A CASE STUDY ON THE FACTORS AFFECTING AFFORDANCE POTENCY DURING THE DEVELOPMENT OF A BUSINESS INTELLIGENCE SYSTEM
master thesis, MSc, Change Management
University of Groningen, Faculty of Economics and Business
February, 2019 LEANDER POTZE Studentnumber: 2377187 Nieuwe Ebbingestraat 75 9712 NG Groningen Tel.: +31 (0)6 13 89 99 69 e-mail: l.n.potze@student.rug.nl Supervisor prof. dr. E.W. Berghout
Co-assessor prof. dr. A. Boonstra
1 Abstract
Due to the growing interest in artificial intelligence solutions in business intelligence systems, it is desirable the enhance the understanding regarding which mechanisms and structures influences the interaction between such systems and the user. This study focused on the strength between relationship between the ability of individuals and implemented system features of a new AI-based BI system, which is called affordance potency. Affordance potency should be as strong as possible in order to reach the full potential of a BI system. By using the critical realist perspective, the study found that the characteristics of the work environment during the development of a BI system are essential to increase affordance potency. In this case study, the impact of feedback is essential in order to increase affordance potency. On top of that, mechanisms and structures should be in line with the properties of the BI system, the users, and the power of their interaction. Only then potency for an actualization of the planned affordance will be strengthened. First, this study will find mechanisms and structures that strengthen that potency. Second, this study enlarges the understanding on how affordances actualization is affected by user perceptions and user goals.
2 Introduction
Due to the increasing dynamism and complexity of the modern-day world the interest in complex analytical applications, which is described as business intelligence (BI), has risen across sectors and in academia (Wixom et al. 2014). Luftman and Ben-Zvi (2010) argued that investments in BI systems has been seen as the largest and fastest growing areas of information technology (IT) expenditures. BI is most often used as an ‘umbrella’ term that describes a process or method which often refers to terms such as business analytics, big data, data warehousing and data mining (Trieu, 2017; Wixom and Watson, 2010). BI allows users to access and use data to support decision-making by presenting knowledge, providing effective analysis and intuitive presenting the right information (Popovič et al. 2012). According to Davenport (2006), BI systems are seen as a source that can be of great value due to their analytical nature and could enhance the competitive advantage of an organization.
Despite the investments, organizations are struggling to achieve expected outcomes and benefits after the transition from legacy IT systems to new BI systems (Malladi & Krishnan, 2013). A lot of research has been done to information systems (IS) and its success (Chau et al., 2007). However, research in the area of BI is lacking and needs further attention (Dremel, Herterich & Spottke, 2017). More research might be needed in order to get to know which capabilities and competencies are required to profit from BI (Dremel, Overhage, Schlauderer & Wulf, 2017). Yeoh et al. (2008) have found which factors and elements are crucial for successful BI implementation. Some years later, Popovič et al. (2012) provided an in-depth analysis of BI systems due to a deeper understanding of the variables affecting a BI system. However, no study has found the characteristics of the working environment that influences the use of a BI system individually.
3
affordance actualization, an organization can provide information to change user’s perception of affordances (Bernhard et al., 2013) or organizations can improve affordance potency (Anderson & Robey, 2017). Nevertheless, the research gap how affordance potency can arise remains. In the context of BI with an artificial intelligence (AI) nature, no research has been conducted about the factors that increases affordance potency. To address this underdeveloped topic, this research will provide new knowledge about the characteristics of the working environment that affect affordance potency during the transition from a legacy IT system to a new AI-based BI system in an organizational context.
Users can attach different meanings to IT (Shaikh & Karjaluoto, 2015). Consequently, the risk that several users are not using IT features as expected is likely, whereas ineffective use of IT features jeopardize the potential opportunities of an Information System (IS) (Li et al., 2013). Therefore, the transition of a legacy IT system to a of new BI system could fail easily, because former affordances between users and legacy IT systems need to be considered during the developed of a new system. By exploring the factors that influence the strength of the relationship between abilities of the individual and the features of a BI system, the formation of affordance potency can be explained. This results in the following research question: “How does affordance potency increase during the development of an AI-based BI system?”.
Previously, affordance potency has been only investigated in health care industry (Anderson & Robey 2017). This study can expand the knowledge on how organizations can increase affordance potency of BI systems and emphasize external factors that might enhance that enlargement. Furthermore, by tracing the antecedents of user abilities and system features, further evidence of the existence of affordance potency is provided. In doing so, the affordances actualization theory of Strong et al. (2014) is further explained by enhancing the knowledge on how affordances are actualized. Furthermore, this study will enhance the understanding of the encompassing theory of imbrication (Leonardi, 2011), which explains how IT affects work practices. Therefore, the power of the concept of affordance theory will be strengthened.
4
developers should be aware of the importance of an interface that is clear and easy to understand, to ensure affordance potency.
In the next chapter, the depiction of the concept of affordances and BI systems will be described. Subsequently, the methodology, the research design and the results will be provided. Finally, the results will be discussed, including the theoretical and managerial implications, limitations and further possible research directions.
Theoretical background
Business Intelligence
Organizations and technologies are currently experiencing enormous changes in function and form. The origin of the term technology can be found in the Greek language. Technology is derived from the Greek τέχνη (tekhnē), which was used to emphasize art or craft, and -λογία (-logia), meaning knowledge. Nowadays the term includes artefacts as well, therefore, the modern use of the term technology is etymologically incorrect but tremendously deep-seated. Since 1950, the utilization of the term technology has risen significantly. Besides the true sense of the term, technologies change through logic obtained by themselves.
5
to be adapted, accepted, routinized and extensively used during business activities of the organization (Zhu et al. 2006).
According to Watson et al. (2002), the main benefits of a BI system are the presented information for supporting decision making and time savings. Gut feeling, or intuition-based decision-making will be reduced to the bare minimum. Despite that BI systems can gather and analyze large quantities of unstructured data, users must understand the system features s in order to reach the full potential of that BI system. Currently, there is a lack of understanding of BI technologies and the user-based decision support environment because organizations are applying artificial intelligence to improve traditional applications (Cearley et al., 2017). AI consists of scoped machine-learning solutions that is a developed to optimize a specific task. Several algorithms are chosen and developed in order to do so. AI-based BI systems have the potential to transform the nature of work by helping users to make better business decisions. Affordance Theory can help to better understand the interaction of an AI-based BI system and the goal-oriented users.
