BIPOLARITY OR NON-EQUIVALENCE IN USERS’ ADOPTION REACTIONS: IT ALL RESIDES TO ISSUE MANAGEMENT

(1)

BIPOLARITY OR NON-EQUIVALENCE IN USERS’ ADOPTION REACTIONS:

IT ALL RESIDES TO ISSUE MANAGEMENT

University of Groningen Faculty of Economics and Business Master‟s Thesis MSc. Business Administration:

Change Management

I.J. (Irene) de Vries (S2178680) i.j.de.vries.3@student.rug.nl

Supervisor: dr. M.A.G. (Marjolein) Van Offenbeek Co-assessor: dr. J.F.J. (Janita) Vos

Date: 06 July 2016

Word Count (excluding Appendices): 15.477 ABSTRACT

(2)

INTRODUCTION

“No one knows how many computer-based applications, designed at great cost of time and money, are abandoned or expensively overhauled because they were unenthusiastically received by their intended users.”

-Markus, 1983, p. 430 Despite the citation‟s year of publication, its message is in no sense outdated. Widespread implementation failures can be identified for management information systems (e.g. Kumar & Welke, 1984, Lucas, 1978; 1981; Zmud, 1983), which have been investigated over the last two decades (Joshi, 1990; 1991). The adoption of Information Technology (IT) and its usage in the workplace are still central concerns within the Management Information Systems (MIS) research (Van Offenbeek, Boonstra & Seo, 2013). The MIS literature found that most problems regarding the success of information systems are related to users‟ attitudes and behaviours (Lucas, 1975; Ives, Olson & Baroudi, 1983; Joshi 1990). Consequently, users must be regarded as the “final customers” of Information System (IS) development activities (Joshi, 1990, p. 786), and probably for good reasons. More specifically, by viewing the implementation of an information system as a change process – where the systems‟ designers can be seen as change agents (Ginzberg, 1978; Zmud & Cox, 1979, Joshi, 1991) – the end users must be treated as change recipients (Van Offenbeek et al., 2013).

A significant implication resulting from this treatment is that sense-making becomes a central concept. This implication is acknowledged by Van Offenbeek, et al. (2013), where the authors postulated that requests for IS adoption often meet “multiple and usually conflicting motives” (Knowles & Linn, 2004, p. 141). Van Offenbeek, et al. (2013), were the first to recognize that most adoption theories did not diagnose these motives, by taking either the acceptance- (e.g. Davis, 1989; Venkatesh & Davis, 2000) or the resistance line of reasoning (e.g. Markus, 1983; Joshi, 1991). Insufficient attention has been given to a connection between these IT Acceptance- and IT Resistance research streams, since the primacy for IS implementation research lies on the IT Acceptance theories (Davis, 1989; Venkatesh & Davis, 2000; Venkatesh, Morris, Davis & Davis, 2003), in which “resistance is considered as the reverse side of the acceptance coin” (Laumer & Eckhardt, 2012, p. 64). But is this implicit assumption of bipolarity correct?

(6)

Van Offenbeek et al., 2013) With typifying ambivalent adoption reactions (i.e. resisting users and supporting non-users), the respective authors provided support for acceptance and resistance to be conceptually non-equivalent. Conceptually non-equivalent in the sense that acceptance and resistance must be acknowledged as independent, distinct behavioural dimensions, which can exist and co-occur next to each other. Stated in other terms, non-acceptance and resistance are not conceptually equivalent to each other. Rather than being “opposite ends of a bipolar continuum”, acceptance and resistance are “two distinct dimensions” (Van Offenbeek et al., 2013, p. 448).

The users‟ implication resulting from adopting this two-factor view, is that users have two pairs of options to choose from, namely “to support or to resist” and “to use or not to use” (Van Offenbeek et al., 2013, p. 439). Consequently, four groups with different user reactions could be detected (i.e. supporting- and resisting users as well as supporting- and resisting non-users), accentuating that encouraging acceptance is not equivalent to an increase in support, where trying to minimize resistant behaviour will not always evoke support (Van Offenbeek et al., 2013). Therefore, rather than perceiving resistance as the “reverse side of the acceptance coin” (Laumer & Eckhardt, 2012, p. 64), resistance must be appreciated as a separate coin, complementary to that of acceptance. In addition to conceptual non-equivalence, Van Offenbeek et al. (2013) found indications for the research streams being functionally non-equivalent. Functional non-equivalence manifests itself – among other things – through effecting different outcomes of an IS implementation, or identical outcomes in a different manner (Van Offenbeek et al., 2013). The former notwithstanding, the manifestation which is held on to in this study is that acceptance- and resistance have different antecedents (Van Offenbeek et al., 2013). As a consequence, it must be acknowledged that acceptance and resistance are both “triggered by a different set of causative mechanisms” (Van Offenbeek et al., 2013, p. 438). Stated more clearly, the indications for functional non-equivalence demonstrate that both acceptance and resistance have different antecedents –functioning as triggers –, thereby setting (divergent) causative mechanisms in motion, consequently leading to different adoption reactions in the form of acceptance or resistance. Van Offenbeek et al. (2013) identified that the antecedents of acceptance originate from a “relatively narrow context of system-in-use” (Van Offenbeek et al., 2013, p. 446), while the antecedents of support / resistance emanate from the “anticipated and experienced changes in a wider context” (Ibid.).

(7)

Offenbeek et al. (2013) (e.g. Haaksema‟s, 2014 Work Interdependencies and Digitalization antecedents). As a consequence, it can be stated that research is lacking in the provision of a deep understanding in the differences in (sources for) antecedents of prospective user adoption reactions (Jensen & Aanestad, 2007).

This lack in the provision of a deep understanding, proved the rationale for this study to test the following proposition, namely whether an association exists between the type of issues voiced by users and the users‟ prospective adoption reactions in an IS pre-implementation phase.

An exploratory research was conducted with the aim to convert the proposition into hypotheses. More specifically, use was made of secondary data from the research of both Cordes (2013) and Van den Bos (2013), for both a clustering procedure and a measure of association. A quantitative design was employed, thereby making a connection with both the Sense-making perspective as well as the Issue Management lens.

By employing the former, the study let to the following contributions:

Offering further evidence for the typology of users‟ prospective IT adoption reactions – including the ambivalent reactions – as characterized in the work of Van Offenbeek et al. (2013), thereby offering further evidence for the disproof of the bipolarity assumption (i.e. conceptual non-equivalence) as held in the primary IT Acceptance theories (Davis, 1989; Venkatesh & Davis, 2000; Venkatesh et al., 2003).

Further validating, exploring and testing Van Offenbeek et al.‟s (2013) indications for functional non-equivalence, by conducting an exploratory analysis. More specifically, the association was assessed between the users‟ prospective IT adoption reactions and the type of issues voiced by the users, thereby making a connection with both the Sense-making perspective as well as the Issue Management lens.

