algorithms are then applied to the sample
of data included in each time window. Results may vary with the number and length of time
windows. Shorter time windows may not identify certain patterns of emergence because they
do not capture a critical mass of documents, while longer time windows may miss cases of
technologies that exhibit emerging features for a shorter period (e.g. promising technologies that
eventually do not emerge). Also, the identified emerging technologies may be biased towards
certain topics. Small et al. (2014), for example, found that topics identified as emerging by
the combined ’direct citation-co-citation’ approach are in areas that are more likely to offer
practical outcomes than non-emerging topics. This may suggest that such areas attract more
resources, which, in turn, may favour the recruitment of researchers (Small et al., 2014). Yet,
the identification of these emerging areas may also be the result of the model and data used.
The field could move forward more confidently if instead of every study using a different data
set, a standard model dataset was developed to which all techniques could be applied and the
results compared (Katz, 1996).
We have argued that qualitative STS approaches can be particularly powerful for overcoming
the limits of scientometrics, for instance, in relation to prominent impact and to uncertainty and
ambiguity. For example, mapping expectations through content analysis of news, review articles,
and policy documents can provide important insights. Because STS focuses on human agency,
the importance of expectations and visions in steering emergence as well as the examination of
niche-regime dynamics is more apparent. Hence, this tradition attempts to address questions
of how emergence happens. This may favour meaningful interpretations of scientometric data
and possibly a better conceptual understanding. Scientometrics, in turn, can bring a more
robust empirical approach to the STS research tradition, including the capability to address
measurement error by means of statistical inference as well as to increase the generalisability of
results. Few studies have followed a combined scientometrics-STS approach. Kuusi and Meyer
(2007), for example, applied a bibliographic coupling approach to identify clusters of patents
and then to map ’guiding images’ used by different actors to develop a consensus around the
goals and directions during different phases of development of an emerging field.15 Yet, there
remains great potential for substantial links and a deeper synthesis between the two traditions
focusing on the examination of emergence in science and technology.
15 As noted earlier in Section 4.4, a similar mixed approach has been adopted by Gustafsson et al. (2015).
31
The conceptualisation and operationalisation of emerging technologies offer a number of op-
portunities for future research. From a conceptual point of view, more understanding of the
origins of emerging technologies is required. In the early phase of emergence (high levels of
radical novelty and of uncertainty and ambiguity, low levels of growth, coherence, and impact)
some technologies acquire a certain momentum to become ’emerging’ (when the levels of at-
tributes are subject to more dramatic change), other technologies instead arrive at the verge of
becoming emergent, but eventually not emerge at all. Funding and research programmes, the
power distribution among actors, communities of practices, and regulations are likely to exert a
significant impact on this process. However, more systematic research is required on the factors
that enable a technology to eventually become emergent. This also extends to the empirical
investigation of emergence. Studies often tend to analyse emerging technologies, without com-
paring them with a counter-factual sample of technologies that had the potential of becoming
emergent, but eventually did not emerge. Likewise, we have limited knowledge of the end point
of the emergence process, i.e. when emergence is over, or perhaps prematurely grinds to a halt
or reverses.
The limitations of the use of scientometrics for the operationalisation of some the attributes
also represent important avenues for future research. In this regard, the use of novel data sources
such as publication-full-text and funding data seems particular promising. For example, publi-
cation full-text data have been mainly used to improve the accuracy of standard scientometric
approaches (e.g. co-citation and co-word clustering) (e.g. Boyack et al., 2013; Glenisson et al.,
2005). However, the analysis of the full-text of publications may also provide information for
operationalising the uncertainty and ambiguity attribute of emergence. Instances of multiple
and competing envisioned applications of an emerging technology may be identified in publica-
tion sections such as the introduction and discussion, which also have the advantage of being
structured in a relatively standard manner across publications as compared to other sections.
Sentiment and narrative analysis techniques may be particularly suitable for the extraction of
this information.