The Principles of Affordance Theory
6
“An affordance is the potential for behaviors associated with achieving an immediate concrete outcome and arising from the relation between an artefact and a goal-oriented actor or actors.” – Strong et al. (2014, p. 69)
The concept of affordance indicates the relation between object and human users. Their interplay leads to different uses and outcomes; hence it is seen as a relational understanding of a human-technology interaction (Volkoff & Strong, 2013). So, what an IT system affords to users and how users interpret the system is an open negotiation between those two (Glowalla et al., 2014). Affordances are the possibilities for goal-oriented actions and afforded to specified groups of actors by technical objects (Markus & Silver, 2008). Technical objects (TOs) are IT artefacts together with their component parts (Markus & Silver, 2008). IT artefacts are the outcome of manufacturing processes and intentional design (Faulkner et al., 2010) with characteristics that possibly have causal potential (Markus & Silver, 2008). On top of that, affordances can both constrain and enable (Volkoff & Strong, 2013).
For example, a bridge affords humans to walk, cycle or drive between two points that are normally separated by water, gaps, roads or other obstacles by making a connection between those two points. The bridge does not own the ‘accessibility affordance’, except in relation to an actor. So, the characteristic of accessibility is not inherent to a bridge, but needs to be evoked by the actor. Affordances arise from the features of an artefact and actor’s capabilities and characteristics. Some affordances might not be obvious at the first glance. Sometimes actors need to recognize or need to be aware of the, for example, ‘bungee jump affordance’ before they can actualize it.
7
possible negative side-effects of affordances. Constraints, the mirrored concept of affordances, are considered as a relational concept between an user and a TO which limits goal-oriented actions (Leonardi, 2011). Since the relation between information technology and the user is crucial, a constraint may apply to all users or it may only limit to specific users. Several characteristics can serve as a basis for a constraint, like the situation, goals or due to the abilities of the user. According to Leonardi (2011), constraints can be seen as an indicator for potentials of future IS development. In case of the bridge the constraints are that it could obstruct free movements of boats or trucks underneath the bridge. By building a drawbridge, a more advanced technology, one could eventually create an even more extensive accessibility affordance.
8
the perception of affordances do not make sense to them. This study enlarges the understanding on how affordances actualization is affected by user perceptions and user goals.
According to Warren (1984) individuals can perceive environmental objects in relation to their action capabilities in order to detect the most economical courses of action in a situation. Anderson and Robey (2017) were the first researchers who implemented that theory in the field of technology affordances. They came up with affordance potency, which is the “strength of the relationship between the abilities of the individual and the features of the system at the time of actualization, conditioned by the characteristics of the work environment” (Anderson & Robey, 2017, p.103). Therefore, affordance potency is the ease of the individual to actualize an affordance. An affordance with a weak potency would require a higher-level energy to actualize than a similar with a strong potency (Anderson & Robey, 2017). As abovementioned, the affordance potency depends on the possible interaction of the implemented system features and the actual user abilities. Therefore, the affordance potency is unique for everyone, because it is based on the capabilities and abilities of the individual.
9
of the work environment are most likely to be exposed and understood by using the critical realism (CR) philosophy.
Table 1 – Model of affordance potency (Anderson & Robey, 2017)
Critical Realism and Affordance Potency
CR is a philosophy that found support at scholars by the writing of Bhaskar (1975). CR is often explained as the philosophy that is situated between interpretivism and positivism (Zachariadis et al., 2013). An important proposition of CR is that there is a reality that exists independent of human perception and cognition (Tsang, 2013; Wynn Jr. & Williams, 2012). Both social entities and the investigation of these entities are differentiated, structured and changing. These interpretations of the social entities and the social entities itself exists independently of each other (Bhaskar, 2013).
10
Standing, 2017). In the field of IS, structures can be a set of rules and practices, contracts, shared norms, values, (Wynn Jr. & Williams, 2012; Elder-Vass, 2005) language, culture, technological artefacts and applications (Fleetwood, 2005) Eventually, the interaction of these mechanisms and structures causes the events that occur or not occur. Mechanisms explain under what conditions event ‘A’ causes event ‘B’ (Zachariadis et al., 2013). Those events form the domain of the actual (Bhaskar, 1975). Finally, the subset of events that that occur and are observed and recognized, become the domain of the empirical, which is often used for scientific research. The empirical consists the knowledge and beliefs about the different entities (Wynn Jr. & Williams, 2012), the transitive dimension. To simplify, this world exists in an independent reality, but our knowledge about that world is fallible, subjective and socially constructed (Mingers, 2004; Leonardi, 2013b). Moreover, our understanding of specific structures and mechanisms is lacking, because we are not able to directly observe them (Wynn Jr. & Williams, 2012).
CR is helpful in the field of IS, because it focuses on what happened in the empirical domain and how humans interpret the socio-technical world (Henfridsson & Bygstad, 2013). This philosophy would seem to generate a more comprehensive understanding of complex phenomena, such as BI systems, by determining which structures, mechanisms and properties affect events and experiences. An affordance is a power of a TO, such as a BI system, and depends on the properties of that same TO and the user (Leonardi, 2011). Therefore, these affordances lead to the possibilities and the opportunities provided by the TO. In the context of this research, the key question is to what extent mechanisms and structures causes the event of the actualization of affordances. In order to increase affordance potency, those mechanisms and structures should be in line with the properties of the BI system, the users, and the power of their interaction. As a result, the potency for an actualization of the planned affordance will be strengthened. This study will find mechanisms and structures that strengthen that potency.
Methodology
11
The understanding how individuals can exploit the full potential of a BI is not explored. In doing so, future implementation of BI systems has a greater chance of being successful. Therefore, it is desirable that a more comprehensive understanding will be constructed regarding which mechanisms, structures and properties influence the interaction between BI systems and users. The goal of this paper is to provide answers to the research question: “How does affordance potency increase during the development of an AI-based BI system?”.
12
Methodological Principle Description
Explication of Events Identify the events being studied from
experiences in order to understand what really happened
Explication of Structure and Context Identify components of social and psychical structure, contextual environment and their joint relation
Retroduction Identify and explain the powers of structures
that may have generated events
Empirical Corroboration Ensure that the proposed mechanisms have
causal power and that they fit with your data.
Triangulation Apply several approaches to support the
analysis by using a variety of data sources.