Generating of hypotheses for the causal relationships between the type of issues voiced and the users‟ prospective IT adoption reactions, by employing the research field of Sense-making for both interpreting the nature and building an argument for the causal direction of the respective associations respectively. Consequently, generating more understanding for the users‟ cognitive-action relationship by employing the sense-making rationale (Thomas, Clark & Gioia, 1993).

(8)

THEORY

IT adoption

Before I will go into detail with regard to the (conceptualizations of) prospective IT adoption reactions, I consider it to be necessary to delimit as well as to provide some foreknowledge with respect to the object of research.

As argued for in the work of Lauterbach and Muëller (2014), adoption can be defined at either the organizational- or the individual level of adoption. Taking the individual (prospective) user as the study‟s unit of analysis, two of the eight idiosyncratic individual adoption phases could be identified as adoption, namely adaptation (here: restricted to the period prior to the users‟ appropriation phase) and the adoption (decision). Prior to the users‟ appropriation phase – in which users commit to initial usage – prospective users first need to make a decision regarding the question whether to use the technology system or not (i.e. adoption decision) (Lauterbach & Muëller, 2014). Furthermore, Lauterbach and Muëller (2014) defined adaptation as the cognitive mechanisms regarding the prospective users‟ intentions to either adopt or use a technology, starting before the technology will be delivered to the individual user. It is only this specific period of the adaptation phase (prior to the initial usage / appropriation phase), which will be considered to be relevant for the study.

This study will hold on to the conceptual mechanism enclosing the adaptation- and adoption decision phases. Characterized as the “adoption / acceptance stream” (i.e. a positivist paradigm), this stream is initiated by raised awareness for the opportunity of-, and need for implementing a new technology, resides to the individual unit of analysis, incorporates activities for developing a positive / negative stance regarding the implementation and ends where the implementation activities commence (Lauterbach & Muëller, 2014, p. 17).

As acknowledged by Meissonier and Houzé (2010), little research has been conducted with regard to the emerging and unfolding of (prospective) IT adoption reactions, in phases prior to the implementation. Nonetheless, the studying of these processes is of utmost importance for implementers‟ anticipation on indications for (negative) future adoption reactions (e.g. resistance) (Meissonier & Houzé, 2010). The earlier users‟ intentions are encountered, the earlier implementers can anticipate in the form of undertaking “customized strategies to turn potential users into “supporting and high usage” users of the new system” (Seo et al., 2011).

(9)

This study will hold on to the prevailing literature on reactions to change, which “tend to focus either on levels of rejection of change (i.e. resistance) or [on levels of] acceptance of change” (Coetsee, 1999, p. 204-205). Consequently, the two streams of the (prospective) IT adoption reactions will be discussed, where for each stream the most prominent theories‟ antecedents, its definition and operationalization for this study will be set forth.

Adoption Reactions: Acceptance Stream

User acceptance has been a long-standing issue in MIS research (e.g. Swanson, 1974; Schultz & Slevin, 1973; Lucas, 1975; Robey, 1979; Ginzberg, 1981; Swanson, 1987; Davis, 1989), where it is still acknowledged as “one of the most mature research areas in the contemporary information systems (IS) literature” (Venkatesh et al., 2003, p. 426; Lucas, Swanson & Zmud, 2008).

Most prominent IT Acceptance Theories. In line with the work of Venkatesh et al.

(2003), Van Offenbeek et al. (2013) and Lauterbach and Muëller (2014), the most prominent IT acceptance theories – from which the predictor variables (i.e. antecedents) will be reviewed – are Fishbein and Ajzen‟s (1975) Theory of Reasoned Action (TRA); Ajzen‟s (1985) Theory of Planned Behavior (TPB); Davis‟ (1989) Technology Acceptance Model (TAM), subsequently extended by Venkatesh and Davis (2000) resulting in the Extended Technology Acceptance Model (TAM2) and Venkatesh et al.‟s (2003) Unified Theory of Acceptance and Use of Technology (UTAUT).

The rationale for choosing these models is Venkatesh and Davis‟ (2000) research, where the authors stated that the TAM “compares favourably” with the TRA and with the TPB (Venkatesh & Davis, 2000, p. 186). Furthermore, the “pervasive application” and “comprehensiveness” of the TAM and the UTAUT respectively, proved the rationale to prefer these models from among choices (Van Offenbeek et al., 2013, p. 435).

(10)

& Slevin, 1975; Robey, 1979; Ginzberg, 1981; Swanson, 1987), by the creation of “high-quality measures for key determinants of user acceptance” (Davis, 1989, p. 319). By reducing the predicting variables to two (i.e. perceived usefulness and perceived ease of use), Davis (1989) developed a “simplified frame for studying individuals’ IT adoption / use behaviors” (Lucas et al., 2008, p. 206), which has held – and still holds – a central position in the IT adoption field.

Acknowledging for the predictive strength of perceived usefulness as an usage intention antecedent Davis‟ (1989) model, Venkatesh and Davis (2000) aimed their arrows on the respective antecedent, thereby increasing the understanding of its key determinants. The authors specifically relied on social influence processes (e.g. image, subjective norm), and cognitive instrumental processes (e.g. output quality, job relevance, result demonstrability) (Venkatesh & Davis, 2000). As a response to the multitude of models available for explaining the variable in individuals‟ intentions to use the new IT system, Venkatesh et al. (2003) aimed for unification. The authors (2003) stated that – due to the multitude of models – researchers must either “pick and choose constructs” (i.e. antecedents) or “choose a favored model and largely ignore the contributions from alternative models” (Venkatesh et al., 2003, p. 426). Consequently, by reviewing, synthesizing and integrating elements of eight prominent models, the UTAUT model resulted in four determinants of both user acceptance and –behaviour, namely performance- and effort expectancy, social influence and facilitating conditions (Venkatesh et al., 2003).

Operationalization and Definition Acceptance. Being perceived as the “immediate

determinant of […] action”, where “people are expected to act in accordance with their intentions” (Ajzen, 1985, p. 12), intention is the IT acceptance focal point (Ajzen, 1991; 2002). The significant role played by intention implies that individual acceptance of technology is operationalized by using (intention for) usage as the dependent variable (Compeau & Higgins, 1995; Davis, Bagozzi & Warshaw, 1989; Venkatesh et al., 2003), where it is measured on a “unipolar continuum from non-use to high-use” (Marakas & Hornik, 1996; Van Offenbeek et al., 2013, p. 436). Furthermore, the study‟s definition of acceptance is based on the work of Burton-Jones and Straub (2006), where acceptance is defined as a “user’s employment of a system to perform a task” (Van Offenbeek et al., 2013, p. 436).

Adoption Reactions: Resistance Stream

Contrasting to the former, the body of knowledge for the resistance research stream proves to be less comprehensive (Cenfetelli, 2004; Lapointe & Rivard, 2005; Kim & Kankanhalli 2009; Laumer & Eckhardt, 2012). Lapointe and Rivard (2005) even stated that IS research has neglected the studying of the resistance stream.