Funding data may also provide relevant information for the operationalisation of emerging
technologies. For example, uncertainty and ambiguity may be indicated by more extensive pub-
lic funding than private investment. Growth in funding may indicate relatively fast growth, thus
overcoming the time lag between actual emergence and emergence detected in publications and
patents. The amount of funding can also cast light on the expected impact of the technology.
32
Relatively large investments, suggest prominent impact is expected. Nonetheless, the coverage
of funding data remains limited (Hopkins and Siepel, 2013). A number of databases (e.g. Re-
searchfish, FundRef, RCUK Gateway to Research, NIH RePORTER) have been recently built
with aim of providing access to these data. Such databases include data on funding from major
funders (e.g. government departments, research councils, large charities and foundations), but
inevitably lack information on a large variety of relatively small funding organisations that may
be important, especially in the early phases of development. The use of funding data as reported
by authors in the acknowledgements section of publications provides better coverage of funders
but no information on the amount of funding.
The use of big data and altmetrics (e.g. download statistics, number of retweets, Mendeley
readers, citations in blogs or news articles) add to the set of potential data sources. Given that
these data are produced in a ’real-time’ manner as compared to conventional scientometric data,
they seem particularly promising in enabling the development of indicators for early detection.
For example, publication download statistics can provide an early indication of relatively fast
growth and perhaps of prominent impact in the academic domain as compared to conventional
citation data. Numbers of tweets and citations in blogs or news articles may instead provide
an indication of attention outside the academic domain. Nonetheless, there is first a need to
improve our understanding of these data as well as how the data can be compared across different
cases of emerging technologies. We hope the framework offered here can be used to structure
exploration of novel data sources for the detection of emerging technology.
6 Conclusions
Emerging technologies have assumed increasing relevance in the context of policy-making for
their perceived ability to change the status quo (e.g. Alexander et al., 2012; Cozzens et al., 2010;
Day and Schoemaker, 2000; Martin, 1995). This has spurred ad hoc governmental actions such as
the ”Future & Emerging Technologies” (FET) initiative funded by the European Commission
in 2013 and the ”Foresight and Understanding from Scientific Exposition” (FUSE) research
program funded by the US Intelligence Advanced Research Projects Activities (IARPA) in
2011. The FUSE program, for example, in pursuit of potential uses of big data, has aimed
to develop methods for the reliable early detection of emergence in science and technology
by mining the full-text of publications and patents. Policy interest has been matched by the
33
academic community who have developed a variety of methods for the detection and analysis of
technological emergence in recent years especially in the scientometric domain (e.g. Gl¨anzel and
Thijs, 2012; Small et al., 2014).
Despite this broad interest, a widely accepted definition of emerging technologies and an
agreed conceptually grounded framework for their operationalisation are both still missing. We
showed that emerging technologies are either loosely defined in the empirical literature or often
no definition at all is provided. As a consequence, operationalisations of emergence tend to differ
greatly even between approaches using the same techniques. In addition, the understanding of
what is an emerging technology differs across actors: some individuals may conceive a technology
to be emergent because they expect impact on the socio-economic system, while others may see
the same technology as old and no longer emergent. This, in turn, has significant implications
for policy making and the governance of emerging technologies.
The present paper has attempted to move the field forward by systematically delineating
the concept of technological emergence linked to measurement options. To do so, we first
developed a definition of emerging technologies that is able to capture the multifaceted nature
of emerging technologies, and then proposed a framework for their operationalisation drawing on,
but not limited to, scientometric analysis. We identified five attributes of emerging technologies:
(i) radical novelty, (ii) relatively fast growth, (iii) coherence, (iv) prominent impact, and (v)
uncertainty and ambiguity, and defined emerging technologies as: a relatively fast growing and
radically novel technology characterised by a certain degree of coherence persisting over time
and with the potential to exert a considerable impact on the socio-economic domain(s) which
is observed in terms of the composition of actors, institutions and the patterns of interactions
among those, along with the associated knowledge production processes. Its most prominent
impact, however, lies in the future and so in the emergence phase is still somewhat uncertain
and ambiguous”.