Table 2 – Five principles of a critical realist case study (Wynn & Williams, 2012)
Case Description
This research took place at a multinational with more than two hundred thousand of professionals working in independent firms throughout the world. Collaboratively they provide consulting, audit and assurance, risk management, risk and financial advisory and tax related services to selected clients. They are active in several industries such as financial services, consumer, government and public services, energy, resources and industrials, life science and health care, media and telecommunications and technology. In general, each independent firm has their own managers and partners, and is structured in accordance with national regulations, laws and other factors. Although this firm is active in one hundred fifty countries, the focus of this research is on their office in Amsterdam, the Netherlands. Especially, the development team which consists of approximately 15 employees.
13
The firm developed a BI system from scratch. At the end of the development process, the BI system allows users to access and to use data to support decision-making by providing additional insights into tax law through network visualizations, automatically generated case summaries, statistical analytics and predictive models. Those advanced analytics is made possible through the AI-based solution of the system. Therefore, the system can automatically structure unstructured data like judgements, can find links between those structured data, can generate summaries and can predict the judgment of European court. Moreover, the BI system encompasses the ability for consolidating, analyzing, gathering and providing access to tax cases and let tax analysts make better and faster business decisions for their customers.
In January 2017, the preparation phase started (Table 3). Afterwards, the firm treated the BI system as a start-up and they managed to create the first proof of concept (poc) within one working sprint of six weeks. The delivery of poc-version 4.0 has been delayed from the 1st of September 2018 to the 1st of October 2018. The development of poc-version 4.0 is commonly seen as an overrun of version 3.0, because the main proposed features and improvements of poc-version 3.0 were not fully completed. Around that point in time, super-users of the firm, the Tilburg University and the Vrije Universiteit Amsterdam gave constructive feedback. Therefore, they managed to improve the design, the engine and the validity. Furthermore, most data are collected in the period when the development of poc-version 4.0 was recently started. Therefore, data collection of the outcome and the processes during the development of poc-version 4.0 is limited. Hence, poc-version 4.0 is not incorporated in the data analysis of this case.
Events Time frame
Preparation phase January 2017 – April 2017
Poc-version 1.0 April 2017 – August 2017
Poc-version 2.0 August 2017 – December 2017
Poc-version 3.0 December 2017 – April 2018
Poc-version 4.0 April 2018 – October 2018
Table 3 – Overview of events
Data Collection
semi-14
structured interviews will be held with developers, managers and super-users of the BI system. Semi-structured interviews were introduced to ensure flexibility during the interviews (Aken et al., 2012). In order to increase the scope and the broadness of the development of the BI system, multiple developers, managers and superusers will be interviewed (Corbin & Strauss, 2008). The interviews were conducted in Dutch and lasted approximately 35-50 minutes for the single interviews and 90 minutes for the interview with two persons. Different interview guides have been made for either the developers, managers and the users. These interview guides are attached in appendices 1, 2, 3, and 4. The name of the firm and AI-based BI-system are anonymized. In this study they are referred as the firm and the BI system. During the interviews, questions were asked regarding the BI system about usage behavior, sensemaking, affordances and opinions of colleagues. After approval, the interview was taped and transcribed to enhance the data analysis. The transcribed interviews are anonymized to ensure confidentially. Unfortunately, the semi-structured interviews with the managers were not taped. A comprehensive summary has been made and send to those respondents. The written summaries have been used for data analysis accordingly. In total 3 superusers, 3 developers and 2 managers were interviewed (Appendix 5) and they all participated voluntary.
Second, the organization gave a tour through their headquarters to give insights about their culture and working environment. They also provided an official demonstration of the BI system. Hereby there has been a good understanding about the features within the system. Information that is gathered due to firm observations and system demonstrations were used for data analysis.
Third, the organization allowed the researcher to access relevant documentations. The documents consisted of the project timeline, the data of feature implementations, print screens of old versions, temporal business model, the outcome of the preparation phase and presentations about the BI system. The documents were analyzed to fully understand the BI system and to understand the timeline of several events.
15
world. To assess construct validity on the pre-established construct affordances, the interview guide was based on the interview guideline of Glowalla et al. (2014) and the suggested questions of Volkoff and Strong (2018). The understanding in what users think when using the technology, what they do with it and how the technology affects them will be increased. By combining several data collection methods, which is referred to as triangulation, external validity will be assured, and a better view of the phenomenon was guaranteed (Eisenhardt, 1989).
Data Analysis
16
has been compared with the factors in the existing model of Anderson and Robey (2017) in order to create the new conceptual model. The three data sources were able to empirical corroborate the composed conceptual model.
Interview text Open and axial codes Selective coding “We are known for looking
ahead, for being occupied with start-ups and new technologies. We invest a lot in these start-ups. This tool completely fits this idea. What we are trying to show with this is that old tools such as Kluwer and NDFR are redundant/dusty and have done the same thing for years.”
The firm is a pioneer in their field, firm is interested in new
technologies, fit between tool and strategy, old system is not good enough. New service, space of possibility
Comparing this passage with other statements about new service and space of
possibility the innovation mechanisms emerged. Other passages mentioned the assembly of components, which is also link within the innovation mechanism.
“They added the case numbers in the tool. It said C100/13 or something. We memorized the names of the cases instead of the number. That is easier. Now they added the names of the cases to the tool. This was only a small change, but you have to notice it. This makes it
substantially more user friendly.”
Lack of content knowledge of developers, feedback for developers, system gets more user friendly Content knowledge
The statement provides a better understanding what kind of feedback is given by superusers.
17 Findings
This section will present the findings of the study. In order to explore the underlying mechanisms, the structures and events need to be identified (Wynn & Williams, 2012). First, the structures will be provided. Second, the events related to the development of the BI system with AI-based solutions will be presented. Third, the description of the events related to the actualization of affordances will be provided.
Structures
The interviews and observations showed that the key entities of this case during data analysis were the super users, business manager, the developers, development team, the partner, the BI system, University of Tilburg, University of Amsterdam, the IT department in Amsterdam and the Tax department in Amsterdam. These key entities are interrelated in various ways. First, the super users, developers, and the business manager work together in the Tax department in Amsterdam, jointly they form the development team of the BI system. Second, the superusers collaborate with the business manager and developers in order to improve data and system quality of the BI system by providing feedback. Third, the developers are depending on the business manager, which need to convince the partner for the obtained required resources. Fourth, the universities and the development team collaborate by improving the validity of the BI system. Fifth, the IT department in Amsterdam maintains the IT infrastructure of the BI system.