(11)

conceptualized as the acceptance‟ opposite (i.e. the implicit assumption would be refuted), “studying acceptance alone will do little to provide insights into user resistance” (p. 64). A minority of researchers already recognized the relevance for studying the resistance stream in the early- and mid-1980‟s, where Keen (1981) recognized resistance as to be a critical concept and Markus‟ (1983) stated that its explanations – indifferent whether implicit and / or informal – influence management‟ and system implementers‟ actions, moreover directing their behaviour. The study‟s premise is that “better theories of resistance will lead to better implementation strategies and, hopefully, to better outcomes” (Markus, 1983, p. 430).

Most prominent IT Resistance Theories. In line with the work of Lapointe and Rivard

(2005), the most prominent IT resistance theories – from which the predictor variables (i.e. antecedents) will be reviewed – are Markus‟ (1983) Interaction Theory, Joshi‟s (1991) Equity-Implementation (E-I) Model, Martinko, Zmud and Henry‟s (1996) Attributional Model of Reactions to Information Technologies (AMRIT), Marakas and Hornik‟s (1996) Passive Resistance Misuse (PRM) as laying the groundwork for Kim and Kankanhalli‟s (2009) Status Quo Bias Perspective and the mixed-determinants model known as the Multilevel Model of Resistance, established by Lapointe and Rivard (2005).

The rationale for choosing these models is derived from the research of Van Offenbeek et al. (2013), where the “pervasive application” of Markus‟ (1983) Interaction Theory and Joshi‟s (1991) (E-I) Model and the “comprehensiveness” of Lapointe and Rivard‟s (2005) Multilevel Model of Resistance, proved the rationale to prefer these models from among choices (Van Offenbeek et al., 2013, p. 435). Furthermore, the selection is based on Lapointe and Rivard‟s (2005) rationale, where the authors relied for their multilevel, mixed-determinants model on those resistance theories which opened the “black box” by singling out four models (i.e. Markus, 1983; Joshi, 1991; Marakas and Hornik 1996 and Martinko et al., 1996) which provided definitions as well as explanations for the reason why resistance would occur and – if so – in what manner (Lapointe and Rivard, 2005, p. 462).

(12)

Joshi (1991) employed the (E-I) Model to assess the effect on user‟s equity status to evaluate change (Laumer & Eckhardt, 2012). The striving of individuals for equity, thereby comparing their net gain and relative outcomes on the account of the change with other users and the employer on a constant basis, proved the rationale for conducting the assessment (Joshi, 1991; Laumer & Eckhardt, 2012; Van Offenbeek et al., 2013).

On its part, Martinko et al. (1996) postulated that the application of the AMRIT would offer a basis for a more integrative, comprehensive resistance theory to what has been established afore (Martinko et al., 1996). The AMRIT argued for causal attributions through the interplay of the new technology with internal, intrapersonal influences (i.e. negative prior experiences, pessimistic attributional style) and external influences (i.e. co-workers and supervisors, technology characteristics, management support). These attributions – on its part – would influence outcome expectations, thereby affecting both the “individual’s affective and behavioural reactions to and use of the technology” (Martinko et al., 1996, p. 315).

In the same year, Marakas and Hornik (1996) proposed the PRM model, thereby providing an explanation for the (in)correct associations regarding threats or stresses of a new system to which an individual can respond in passive-aggressive manners (Marakas & Hornik, 1996; Laumer & Echkardt, 2012). As identified in the work of Laumer and Eckhardt (2012) “their explanation of user resistance is among the first to focus on resistance as a resulting behavior during IT implementation projects”, where the authors stated that Kim and Kankanhalli‟s (2009) work built upon this approach by identifying resistance as their Dependent Variable (DV) (Laumer & Eckhardt, 2012, p. 75). Kim and Kankanhalli (2009) postulated with their Status Quo bias theory that people prefer to maintain their current situation and / or status. For the explanation of this human beings‟ preference, three main categories could be identified, namely rational decision making, cognitive misperceptions, and psychological commitment. Where rational decision making entails the assessment of the net benefits of the change (encompassing transition- and uncertainty costs), cognitive misperceptions represent the loss aversion principle and psychological commitment encloses the contributing factors of sunk cost, social norms, and efforts to feel in control (Samuelson & Zeckhauser 1988; Kim & Kankanhalli, 2009).

By examining the former models (besides Kim and Kankanhalli‟s Status Quo bias theory), Lapointe and Rivard (2005) specified the nature of the relationships between the identified components, thereby refining the understanding of the resistance‟ multi-level nature. Consequently, five basic, semantic primitives could be singled out from the resistance definitions, identified as to be the initial conditions, the subject- and object of resistance, perceived threats and the resistance behaviours „an sich‟ (Lapointe & Rivard, 2005; Van Offenbeek et al., 2013).

(13)

both the study of Lapointe and Rivard (2005) and Van Offenbeek et al. (2013), the resistance concept is operationalized by making use of Coetsee‟s (1993; 1999) typology. By employing Coetsee‟s (1993; 1999) resistance typology, the continuum ranges from enthusiastic support and constructive cooperation through Coetsee‟s resistance classification of apathy (here: neutrality), passive- and active resistance respectively, to aggressive resistance (Lapointe & Rivard, 2005; Seo et al., 2011; Van Offenbeek et al., 2013) (See Appendix B for the pure definitions).

This study holds on to the resistance definition of Van Offenbeek et al. (2013), entailing the “behavioural reactions expressing reservation in the face of pressure exerted by change supporters seeking to alter the status quo” (Waddell & Sohal, 1998; Coetsee, 1999; Lapointe & Rivard, 2005; Meissonier & Houzé, 2010; Van Offenbeek et al., 2013, p. 436). The former notwithstanding, by treating the positive end of the resistance continuum as the study‟s focal point, resistance is assessed by employing its concept of support. Coetsee (1999) identified support – in the acceptance context – as the “saying that you are prepared to throw your weight behind it”, which implies the voting of users for the system implementation (Coetsee, 1999, p. 211). Since Coetsee (1999) specifically referred to the users‟ “saying”, the author postulated that the concept does not entail the active contribution to the promotion of the IT system (Ibid.).

Comparison IT Acceptance and -Resistance theories

The primacy for IS implementation research lies on the IT Acceptance theories (Davis, 1989; Venkatesh & Davis, 2000; Venkatesh et al., 2003), in which “resistance is considered as the reverse side of the acceptance coin” (Laumer & Eckhardt, 2012, p. 64). Cenfetelli (2004) stated that the underrepresentation of “perceptions (which) uniquely inhibit usage” is due to the implicit assumption held that “the inhibitors of usage are merely the opposite of the enablers” (Cenfetelli, 2004, p. 472). Is this implicit assumption correct? In order to assess the legitimacy of the implicit assumption, a comparison has to be made between the respective research streams, on top of those as already identified in the work of Van Offenbeek et al. (2013).