We then developed a coherent and systematic framework for operationalising these attributes
of emergence. Scientometric literature was the main source of potential measures. Relevant
studies were reviewed and linked to the attributes of emergence. Our analysis showed that
scientometric analysis is particularly appropriate for the operationalisation of growth, novelty
and coherence. Relatively fast growth is operationalised in many studies and often evaluated by
counting documents over time (such as news articles, publications, and patents) (e.g. Porter and
Detampel, 1995). Radical novelty is identified with the appearance of new clusters of documents
34
or words in citation or co-word analyses (e.g. Kajikawa and Takeda, 2008), while other studies
point to the importance of also considering the extent to which the new cluster is connected to
clusters in the same year of observation or to clusters identified in previous years (e.g. Small
et al., 2014). Indicators based on entropy measures or on the appearance of new categories (e.g.
journals, technological classes, terms in institutionalised vocabularies) were identified as more
suitable for assessing coherence (e.g. Cozzens et al., 2010).
Nonetheless, important limitations exist on the scientometric contribution to the opera-
tionalisation of emerging technologies. The evaluation of uncertainty and ambiguity as well as
prominent impact is, for example, largely unexplored in scientometrics. Also, methods often
rely on few data sources, mostly publication and patent data, which tend to be not suitable for
the analysis of contemporary cases of emerging technologies these data require time to be
generated. The risk that detecting apparent emergence may be merely an artefact of selected
models adds to these limitations.
We have argued that the qualitative investigation of emerging technologies conducted by
STS researchers seems particularly promising in complementing scientometrics for the purpose
of operationalising the attributes of emergence. The mapping of expectations of emerging tech-
nologies by mean of qualitative analysis of documents such as news, review articles, and policy
documents can, for example, provide important insights on the uncertainty and ambiguity and
the prominent impact attributes of emergence, especially in the case of contemporary analyses.
STS approaches can also provide meaningful interpretation of the results of scientometrics, thus
potentially reducing the likelihood of detecting false positives or missing patterns.
We envisage a number of opportunities for future research. First, future research should pay
more attention to the origins of emerging technologies. We have limited knowledge of factors
that enable certain technologies to become emergent while other do not emerge at all. This also
extends to the research design used for the investigation of emerging technologies. Studies often
examine emerging technologies without delineating a counter-factual sample of technologies that
did not emerge but which nevertheless had the potential to emerge. Similarly, we have limited
knowledge on when a technology ceases to be emergent and what factors shape this process. Sec-
ond, the increasing access to publication full-text, funding data, altmetrics, and, more generally
big data, may provide significant opportunities for future research in scientometrics to develop
indicators and methods for the evaluation of attributes of emergence for which the current ’state
of the art’ in scientometrics can provide only a limited contribution.
35
In summary, we have showed that considerable disagreement exists on what is technological
emergence and how it should be operationalised. This has important implications for policy-
making in the context of emerging technologies (e.g. resource allocation, creation of research
programmes, drawing up of regulations), which, in turn, exerts a direct effect on the emergence
process itself. The present paper has attempted to contribute to this ongoing and urgent debate
in science policy research through conceptual clarification of the phenomenon of emergence.
This is a necessary precondition for a coherent and systematic operationalisation of emerging
technologies, for future research developments, for a better understanding of the phenomenon,
and, therefore for more informed policy-making and governance of emerging technologies.
Acknowledgements
We acknowledge the support of the People Programme (Marie Curie Actions) of the Euro-
pean Union’s Seventh Framework Programme (FP7/2007-2013) (award PIOF-GA-2012-331107
- NET-GENESIS: Network Micro-Dynamics in Emerging Technologies”). We are grateful to
Loet Leydesdorff, Nils Newman,