Additionally, there were also important contextual factors that contributed to the occurrence of the development of the BI system. Currently, the firm is using an outdated software application that is not significantly improved for several years now. Furthermore, the firm need to pay a fee for every user of that application. By owning an even more comprehensive BI system, which has the functionalities of the legacy IT system and extra functionalities, costs can be reduced. So, during the beginning of 2017 there was a gap in the market of European Tax Law and artificial intelligence. In that period of time, rapid action was essential due to fierce competition in the market:
18 Development Process
The scope of the development process has been established between January 2017 till June 2018. As a result, the most important events of this study were the preparation phase, poc-version 1.0, poc-version 2.0 and poc-version 3.0 and their overrun.
preparation phase. During the preparation phase they wanted to have a project as flexible as possible. Therefore, a business model has not been written in order to keep the possibility to change their path easily. Their aim was to develop an AI-based BI system from scratch, that has the features of the legacy IT system plus the ability to predict the judgment of the European court. That leads to the first planned affordance: prediction affordance. They expected that this affordance is potential, because they anticipated on the interaction of the expected ability of the user and the expected designed system features. They anticipated on the characteristics of the employees in this firm; whom uses a comparable IS, is commonly adaptive to change, is eager to learn and the fact that the average age in this firm is in the early thirties. Back then, they needed to devise what the design affords and consider on which level it corresponds with the actors’ characteristics. In order to guarantee further growth of the BI system the firm is also anticipating on the abilities of the future applicants. The Tax department of the firm is striving to accept applications with a beta and IT background only. Furthermore, the firm tries to affect current employees to prepare future users for the possibilities for actions with technological events:
“One or two weeks ago we had an event. Kasparov, the chess player, came by. The entire event was about technology. Each industry, such as Tax & Legal and Audit, had pitches and updates about what we are offering to the clients at this moment. This is to keep each other informed and to exchange ideas.” (superuser 6)2
poc-version 1.0. In order to realize the outcomes of the preparation phase, the business
1 Translated from Dutch transcript: Stel dat wij wel dingen automatiseren en de competitors niet laten we zeggen, dan kunnen wij
onze prijs omlagen gooien. Hopelijk kunnen we op deze manier meer klanten krijgen, en hopelijk ook meer uren.
2 Translated from Dutch transcript: Zo hadden we 1 à 2 weken geleden een evenement. Hier kwam Kasparov, de schaker, kwam
19
manager, who studied tax law, asked several developers to help him with building a demo; a proof of concept, of the BI system. Those colleagues have a rich experience in data analytics. So far, the system was an abstract idea. So, their main goal is to have a physical and tangible system in a period of six weeks. The legacy IT system affords the user to find and present judgments of the European court. However, the usability of that legacy IT system is not intuitive. In this stage, they are trying to build a system that has the functionalities of the legacy IT system, which is intuitive and is capable to predict the judgment of European court. The functionality of the prediction is called the ‘verdict generator’. Roughly, they needed to scrape data (available for free), develop the verdict generator and design an intuitive web interface.
Data is needed in order to develop the verdict generator and ensure the planned prediction affordance. To ensure data quality of the demo, all relevant data should be implemented. Therefore, the developers scraped all possible judgments of the European court and implemented that data in a database. Furthermore, the developers designed a code that could find linkages between data. Eventually, through these linkages the BI system was expected to compose a written prediction of the verdict from European court.
During the development of the demo, the developers encountered a serious problem. It was not feasible to design a code that could compose a written prediction of the verdict from European court. After all, they managed to write a code that can predict the judgment of the European court by providing the chance of approval or rejection by giving a percentage. The previously written code was not useless, because it could operate as the basis of other features within the system. Three planned affordances were expected: linkages affordance, finding cases affordance and scanning affordance.
20
“It is relaxed that it is possible to combe through those summaries, this way you know instantly that you do not have to read this one, but this one is interesting. After that I can read the entire case/judgement. This ensures that you can easily distinguish between irrelevant information and information that can help you find the answer.” (superuser 6)3
Secondly, the property of linkages and finding cases affordance can be designed by implanting the feature ‘network chart’. This feature shows the linkages between cases through their references and jurisprudence. Jurisprudence is the law that is determined by historical judgments and cases. Due to this network chart, users can easily find related cases. As a result, less popular but relevant cases can arise as well. In practice, the users did not immediately understand those functionalities. In order to increase the affordance potency of that planned affordance, system features needed to be redesigned in subsequent poc-versions. Interviewee 1 provided feedback how to increase usability:
“Think about those white dots and those links. You have to be a tax professional if it says something to you. If you would see that example, you would think “nice”, however, you do not know what is meant by that. For usability it would be good if you would categorize all the expects of taxes into certain topics.” (superuser 7)4
After implementing those features, the proof of concept was finished. At that point in time, it was basically a tool to deliver evidence that AI can be useful in Tax law. However, it was not yet possible to fully interact with the BI system.
poc-version 2.0. During the development of poc-version 2.0 superusers were involved as well. The goal of the development team was to improve certain functionalities of poc-version 1.0 and to clearly design the perceptual information that specifies the planned affordances. The
3 Translated from Dutch transcript: Heel relaxed dat je even die samenvatting kan doorspitten, dan weet je meteen deze hoef ik niet
te lezen, maar deze is wel interessant. Dan lees ik vervolgens het hele arrest eropna. Dus dan kan je heel snel onderscheid maken tussen irrelevante informatie en informatie die je gaat helpen met het antwoord.
4 Translated from Dutch transcript: Denk aan die witte bolletjes en die links. Je moet wel een btw-expert zijn, mocht het jou wat
21
developers implemented a feedback system in order to improve the usability of the system. Henceforth, the superusers could easily provide feedback when usability is lacking. The received feedback must decrease the discrepancy between the ability of the user and the implemented system features. As a result, the developers implemented small adjustments in the system in order to increase usability of those system features:
“Because of that he could collect all input and subsequently improve or adapt the tool. We had such a session, it took about three months. You can still do this every day, but then by walking by or sending an email.” (superuser 7)5
“They added the case numbers in the tool. It said C100/13 or something. We memorized the names of the cases instead of the number. That is easier. Now they added the names of the cases to the tool. This was only a small change, but you have to notice it. This makes it substantially more user friendly.” (superuser 6)6
The development team had not enough resources to improve the summary feature, but they did improve the network chart significantly due to obtained feedback. The network chart was unclear and chaotic according to the superusers. They advised that it would be very helpful to add a timeline in the network chart in order to scan quickly which year the case has been released. Furthermore, the superusers demanded insights about prediction procedure. The developers replied by adding the ‘legal analytics’ feature. From now on, the employees had more transparency on how the system came up with the prediction by receiving from the system which cases and information was used. Furthermore, the developers improved the prediction feature by adding an extra variable in the system. The prediction is thereby more accurate than before.