Inferring from the former IT acceptance and -resistance reviews as well as from Appendix A and B, the sources for IT acceptance‟ antecedents stemming from the work of Van Offenbeek et al. (2013) could be reviewed as well as complemented as displayed in Appendix C.

(14)

Consequently, it is argued that the implicit assumption for bipolarity could be refuted. Stated in other terms, the work of Laumer and Eckhardt‟s (2012) is supported, leading to the conclusion that “studying acceptance alone will do little to provide insights into user resistance” (Laumer & Eckhardt, 2012, p. 64).

Conceptual non-equivalence

Laumer and Eckhardt‟s (2012) statement is backed by Van Offenbeek et al. (2013), advocating the disproof of the implicit assumption, by acknowledging that “each of the acceptance and resistance research streams provides a valuable but incomplete understanding of IS adoption” (Van Offenbeek et al., 2013, p. 437). Van Offenbeek et al. (2013) took the critical stance that when users‟ prospective adoption reactions would lie on a single continuum, the studies should adhere to similar methods, attributes and theories (i.e. draw on similar underlying assumptions).

With the aim to increase understanding with regard to the relationship between, and the co-occurrence of the IT Acceptance- and IT Resistance research streams, Seo, Boonstra and Van Offenbeek (2011; 2013) created – and subsequently empirically evaluated – a framework in which the acceptance and resistance research streams were conceptually connected. A two-factor view model (i.e. ACRES-Model) was developed, in which the prospective user reactions / “intended users’ attitudes” regarding the implementation of information systems could be identified and typified (Seo et al., 2011, p. 70; Van Offenbeek et al., 2013)

FIGURE 1

Van Offenbeek, Boonstra and Seo’s (2013) ACRES-Model

(15)

independent, distinct behavioural dimensions, which can co-exist and co-occur next to each other. Stated in other terms, non-acceptance and resistance are not conceptually equivalent to each other. Rather than being “opposite ends of a bipolar continuum”, acceptance and resistance are “two distinct dimensions” (Van Offenbeek et al., 2013, p. 448). By adopting the two-factor view, it is acknowledged that users have two pairs of options to choose from, namely “to support or to resist” and “to use or not to use” (Van Offenbeek et al., 2013, p. 439). Consequently, four groups with different user reactions could be detected (i.e. supporting- and resisting users as well as supporting- and resisting non-users), accentuating that encouraging acceptance is not equivalent to an increase in support, where trying to minimize resistant behaviour will not always evoke support (Van Offenbeek et al., 2013). Therefore, rather than perceiving resistance as the “reverse side of the acceptance coin” (Laumer & Eckhardt, 2012, p. 64), resistance must be appreciated as a separate coin, complementary to that of acceptance.

Functional non-equivalence

In addition to conceptual non-equivalence, Van Offenbeek et al. (2013) found indications for the research streams to be functionally non-equivalent. Functional non-equivalence manifests itself – among other things – through effecting different outcomes of an IS implementation, or identical outcomes in a different manner (Van Offenbeek et al., 2013). The former notwithstanding, the manifestation which is held on to in this study is that IT Acceptance- and the IT Resistance research streams have different antecedents (Van Offenbeek et al., 2013). As a consequence, it must be acknowledged that acceptance and resistance are both “triggered by a different set of causative mechanisms” (Van Offenbeek et al., 2013, p. 438). Stated more clearly, the indications for functional non-equivalence demonstrate that both acceptance and resistance have different antecedents – functioning as triggers –, thereby setting (divergent) causative mechanisms in motion, consequently leading to the different research streams of IT Acceptance and IT Resistance. Van Offenbeek et al. (2013) identified that the antecedents of acceptance originate from a “relatively narrow context of system-in-use” (Van Offenbeek et al., 2013, p. 446), while the antecedents of support / resistance emanate from the “anticipated and experienced changes in a wider context” (Ibid.).

(16)

Sense-making

By considering the implementation of an information system to be a change process, Van Offenbeek, et al. (2013), postulated that requests for the adoption of information systems often meet “multiple and usually conflicting motives” (Knowles & Linn, 2004, p. 141). In their Approach-Avoidance model, Knowles and Linn (2004) acknowledged that attitude objects, goals, offers as well as opinions are “complex stimuli that engage multiple motives” (Knowles & Linn, 2004, p. 119). The authors continued by postulating that these motives can either be typified as approach motives, “pushing opinions and behaviours toward the goal”, where others can be typified as avoidance motives “pushing opinions and behaviours away from the goal” (Ibid.)

In order to understand the underlying reasoning for these motives regarding prospective users‟ attitudes, subsequently influencing their opinions as well as their behaviours, the sense-making theory can help. More specifically, sense-making is reviewed to increase our understanding of the cognitive-action relationship (Thomas et al., 1993). Moreover, by taking a Sense-making perspective, it will also become more clear why the coupling with the Issue Management lens is relevant for this study. Sense-making is regarded as a process of “turning circumstances into a situation that is comprehended explicitly in words and that serves as a springboard into action” (Weick, Sutcliffe and Obstfeld, 2005, p. 409). The respective authors identified three sub-processes incorporated in the process of sense-making, namely, “noticing and bracketing”, “labelling” and “action” (Weick et al., 2005, p. 411-412).

Noticing and Bracketing (Scanning). Sense-making starts with “noticing and bracketing”

(Weick et al., 2005, p. 411) where it must be recognized as the embryonic phase of sense-making, in which meaning is given to a process / event / object which has already taken place during the process of organizing but did not have a name (Weick et al., 2005). This phase is also recognized in the work of Thomas et al. (1993), being labelled as the scanning process. The scanning process is being considered as the “antecedent to interpretation and action” (Thomas et al., 1993, p. 241), since users first need to observe issues‟ features and attributes in order to form cognitive categories (Dutton & Jackson, 1987).

(17)

action” (Thomas et al., 1993, p. 241). The structure for attaching meaning to issues (i.e. labelling) is often the product of decision makers‟ (here: users‟) classification (Thomas, et al., 1993).

This users‟ classification is taken up by researchers, whom explained this phenomenon in the Categorization Theory, in which the cognitive categorization of users‟ observations of issues by employing natural- and social concepts is set forth (Dutton & Jackson, 1987). Moreover, Dutton and Jackson (1987) were able to establish a model in which the labelling of the issue could be connected to the organizational action inherently resulting from it. The authors postulated that the cognitive categories have impact on both the “… “cool” cognitive processing and the “hot” affective reactions of decision makers”(here: users) (Dutton and Jackson, 1987, p. 79). The former is backed by Thomas et al. (1993), stating that not only decisions, but also the motivations as well as the cognitions of users are affected by the issue categorization.