Up till now, data is scraped manually. The main goal of the developers was to automatize
5 Translated from Dutch transcript: Op basis daarvan kon hij alle input verzamelen, en vervolgens kan je de tool verbeteren of
aanpassen. Zo'n sessie hebben we gehad en duurde ongeveer 3 maanden. Dat kan je in principe nog steeds elke dag doen, maar dan langslopen of een mail sturen.
6 Translated from Dutch transcript: Zo hadden ze de zaak nummers opgeschreven in de tool, dan staat er C100/13 ofzo. Wij
22
as much processes as possible. Therefore, the developers created an automated labeling system that automatically scrapes and labels the data into the database. The main risk is the quality of that labeling system, which can jeopardize the validity of the BI-system. The development team aimed to redesign the demo towards a system where users can interact with:
“We wanted to make sure that everything became more stable and a real product instead of solely a proof of concept.” (developer 3)7
Therefore, a great deal of work has gone into improving the data model, infrastructure and security. Firstly, the BI-system is better when there is a lot of data available. Within the domain of Tax, it will be a continuous challenge to get enough data. It is not possible to work with big data in this domain, because the amount of cases is limited. Therefore, the developers tried to improve the data model by increasing its validity. In order to do so, the university of Tilburg and Amsterdam, and an internal team validated the automated labeling system and data linkages:
“I re-joined the group from the moment there was something tangible and when we started to validate it contents. We looked at what the tool would do and if it was correct what it did. For example, two cases were linked, but is that an appropriate link?” (superuser 7)8
In order to increase validity easily, an advanced feedback system was implemented. Afterwards, specific users were able to overwrite certain codes that were insufficient. Secondly, the scalability of the system and the speed of data retrieving was lacking. Therefore, the IT infrastructure has been improved in collaborating with the IT department. This was needed to be fixed in order to increase usability of the BI system. Otherwise, the initial planned affordances would never be situated, because superusers stopped using the system due to their slowness:
7 Translated from Dutch transcript: We wilden sowieso dat alles stabieler wordt en dat het uiteindelijk ook een echt product wordt
in plaats van alleen een proof of concept.
8 Translated from Dutch transcript: Ik kwam eigenlijk pas weer bij toen er wat stond, en dat wij inhoudelijk gingen valideren. We
23
“[..] Mostly not fast enough. Sometimes it can take thirty seconds to load. I might not have the patience for it yet. Then I take the case of which I know it is needed. However, when he would do it, it might come up with three cases I did not think of. That is sort of sad.” (superuser 6)9
Thirdly, a constraint of the BI system was its insecurity. That constraint needed to be repaired, because other companies are consulting the firm in how to ensure cyber-security. A flaw in the security of their own system will risk the firm’s reputation and jeopardize the credibility of the cyber security department:
“We have an enormous department that stops them from being hacked, or in the case that they are hacked, knows how to deal with the situation. If we are being hacked ourselves it could be an immense hit on our business.” (developer 4)10
poc-version 3.0. During the development of poc-version 3.0 the business manager
encountered some difficulties during the interaction of the superusers and the BI system. To get a prediction of the chance of success of legal proceedings, users should formulate their problem definition properly. As soon as the problem definition is not formulated accordingly to the requirements, the BI system can not provide a good prediction. As a respond, the developers designed a research assistant, which is called the ‘chatbot’. The chatbot can assist the user how to formulate what they are looking for:
“The chatbot would be developed for the clients as well. You can use the predictor just to put in free-form-text. The question is what they will fill in and if it is helpful for the predictor algorithm. The predictor algorithm compares the written text with laws and jurisprudence. When you formulated that problem definition insufficient, the predictor algorithm is not useful.” (developer 3)11
9 Translated from Dutch transcript: Dat komt hij nog niet per se goed genoeg is en vooral niet snel genoeg is. Het kan soms 30
seconden duren voordat hij wordt geladen. Ik heb daar misschien nog niet het geduld voor, dan pak ik maar gewoon de zaak waarvan ik weet dat ik die nodig heb. Want als hij het snel zou doen komt hij misschien met drie zaken waar ik niet aan gedacht zou hebben. Dus ergens is dat ook wel jammer.
10 Translated from Dutch transcript: We hebben een hele grote afdeling hoe ze niet gehackt moeten worden, of als ze gehackt zijn
24
There initial plan was to expand to other fields in Law to get a comprehensive source of
information for the consultants. Due to the development of the chatbot, data model, infrastructure and security no resources were available to implement Dutch Law cases. The development team decided to increase usability instead of the broadening the scope of the project.
Affordance actualization
As previously discussed in the theory background section, the actualization of affordances depends on the user perceptions and goals (Table 1). The collected data and the interrelations of the constructs will further be explained below.
user perception. After consulting the information from the interviews and the provided demonstrations by the firm, it became obvious that the user’s perception of the new BI system influences the actualization of affordances due to sensemaking. The firm and their employees were introduced with the possibilities of the BI-system through the marketing department. Since that they knew that the Tax department had developed a BI-system that could predict the
outcome of the European court, they edited several promotion films for both internal and external stakeholders of the firm. Within those films, the functionality to predict was mainly highlighted, which is not the most important feature:
“A lot of time is invested in the data model for the search engine and that kind of stuff. In my opinion, that is the most important thing, because people will not use the tool to predict courts in the first place.” (developer 3)12
As a result, most future users expect that this is the most important functionality of the BI system and were inclined to actualize the prediction affordance. The collected data indicated that internal communication influence future users’ perception of affordances but does not influence affordance
11 Translated from Dutch transcript: En deze chatbot zou dan zijn voor de klanten. Wat je nu namelijk met de voorspeller hebt is
dat mensen gewoon free form text in kunnen vullen. Nou is de vraag of wat ze invullen nuttig is voor de voorspelalgoritme, die vergelijkt het met de tekst in de wetten en de jurisprudentie. Als je het dan verkeerd formuleert heb je er dus niet veel aan.
12 Translated from Dutch transcript: Verder hebben we veel tijd gestoken in een datamodel voor de zoekmachine en dat soort
25
potency directly. Future users might neglect or overlook the other situated affordances due to such messaging. However, external communication of this BI system can possibly lead to higher level affordances, such as a job application affordance:
“By doing that, like this tool, the organization is able to make a sound on the market and the labor market. We are able to show that we are doing cool stuff. Hopefully, people will apply for a job.” (developer 4)13
user goal. On the other hand, affordance actualization is also conditioned by the goal of the user. The job function was influential for the actualization of affordances. The job function determines the responsibilities and tasks of the user. The range of tasks was influential in what affordances were actualized, and which affordances were more beneficial for the user. Those users at the firm whom are not working with European Law will not use the BI system at all. Furthermore, the users who need to analyze a case according to jurisprudence will rather actualize the finding cases affordance than the prediction affordance. A consultant of the firm said the following:
“Eventually, the most important thing is that you can find accurate information”. (superuser 6)14
Mechanisms
The key entities, and their relationships, offers powers that might have generated the highlighted events (Wynn & Williams, 2012). Regarding the data analysis, several mechanisms were responsible for those events. In the following paragraphs, the explanation of the interaction of the innovation, knowledge and reputation mechanisms that causes the start of the preparation phase and further development are provided. Followed by two mechanisms that are responsible for the
13 Translated from Dutch transcript: Door zoiets wel te doen, zoals Tax-I, kan je wel weer geluid maken op de markt en
arbeidsmarkt. En laten zien dat we toch wel vette dingen aan het doen zijn, en hopelijk gaan mensen hierdoor wel solliciteren.
26 actualization of affordances.
innovation mechanism. The firm is constantly trying to improve and enhance their techno-organizational context. Therefore, the firm is probing to innovate their current IT systems as well. In this case, the Tax partner encountered some space of possibilities in the field of Tax due to the outdated legacy IT system and the application of AI:
“We are known for looking ahead, for being occupied with startups and new technologies. We invest a lot in these startups. This tool completely fits this idea. What we are trying to show with this is that old tools such as Kluwer and NDFR are redundant and have done the same thing for years.” (superuser 6)15
The preparation phase with the business manager started with the awareness of the space of possibility, which resulted in an idea for a new service. The assembly of new components lead to new services and the enhancement of their techno-organizational context, therefore this is a reinforcing mechanism.
knowledge mechanism. The firm is consciously registering new technological trends, think about cryptocurrency, blockchain, AI and the future of mobility. New organizational trends could lead to future questions of their clients. Their tendency to stay up-to-date and to have experience in technological trends will lead to a fast and constructive advice for clients. Eventually, such consults will increase their experience and knowledge in new technological trends even more, which reinforces knowledge. In this case, the firm is trying to be prepared when clients have questions about AI:
“Obviously we are a big ponderous machine. We have to continue to answer questions given to us by the market. When those questions evolve, we have to evolve as well.” (developer 4)16
15 Translated from Dutch transcript: We staan er bekend om dat we onze blik voorruit hebben, dat we ons bezig houden met
startups en nieuwe technologieën. Hier investeren we veel geld in. Wat betreft pas deze tool er dus helemaal bij, juist. Wat we hiermee aangeven is dat de oude hulpmiddelen, zoals Kluwer en NDFR, stoffig zijn en al jaren hetzelfde doen.
16Translated from Dutch transcript: We zijn natuurlijk een hele grote loge machine, we moeten natuurlijk vragen uit de markt
27
As a result, the firm has sixteen applied AI cases, which will enhance their techno-organizational context. According to this firm, such knowledge is a powerful source to attract future clients or projects. This BI system, which is one of those sixteen, is more than a physical product, because this BI system enlarges their knowledge about AI:
“That is a part of the proposition of the BI system as well, to learn stuff.” (developer 5)17 reputation mechanism. The reputation mechanism enforces the innovation and knowledge mechanism. This firm is recognized in the market as the most leading firm in innovation and technology:
“Company X is specified as the number one by the markets large accountancy businesses who are leaders in technology and innovation. That is the perception of the market what we are trying to maintain and to realise. You can say something for a substantial amount of time, however, this too has an expiration date. Eventually you have to show results, that is why we are working as hard as we do. That perception of the market is just there, we try to maintain it.” (developer 4)18
The firm’s main pillar is to maintain that perception of the market. In order to that, the firm is forced to find new possibilities in the market (innovative mechanism) and keep their knowledge about technological trends up-to-date (knowledge mechanism). In doing so, the reputation mechanism will be reinforced.
indispensable mechanism. Because of the partner-model structure in the firm, an individual must contain an added value for the firm. If not, dismissal is unavoidable. In order to retain an added value for the firm individuals should possess knowledge or experiences that are hard to replace. Therefore, employees are eager to learn more about new technologies and
17Translated from Dutch transcript: Dat is ook een onderdeel van de propositie van Tax-I, om dingen te gaan leren.
18 Translated from Dutch transcript: X wordt wel in de markt gespecificeerd als nummer 1 vanuit de grote accountancy kantoren
28
innovations, and are willing to work hard. Hence, this project is a possibility to own such scarce knowledge or experience:
“From the moment that you stop adding value to the organisation the end is near. It is not a charity, then it is over. You have to keep improving yourself in order to continue adding value. It still is just a commercial business with a partner-model. Eventually, if you do not add value it means that it costs someone else money. It is really simple.” (developer 4)19
Because the developers and superusers are involved within this project from the start, they possess certain knowledge and experiences that others not have. As a result, they evolve as a key employee due to their specialization and indispensability. These knowledge and experience are valuable for future projects or poc-versions as well. Therefore, their chance to get involved with comparable activities will rise, which reinforces their indispensability. The data shows that the indispensable mechanism influences the user goals in a direction that they get encouraged to use new technologies like the BI system.
trust mechanism. Due to the first poc-version 1.0 colleagues and partners were delighted about the possibilities of AI in the field of Tax. For individuals without a background in the field of data science, those possibilities of AI in the field of Tax are difficult to understand. The colleagues and partners were delighted about those possibilities. This has been done by creating a more physical and tangible demo instead of having an abstract idea only. The distance between AI and Tax has been reduced, which eventually leads to an increased involvement. This phenomenon affects the development team itself as other stakeholders and is reinforcing. Multiple individuals supported further development of the BI system and recognized the relevance of it. Trust in the developers and the BI system itself is important for sustainability of that same BI system. Positive signals from individuals in the firm can be influential on the user’s perception in a way that user will be encouraged to use the BI system more often:
19 Translated from Dutch transcript: Op het moment dat je geen waarde meer toevoegt aan de organisatie is het snel afgelopen. Je
29
“By adding stuff intermediate, people’s attractiveness towards the tools remains. After that, they would like you to continue, which provides time and rest for developing the prediction. It is not necessary to have a perfect and usable product from the beginning. It is necessary to have a clear vision for the longer term. Eventually, that vision is to predict courts. Intermediate we need to give something to the people who are sponsoring you. Thereby, we display our progress of the new features.” (developer 4)20
algorithm mechanism. An AI-algorithm is able to improve itself when used by an actor. Therefore, when an individual uses the BI-system, future use will be more streamlined, reliable and adapted according to the desires of the users. Those changes will stimulate future use as a reinforcing mechanism:
“When he is learning what users are clicking on, like someone is searching for a word and he clicks on a certain case. The tool is able to understand that this case is important. In the future, the tool understands that this case needs to come to the foreground earlier. Those functionalities are useful to help predict what users would like to do. That is important.” (developer 3)21
Towards a conceptual model
The study found evidence for the conceptual model of Anderson and Robey (2017). Therefore, that model is used as basis for the conceptual model of this research (figure 1). This model contributes how the work environment influences the strength of the affordance potency and the actualization during the development of an AI-based BI system.