Two of the most salient strategic issue categories developed by researchers are the opportunity and the threat (Dutton & Duncan, 1987; Mintzberg, Raisinghani, & Theoret, 1976; Nutt, 1984; Perrow, 1970; Thomas et al., 1993). Dutton and Jackson (1987) identified three attribute dimensions for the discrimination of the information categorization in either opportunities or threats by: users‟ Positive / Negative Affect, potential for Net Gain / -Loss and High / Low Perception on Outcome Controllability respectively.

Tailored to the EHR technology and coupled with the sense-making process, Jensen and Aanestad (2007) identified three more themes which can be employed as conception issue categories in this study. Firstly, the Conceptions of Technology, encompassing the “images and perceptions of the EPR technology and their understanding of its affordances in relation to functionality and capability through bracketing” (Jensen & Aanestad, 2007, p. 35). Secondly, the Conceptions for Work Practices and the Professionals‟ Role, referring to as the “understanding of how the EPR will be used in their clinical practice as well as actual conditions in using the EPR in relation to their perceived identities and roles” (Ibid.). Lastly, the Conception of Implementation Issues, referring to the “professionals’ understanding of adoption aspects associated with the use of the EPR” (Ibid.).

Tailored specifically to the ambivalent (prospective) IT adoption reactions, Seo et al. (2011) identified a classification of the most prevalent issues. For resisting users, system perception could be “incompatible with their personal or organizational norms and values” (i.e. Cultural Resistance), the Loss for Autonomy could be feared and Uncertainty could be felt with regard to the consequences of system implementation (Seo et al., 2011, p. 72). For the supporting non-users, practical- and technological Issues (i.e. Materializing Support), shortfall in necessary knowledge and / or skills (i.e. Technical Barriers) as well as preference for the status quo due to the difficulty to switch (i.e. High Sunk or Switching Costs) could be encountered (Seo et al., 2011).

(18)

Action (Issue Management).

Once the labels are applied as a result of the categorizing process, subsequent motivations and cognitions are affected, subsequently affecting both the content as well as the process of the users‟ actions (Dutton & Jackson, 1987). The positive / negative attributes associated with the opportunity / threat are reflective for evaluative appraisals, which make the cognitions “hot” (Dutton & Jackson, 1987, p. 82). In turn, these evaluative appraisals (or affective tags as specified by Fiske and Taylor, 1984) “may attract people to become associated with an opportunity [approach motives] and repel people from becoming involved with an issue labelled a threat [avoidance motives] because threats are aversive stimuli from which people withdraw, while opportunities bestow status and prestige to those who deal with them” (Ibid.).

At this phase, issue management is of essential value. Issue management could be defined as the “organized activity of identifying emerging trends, concerns, or [here] issues likely to affect an organization in the next few years and developing a wider and more positive range of organizational responses toward that future” (Gaunt & Ollenburger, 1995, p. 201). Rather than reacting after-the-fact – thereby forcing to accept the new situation at hand – issue management could be employed as an identification tool for emerging issues, responding to them before they would become public (Gaunt & Ollenburger, 1995).

Consequently, the change agents‟ implications concern the employment of persuasion strategies in order to „steer‟ prospective users towards the preferred user reaction‟ typology (Seo et al., 2011). As stated by Seo et al. (2011), it must be the aim of implementers to “develop customized strategies to turn potential users into “supporting and high usage” users of new systems” (Seo et al., 2011, p. 68). By delimiting this discussion to the persuasion strategies of Knowles and Linn (2004), Alpha strategies could be used in order to increase the users‟ motivation for moving towards the IT system implementation (i.e. increasing the approach motives). In contrast – for the minimization of the avoidance motives, Omega strategies could be employed which would decrease the movement of the subject (here: the user) away from the goal (here: IT system implementation).

Proposition

Consequently of the former discussion of both the IT Acceptance- as well as the Resistance theories, along with their (conceptual / functional ) non-equivalence, the following proposition is made:

(19)

METHODOLOGY

In order to deepen the understanding regarding the former relationship(s) presence and nature, I made use of an exploratory analysis. Rather than looking for confirmations of relationships specified prior to the analysis, an exploratory analysis relies on the data and method to determine the relationship‟s nature (Hair, Black, Babin & Anderson, 2010). Since, thus far, the IS adoption research field has not been integrated with the research field of Issue Management, an exploratory analysis particularly suits this study (e.g. Brown, 2006).

Research Design

To empirically test whether an association could be identified, an exploratory research was conducted with the aim to convert the former proposition into hypotheses. Use was made of secondary, cross-sectional data from the correlational / cross-cross-sectional research of both Cordes (2013) and Van den Bos (2013) within the EPR research of the RUG Healthwise center (Blumberg, Cooper & Schindler, 2014; Field, 2013). The setting concerned the implementation of an EPR programme at a large teaching hospital in The Netherlands (hereinafter referred to as: hospital). By consultation between the researchers from the University of Groningen together with the project staff of the EPR programme, a web-based survey was developed as an element of a larger research conducted by researchers from the University of Groningen (Cordes, 2013; Haaksema, 2014).

Research site

Both the study of Cordes (2013) as well as the study of Van den Bos (2013) used primary data regarding the implementation of a new Patient Record at the hospital. During the time of the data gathering, the hospital was found to be in the early phase of the implementation process (i.e. the pre-implementation phase) of an organization-wide EPR project (Van den Bos, 2013). The objective of the hospital‟s EPR implementation regarded the replacement of the current stand-alone IS applications (i.e. legacy health-care systems), thereby attaining an integral EPR. Stated in other terms, with the implementation of the integral EPR, one digital file of each patient would exist, which would be employed by all the health professionals and administrators within the hospital (Van den Bos, 2013; Van Burgsteden, 2015).

Data and Instrument

(20)

originating from 1 („Fully Disagree‟) to 5 („Fully Agree‟). In addition to these five options, the opportunity to check: Not Applicable (i.e. N.A.) / No Opinion was provided by a sixth box in addition to the five summated ones (Cordes, 2013).

Along with these 25 statements, the survey provided the respondents with the opportunity to voice their most significant issues with regard to the implementation of the new Patient Record, by developing an open question (i.e. number 26 in the survey) (Cathain & Thomas, 2004). Subsequently, the respondents were asked to answer six more questions regarding their personal- and work situation (e.g. department, profession group, age and tenure). Stated in other terms, the survey comprised a mix of open- and closed-ended questions.

In two consecutive pilots, the survey‟s statements were controlled for construct validity, the ordering and scaling of the statements, the degree to which the statements were understandable (i.e. its difficulty) and not too time-consuming (Cordes, 2013) (i.e. check for clarity and unambiguity) (Van den Bos, 2013). Based on the pilot outcomes, alterations were made, where subsequently the survey was send by email to the participants (See Appendix D for the State-of-Mind (SoM) Survey).