During the beginning of the development of a BI system, developers are considering the planned affordances of the intended BI system. In order to do so, the developers anticipated on the
20 Translated from Dutch transcript: Doordat je tussentijds zulke dingen erin gooit, blijven mensen de applicatie vet vinden. En
willen ze dat je er mee door gaat, dat geeft je de tijd en de rust om de voorspelling te perfectioneren. Je hoeft niet op moment één een perfect product te hebben, maar je moet wel een visie hebben waar je op langer termijn wel naartoe wilt gaan. En dat is eigenlijk het voorspellen van die rechtszaken. In tussentijd zou je de mensen toch wel wat moeten geven die jou sponsoren, om te laten zien dat er vooruitgang in zit en dat er nieuwe features in zitten.
21 Translated from Dutch transcript: Als die leert wat gebruikers aanklikken, dus als iemand op een woord zoekt, en hij klikt een
30
expected abilities of users and the expected designed system features. Eventually, the interaction between the implemented system features and the actual ability of the user will lead to
affordance potency (Anderson & Robey, 2017). During the development of a BI system the challenge is to minimize the discrepancy between the designed system features and implemented system features, and the anticipated user abilities and actual user abilities. First, in order to minimize the discrepancy between designed system features and implemented system features resources and highly skilled developers are needed. Furthermore, individuals with content knowledge have to provide feedback to increase the usability of the interface. Without that feedback the perceptual information that is provided by the BI system might not specify the intended planned affordance. Moreover, a firm must have a suitable IT infrastructure in order to run the BI system faultlessly. Otherwise, users will avoid using the BI system due to
uncomfortable interaction with the system. Furthermore, the quality of a BI system depends on the quantity and quality of data that is available. Because of the limited availability of data, the quality of the data model had to be high. This could be obtained by validating both the data linkages and data labelling. Second, the developers had to minimize the discrepancy between the anticipated user abilities and actual user abilities. This was possible by analyzing the
characteristics of current employees. However, the discrepancy could be minimized by hiring applicants with a background in IT or by providing ongoing training and workshops for
employees. The developers miscalculated the actual ability of users to properly formulate their problem definition. As a result, the functionality to predict the chance of success of litigation is not accessible. The absence of something in an IS can be necessary for a redesign to rectify this (Mingers & Standing 2017). After receiving that information in the feedback system, the
developers implemented a research assistant, which supports the user to properly formulate their problem definition. Therefore, a miscalculation of actual ability of the user can be solved by redesign system features as well. Furthermore, the AI-based algorithm modifies the implemented system features after the BI system is used. This will result in an even more valid and optimized BI system.
31
system features are not implemented correctly, adjustments to the system features can rectify this in order to strengthen affordance potency. Furthermore, the AI-based algorithm modifies the implemented system features when the BI system is used. This will result in an even more valid and optimized BI system.
32
33 Discussion
The purpose of this study was to examine how the work environment can increase affordance potency during the development of a BI system. A single case study on the development of a BI system between January 2017 and June 2018 has been analyzed by using CR.
The analysis revealed that the occurrence of the development of an AI-based BI system depends on a market gap in the field of European Tax Law, a fierce competitive market and the presence of certain mechanisms. The presence of the interaction of the innovation, knowledge and reputation mechanisms in the organization caused the development of the BI system. Those mechanisms were constantly present during the development of the BI system and resulted that the developers were constantly eager to improve the BI system. Without the presence of those mechanisms, the BI system was neither developed or improved.
During the development process, two mechanisms influenced the actualization of affordances. First, the perception of the users of the BI system are affected by the trust mechanism. The process of making sense of the BI system by users and future users has been altered. Second, the BI system was an opportunity for the employees to increase their added value, which situated in the indispensable mechanism. Such opportunity will lead to a change in the user’s goal, because users will try to synchronize their goals with the new BI system. Both mechanisms conditioned the affordance actualization of the user. The information when users fail or achieve to perceive certain affordances is important for further development of the BI system. That information is only valuable when the developers receive feedback. Therefore, an important element in the model is the concept of feedback. Due to feedback the developers can match the abilities of the individual and the features of the BI system more accurate. When designed system features are not implemented correctly, adjustments to the system features can rectify this in order to strengthen affordance potency. Furthermore, the AI-based algorithm which is implemented in the BI system is able to improve itself after an interaction with users.
34
(2017) and increase the understanding of their introduced concept ‘affordance potency’ in the field of IS. Especially, the characteristics in the work environment that conditioned the relationship between the abilities of the individual and the features of the BI system. On top of that, this study adds the concept of feedback to their model due to the self-improvement nature of the AI-based BI system and the iterative process of strengthening affordance potency.
Conclusion
35
employees to be indispensable, affects the user’s goal and consequently the interaction with the new BI system. Furthermore, developers will be made aware of the importance of an interface that is clear and easy to understand to ensure affordance potency. Increased usability can be achieved by consulting individuals with content knowledge or by receiving feedback when user fail to perceive certain affordances.
This study is not without limitations. Because of time constraints, it was not feasible to get a complete overview of the development of the last version of the BI system. As a result, data collection is not complete and can affect the data analysis. Furthermore, the actual implementation of the BI system is not analyzed as well, which could give even more insights about situated affordances. Moreover, the partner that was responsible for the assignment of the task to assess the opportunity of AI in the field of Law was not available to interview. As a result, the collected data might be incomplete. Lastly, the CR perspective can provide interesting findings, but generalizability of such studies is limited (Wynn Jr. & Williams, 2012).