Population and sample

The project staff relied for the sampling on the intensity of usage regarding the hospital‟s new EPR system, by identifying the health professionals and administrators that would become the most active users of the new EPR system (Cordes, 2013; Van den Bos, 2013).

The participants who were approached in first instance proved to be the Management Staff and the following profession groups: Nursing Staff, Doctors, Paramedics / Perimedics (Para/Peri) as well as the Administrative personnel. In addition to these participants, the following departments‟ employees were chosen: Obstetrics (plus Gynaecology), Ophthalmology, Orthopaedics and Internal Medicine (including Geriatrics). Moreover, a random sample was drawn from the profession groups, ultimately resulting in a total number of 1.579 respondents whom were asked to participate (Cordes, 2013; Van den Bos, 2013). By ultimately receiving a response from 587 (Cordes, 2013; Van den Bos, 2013), the response rate was calculated to be 37.18 percent.

Descriptives Observations

(21)

the respondents‟ industry. Furthermore, the nursing staff signalled to be the largest profession group, where more than two-third of the staff proved to be female (68.7 %).

(22)

ANALYSIS

In order to develop constructs for subsequent multivariate analysis (i.e. cluster analysis and measures of association) with “well-rounded perspectives” (i.e. representing all the different aspects of the concept), composite measures were created through data reduction (Hair et al., 2010, p. 8). Data reduction relies on factor analysis as the empirical basis for the composite measures, by clarifying the structure which underlies the data as well as the variables‟ interrelationships (Hair et al., 2010).

In contrast to the former data reduction, this study could not rely on a quantitative procedure for the issues voiced. The respondents‟ issues were measures as non-metric data in Cordes (2013) and Van den Bos (2013). In order to represent metric categories, the data had to be encoded in dummy variables. The respective dichotomous nature of the data proved to be less appropriate for the quantitative data reduction procedure (Hair et al., 2010). Consequently, a qualitative data reduction procedure was employed, by inductively and deductively coding the issues‟ nature into categories.

Data reduction

Quantitative Procedure. Composite measures were created for the constructs representing

the prospective IT adoption reactions, by making use of data reduction. With the aim of getting an indication of the strength and direction of the bivariate relatedness, a two-tailed Pearson Correlation was conducted after reverse scoring a few items (Hair et al., 2010; Field, 2013). Furthermore, the degree of internal consistency was determined, by relying for the assessment of the Cronbach‟s Alpha‟s significance on the threshold level of .70, as identified by George and Mallery (2003). However, the cut-off point was a bit more nuanced, since Kline (1999) noted that with psychological constructs – measuring diverse items – values below .70 could realistically be expected (Field, 2013). When the items proved significantly related and internally consistent, the factor analysis was conducted. Firstly, the items‟ communalities were assessed, where the desired threshold was set on .50 indicating that at least half of the items‟ variance was explained by the latent factor (Hair et al., 2010). Secondly, the items‟ loadings on the latent factor were evaluated, where Stevens‟ (2002) tresholds were held on to (e.g. .298 for the sample size being 300, and .210 for the sample size being 600 respectively). Lastly, the percentage of the variance criterion was assessed, where –for social sciences – a cumulative variance of 60 percent was considered to be satisfactory (Hair et al., 2010).

(23)

Acceptance.

The degree of acceptance (i.e. users‟ willingness for prospective usage of the

new system) is measured by means of two items concerning „using EHR as quickly as possible‟ (i.e. Ac_1) and „actively getting to work with EHR when available‟ (i.e. Ac_2), demonstrating their significant positive relatedness (r = .713, ρ <.01), internal consistency (α = .829), thereby explaining more than 85 percent of the variance of the aggregated construct (85,634 %). Moreover, the loadings can be interpreted to be highly significant, by demonstrating to be above Stevens‟ (2002) thresholds of .298 for the sample size being 300, and .210 for the sample size being 600 respectively (Field, 2013). With significant component loadings of .925 with the sample size being 536, furthermore explaining more than 85 percent of the aggregated construct, it can be inferred that the factors encompass convergent validity (Field, 2013). Consequently, a new aggregated construct is being calculated by averaging the scores on all items (see Table B.1, Appendix F) (see Appendix D for codebook survey).

Support.

The support items in the questionnaire measure the degree to which the respondents

favour the implementation of the new technology system and are stating that they would be prepared to “throw [their] weight behind it” (Coetsee, 1999, p. 211). For the proper interpretation, I reversed the conceptualization of the support scale for the profiling of the clusters (see under the heading: Profiling the Clusters) to such an extent that high values on the Likert scale would indicate the favouring of the implementation of the new Patient Record by the respondents (i.e. enthusiastic support).

(24)

After developing the new aggregated constructs for both Acceptance as well as for Support, the descriptive statistics could be calculated. Inferring from Table B.3 (displayed in Appendix F), it can be seen that the new constructs for both Acceptance as well as for Support lie within the tolerable range for univariate normality, since skewness and kurtosis only indicate non-normality issues when they are greater than 3.00 and 10.00 respectively (DeCarlo, 1997; Kline, 1998; Widener, 2007).

Qualitative Procedure. I could not rely on the respective quantitative procedure for the

issues voiced by the respondents. Considering the fact that the issues voiced by the respondents, which were measured as non-metric data in Cordes‟ (2013) and Van den Bos‟ (2013) research, had to be encoded in dummy variables in order to represent metric categories, the dichotomous nature proved to be less appropriate for the quantitative data reduction procedure (Hair et al., 2010). A factor analysis proved to be less appropriate (Hair et al., 2010), concerning the potential for „difficulty factors‟, in which the items‟ identical distributions, rather than the content-similarities are loaded on to a factor (Gorsuch, 1983). Ergo, the baseline is the assumption underlying the factor analyses, thereby calculating the matrix of association by assuming the data to be interval or ratio (Starkweather, 2014). Furthermore, assessing the variables‟ internal consistency by conducting a reliability analysis would also be less suitable. Even by calculating the Kuder-Richardson (KR20) coefficient as an alternative for the Cronbach‟s Alpha, taking the dichotomous nature of these issues together with its „green‟ character by being in the early stages of its research, practically all values could be argued for (e.g. Nunnally, 1978, Kline, 1999). Therefore, it was decided to rely on a qualitative data reduction procedure instead.

A qualitative data reduction procedure was employed, by deductively as well as inductively coding the issues nature into categories. The qualitative data reduction procedure was used in a primarily quantitative way, where the dummy-coded categories formed the basis for the subsequent statistical, multivariate analysis (Marks & Yardley, 2004).

Issues Voiced. Considering the fact that, thus far, the IS adoption research field has not been

integrated with the research field of Issue Management, the selection of issues proved particularly suitable for this study, in which the focal point is on the exploratory nature of the research. As argued for in the work of Brown (2006) “Exploratory research tends to tackle new problems on which little or no previous research has been done” (p.43). For the categorization of the issues, I relied on the typology of Van den Bos (2013), subsequently refined by senior researchers.

Issues.

The qualitative data reduction procedure began with deductive coding. By relying on

(25)

category (Employee commitment), and the inclusion of Administrative personnel as an additional profession group, ultimately resulting in a total of 928 issues. As displayed in the coding process (see Appendix G), I became more familiar with the categorization, consequently also more critical.

(26)

RESULTS

The former coding of data formed a proper starting point for a statistically based analysis such as measures of association (Marks & Yardley, 2004). The reasoning for the employment of a cluster analysis was fourfold. Firstly, the respective analysis was used to offer further evidence for the ACRES typology of Van Offenbeek et al. (2013). Secondly, the cluster analysis was used as a means to simplify the observations (i.e. the data) for the subsequent exploratory research conducted (i.e. the measures of association). (Hair et al., 2010; Field, 2013). Thirdly, the cluster analysis laid the groundwork for testing the proposition by exploring the association between the respondents‟ adoption reactions (i.e. defined clusters) and their type of issues voiced. Consequently, by structuring the data through the cluster analysis, the fourth objective could be realized, namely the generating of hypotheses related to the causal direction of the respective relationship (Hair et al., 2010).

Clustering Procedure

The clustering procedure regarded the combination of an hierarchical- as well as a non-hierarchical procedure. The hierarchical procedure generated a comprehensible set of cluster solutions and established a suitable number of clusters, where the non-hierarchical procedure refined the cluster number obtained, by selecting the definite cluster memberships (Milligan, 1980; Hair et al., 2010).

For the hierarchical clustering procedure, the agglomerative method was used, with the employment of the Ward‟s method as the clustering algorithm in combination with the Squared Euclidean distance being the similarity measure. Rather than relying on a single measure of similarity, the Ward‟s method was chosen for its robust procedure (Hair et al., 2010).

To select the appropriate number of clusters to represent the data structure, while at the same time functioning as the input for the non-hierarchical analysis, the measures of heterogeneity change were used as the stopping rule. In order to assess the exact increase in heterogeneity, the agglomeration coefficient was employed (Hair et al., 2010). By taking the average proportionate increase as a rough guidance (here: 26.501 percent), the cluster solution can be regarded as the final solution at the respective stage where the combination of the two previous clusters (i.e. from one stage to the next) would be significantly larger than the proportionate increase at the previous stage(s) (Hair et al., 2010). The proportionate increase from four to three clusters demonstrated to be significantly large (42,74 percent), favouring the four-cluster solution over three-cluster-solution, indicating the stopping point (see Appendix J for the agglomeration schedule; See Appendix K for the forming, validating and profiling of the clusters).

Interpreting the Clusters

(27)

clusters were profiled. Considering the aim of the clustering being the further validating of past research, the typology as used in the work of Van Offenbeek et al. (2013) was held on to, leading to the interpreting of the following clusters (see for a graphical representation of the derived clusters as represented in the ACRES-Model Figure 2).

Cluster 1: Supporting users (n = 159). These health professionals and administrators

display positive adoption behaviours, in the sense that they state to favour the IT implementation and demonstrate their willingness for prospective use of the EPR (M = 3.979; 4.720), (SD = .538; .408) respectively. Where the predictor variable age proved not to be significantly different for this cluster in relation to other clusters, the predictor variable of the employees‟ profession group did (χ2 = 19.898, d.f. = 4, ρ = .001). Management demonstrates to be overrepresented in this cluster (expected count: 9.7), also signalling more Para- and Perimedics than expected (expected count: 27.9). Moreover, Doctors are well-represented as supporting users (expected count: 46.7) (see for the observed counts, Table C.3 in Appendix K).

Cluster 2: Resisting non-users (n = 68).

The hospital‟s employees whom reside to this cluster display negative prospective adoption behaviours in that they are not favouring (i.e. resisting), nor are willing to use, the new Patient Record (M = 2.172; 2.449), (SD = .635; .647). The employees within this cluster demonstrate to be the oldest in general (M = 46.118, SD = 11.180), where the respective predictor variable marginally differed with the third cluster (ρ = .104). Furthermore, the profession group characteristics demonstrated to be significantly different for cluster two in relation to the other clusters (χ2 = 28.950, d.f. = 4, ρ = .000). More specifically, the cluster is underrepresented by Doctors (expected count: 20) and Administrative personnel (expected count: 11.9), where the Para- and Perimedics are overrepresented in the cluster (expected count: 11.9).

(28)

Cluster 4: Supporting / neutral (low) users (n = 153). The respective employees

display a more modest adoption reaction in comparison to the former clusters. For both the Support- as well as the Acceptance of the new Patient Record, the employees signalled moderate values for the prospective adoption reactions (M = 3.314; 3.677), (SD = .373; .374). While profession group did not significantly differ to other clusters (χ2 = 4.105, d.f. = 4, ρ = .392), the employees‟ age demonstrated to be significantly different to the former cluster (i.e. cluster three) (ρ = .081). Just as the scores on the adoption reactions, the age of the employees was not extremely high or low (M = 45,517, SD = 11,944), thereby more belonging to the average of the sample as displayed in Table 1.4 (see Appendix E) (M = 44.26, SD = 11.370).

FIGURE 2

Clusters represented in Van Offenbeek, Boonstra and Seo’s (2013) ACRES-Model

Descriptives Issues voiced per cluster

(29)

variables (see under heading: Issues Voiced) a Chi-square cross-tabulation was conducted of the clusters on the issues voiced by the respondents. Inferring from Table D.1 in Appendix L, a total of 776 valid issues could be observed (both in numbers as well as percentages), voiced by the respondents distributed over the different clusters. These absolute values could only be interpreted by dividing the total of issues voiced per cluster by the total number of respondents belonging to the respective cluster. By summing the issues voiced over all four clusters (n = 776), consequently dividing this total by the total of respondents adhering to all four clusters (n = 493), the findings demonstrate that – on average – one and a half issue was voiced per respondent ( ̅ = 1.574). The respondents adhering to cluster one (i.e. the supporting users) named the most issues on average ( ̅ = 1.805), followed by cluster two (i.e. the resisting non-users) ( ̅ = 1.691), cluster three (i.e. the resisting users) ( ̅ = 1.434) and cluster four (i.e. supporting / neutral (low) users) ( ̅ = 1.386).

In order to assess the data‟s general tendency per cluster, the median is calculated considering the non-symmetrical distributed nature of the data per cluster. By calculating the median as an approximate of the average of the issues voiced per cluster, it could be examined whether respondents adhering to a particular cluster named many or few issues adhering to the different issue categories or whether the issues were voiced with a high or low frequency.

A high value for the median is interpreted as the situation in which the respondents voiced many issues adhering to some specific issue categories (i.e. acknowledging the significance of (one or more) issue categories in particular), and a few (or none) adhering to other, or the situation in which all issues were named very frequently by the respondents in the respective cluster. Reasoning by analogy, a low value for the median is interpreted as the situation in which the respondents did not acknowledge the significance of issues in particular, thereby voicing issues adhering to many (or all) of the different issue categories, or the situation in which all issues were not voiced in a frequent manner by the respondents in the respective cluster.

By calculating and subsequently displaying the median‟s value per cluster in the Figures B.1-B.4 in Appendix L, indications are provided for the association between the type of issues voiced and the respondents‟ adoption reactions. These indications laid the groundwork for the measures of association per cluster (see Appendix L for the indications).

Measure of association

(30)

procedure. Taking into account that the issues were also encoded into dummy variables (see under heading: Issues Voiced), bivariate measures of association had to be conducted between two variables with nominal (more specifically: dichotomous) measures.

The Phi coefficient was used as the measure of magnitude of the association. The Phi coefficient is a Pearson product-moment, suitable for nominal-dichotomous variables (Warmbrod, 2001). The advantage of the Phi coefficient over the Cramer‟s V – which is also a measure of association for nominal-dichotomous variables – is its indication for the nature of the association (either being positive or negative) (Kraska-Miller, 2013). However, due to the fact that the Phi coefficient relies on the marginal proportions of the variables which have to be correlated, it must be acknowledged that perfect associations (in the form of ± 1) would be hard to attain as well as to defence (Davenport & El-Sanhurry, 1991). Furthermore, it must be kept in mind that the Phi coefficient – just as the Cramer‟s V statistic – is a symmetrical measure (Kraska-Miller, 2013). This implies that no inferences can be made with regard to the direction of the relationship. Stated in other terms, the variables on both the rows and columns of the association matrix could be interchanged, thereby not losing any of its interpretational value (Kraska-Miller, 2013).

The Phi coefficient was computed per issue category on the four clusters derived afore. By holding on to the 99-, 95- as well as the 90 percent confidence interval, most issues demonstrated to be significantly related with one or more clusters. However, by taking a critical look at the Phi coefficients, a considerable amount of the significant associations could be identified as to be „negligible‟. In line with the associations‟ magnitude conventions developed in the work of Rea and Parker (1992), values between .00 and under (-).10 were regarded to be negligible (see Appendix M). With the rest of the significant values ranging between (-).10 and (-).20, the associations could be defined as „weak‟ (Rea & Parker, 1992).

(31)

Qualitative clustering

(32)

DISCUSSION AND CONCLUSION

Theoretical Implications

This study offered further evidence for the ACRES typology of Van Offenbeek et al. (2013).With the exception of prospective non-usage, the empirically derived clusters confirmed the respective two-factor view (see Figure 2). Particularly, by demonstrating users adhering to the resisting usage‟ cluster – an ambivalent prospective adoption reaction –the disproof of the implicit assumption of bipolarity could be advocated for. Consequently, Van Offenbeek et al.‟s (2013) postulation that “each of the acceptance and resistance research streams provides a valuable but incomplete understanding of IS adoption” (p. 437) could be further validated.

The former laid the groundwork for further exploring the indications for functional non-equivalence. The association was assessed between the users‟ prospective adoption reactions (i.e. the defined clusters) and their type of issues voiced, thereby making a connection with both the Sense-making perspective as well as the Issue Management lens. The results demonstrated functional non-equivalence in some respects, where support could be found for “acceptance [being] triggered by a different set of causative mechanisms than resistance” (Van Offenbeek et al., 2013, p. 438). The results signalled the presence of different causative mechanisms for both the IT Acceptance and the -Resistance research stream (i.e. Functional Non-equivalence), where significant associations exists between the type of issues voiced and the supportive usage- and resisting (non-usage) prospective adoption reactions. Albeit the former, the associations between the type of issues voiced and the ambivalent user adoption reactions (here: resisting usage) and supportive / neutral (low) usage prove to be negligible. As a consequence, the study‟s proposition that an association exists between the type of issues voiced by users and the users‟ prospective adoption reactions in an IS pre-implementation phase holds partially.

The last objective of this study that rests, is the generation of hypotheses. To that end, the nature of the associations between the prospective IT adoption reactions and the issues voiced will be interpreted below. Subsequently – after building an argument for the relationships‟ causal direction – the hypotheses could be generated.

(33)

Conceptions of Technology

Supportive Usage, Positively Related. The three issue categories which were more

voiced by users with supportive usage as their prospective adoption reaction, are Integral EHR, User friendliness and New features and expectations (See Appendix I for the refinement of the Codebook). For interpreting these relations, Jensen and Aanestad‟s (2007) category of Conceptions of Technology is employed (as defined afore). Jensen and Aanestad (2007) related the affordances to the extraction of cues as the sense-making property. When the cues for affordances of the technology are enlarged through the design of the system, sense-making will become focused by these cues. (Nijhof & Jeurissen, 2006; Jensen & Aanestad, 2007). For the Integral EHR issue category, the EPR affordance resides in the system‟s ability to assist in the creation of a better patient-overview, where the EPR affordance for the User friendliness issue category exists in the conception of the EPR system to ease the users‟ work (Jensen and Aanestad, 2007). The EPR affordance for New Features and Expectations resides in the functionalities of the technology bound to the users‟ classical work practices, such as the retrieval of information and purposes for documentation. (Jensen & Aanestad, 2007).

Consequently, this study argues for a high users‟ perception of Dutton and Jackson‟s (1987) Net Gain (Nijhof & Jeurissen, 2006; Jensen & Aanestad, 2007). Considering Dutton and Jackson‟s (1987) orthogonal nature of issue categorization, it is argued that the respective issue categories are perceived as an opportunity through its extraction of cues, which would focus on the technological affordances.

Albeit the fact that in this study no inferences can be made with regard to the causal direction of the relationships, I draw on Dutton and Jackson‟s (1987) general process model – developed by integrating the interpretive view with the categorization theory – for conceptualizing the relationship between categorizing issues and organizational actions. It is argued that once the opportunity labels are applied as a result of the categorizing process, subsequent motivations and cognitions are affected, subsequently affecting both the content as well as the process of the users‟ actions (Dutton & Jackson, 1987).

With the positive attributes associated with the opportunity being reflective for approach motives (as argued afore), “movement toward the goal” is promoted, making the positive relatedness with supporting usage plausible (Knowles and Linn, 2004, p. 218; Dutton & Jackson, 1987). In line with this explanation, this study will hold on to Thomas et al.‟s (1993) basic model for sense-making in terms of the “Scanning (Noticing and Bracketing) – (Issue)Interpretation (Labelling) – Action – Performance sequence” (Thomas et al., 1993, p. 240) as the rationale for the relationships‟ causal direction.

Consequently, the following relations can be hypothesized:

BIPOLARITY OR NON-EQUIVALENCE IN USERS’ ADOPTION REACTIONS: IT ALL RESIDES TO ISSUE MANAGEMENT