Interestingly, despite several shortcomings of the BI system, individuals recognized the proposed relevance of the system. So, although the interaction with the BI system is this not yet smoothly or fully operational, affordances can already be perceived by others. This was possible due to the internal communication through several promotion films about the BI system. How does provided information influence the perception of affordances both individually or organizationally? Future research can investigate if future users neglect or overlook other situated affordances when such information is provided. Furthermore, how do affordances emerge on organizational level when a BI-system is fully operational? And how can affordance potency get strengthened after going live?
36
37 References
Ackroyd, S. (2010). Critical realism, organization theory, methodology, and the emerging science of reconfiguration. In Elements of a Philosophy of Management and Organization (pp. 47-77). Springer, Berlin, Heidelberg.
Aken, J. E., Berends, J. J., & Bij, J. D. (2012). Problem solving in organizations: a methodological handbook for business and management students. Cambridge University Press.
Anderson, C., & Robey, D. (2017). Affordance potency: Explaining the actualization of technology affordances. Information and Organization, 27(2), 100-115.
Bernhard, E., Recker, J. C., & Burton-Jones, A. (2013). Understanding the actualization of affordances: A study in the process modeling context. in Proceedings of the International Conference on Information Systems (ICIS), R. Baskerville, and M. Chau (eds.). December 15-18, Milano, Italy.
Bhaskar, R. (1975). Feyerabend and bachelard: two philosophies of science. New Left Review, (94), 31.
Bhaskar, R. (2013). A realist theory of science. Routledge.
Bloomfield, B. P., Latham, Y., & Vurdubakis, T. (2010). Bodies, technologies and action possibilities: When is an affordance?. Sociology, 44(3), 415-433.
Brown, A. D., Stacey, P., & Nandhakumar, J. (2008). Making sense of sensemaking narratives. Human relations, 61(8), 1035-1062.
Bunge, M. (2004). How does it work? The search for explanatory mechanisms. Philosophy of the social sciences, 34(2), 182-210.
Bygstad, B., Munkvold, B. E., & Volkoff, O. (2016). Identifying generative mechanisms through affordances: a framework for critical realist data analysis. Journal of Information Technology, 31(1), 83-96.
Cearley, D. W., Burke, B., Searle, S. & Walker., M. J. (2017). Top 10 Strategic Technology Trends for 2018. The Top, 10.
Cearley, D. W., & Claunch, C. (2012). The top 10 strategic technology trends for 2013. The Top, 10.
38
Danermark, B., Ekström, M., Jakobsen, L., and Karlsson, J. C. (2002). Explaining society. Critical realism in the social sciences.
Davenport, T. H. (2006). Competing on analytics. harvard business review, 84(1), 98.
Dremel, C., Herterich, M. M., Wulf, J., & Spottke, B. (2017). Actualizing Affordances: A Socio-Technical Perspective on Big Data Analytics in the Automotive Sector. ICIS 2017 Proceedings. 14.
Dremel, C., Overhage, S., Schlauderer, S., & Wulf, J. (2017). Towards a Capability Model for Big Data Analytics. in Proceedings der 13. Internationalen Tagung Wirtschaftsinformatik (WI 2017), St. Gallen, Switzerland.
Eisenhardt, K. M. (1989). Agency theory: An assessment and review. Academy of management review, 14(1), 57-74.
Elder-Vass, D. (2005). Emergence and the realist account of cause. Journal of Critical Realism, 4(2), 315-338.
Faulkner, D. R., Jackson, C. A. L., Lunn, R. J., Schlische, R. W., Shipton, Z. K., Wibberley, C. A. J., & Withjack, M. O. (2010). A review of recent developments concerning the structure, mechanics and fluid flow properties of fault zones. Journal of Structural Geology, 32(11), 1557-1575.
Fleetwood, S. (2005). Ontology in organization and management studies: A critical realist perspective. Organization, 12(2), 197-222.
Gibson, J. J. (1978). The ecological approach to the visual perception of pictures. Leonado, 11(3), 227-235
Gibson, J. J. (1977). The theory of affordances. In R. E. S. J. Bransford (Ed.), Perceiving, acting, and knowing: Toward an ecological psychology (pp. 67-82). Hillsdale, NJ: Lawrence Erlbaum Associates.
Gioia, D. A. & Chittipeddi, K. (1991). Sensemaking and sensegiving in strategic change initiation. Strategic management journal, 12(6), 433-448.
39
Henfridsson, O., & Bygstad, B. (2013). The generative mechanisms of digital infrastructure evolution. MIS quarterly, 37(3), 907-931.
Hutchby, I. (2001). Technologies, texts and affordances. Sociology, 35(2), 441-456.
Leonardi, P. M. (2011). When flexible routines meet flexible technologies: Affordance, constraint, and the imbrication of human and material agencies. MIS quarterly, 147-167.
Leonardi, P. M. (2013a). Theoretical foundations for the study of sociomateriality. Information and organization, 23(2), 59-76.
Leonardi, P.M. (2013b). When does technology use enable network change in organizations? A comparative study of feature use and shared affordances.
Li, X., Hsieh, J. P. A., & Rai, A. (2013). Motivational differences across post-acceptance information system usage behaviors: An investigation in the business intelligence systems context. Information systems research, 24(3), 659-682.
Lönnqvist, A., & Pirttimäki, V. (2006). The measurement of business intelligence. Information systems management, 23(1), 32.
Luftman, J., & Ben-Zvi, T. (2010). Key issues for IT executives 2010: judicious IT investments continue post-recession. MIS Quarterly Executive (9:4), pp. 263-273.
Malladi, S., & Krishnan, M. (2013). Determinants of usage variations of business intelligence & analytics in organizations–an empirical analysis. in Proceedings of the International Conference on Information Systems (ICIS), R. Baskerville, and M. Chau (eds.). December 15-18, Milano, Italy.
Markus, M. L., & Silver, M. S. (2008). A foundation for the study of IT effects: A new look at DeSanctis and Poole's concepts of structural features and spirit. Journal of the Association for Information systems, 9(10), 5.
McGrenere, J., and Ho, W. 2000. “Affordances: Clarifying and Evolving a Concept,” Proceedings of Graphics Interface 2000; Montreal.
Mingers, J. (2004). Realizing information systems: critical realism as an underpinning philosophy for information systems. Information and organization, 14(2), 87-103.
40
Popovič, A., Hackney, R., Coelho, P. S., & Jaklič, J. (2012). Towards business intelligence systems success: Effects of maturity and culture on analytical decision making. Decision Support Systems, 54(1), 729-739.
Pozzi, G., Pigni, F., & Vitari, C. (2014). Affordance theory in the IS discipline: A review and synthesis of the literature. Presented at the 20th Americas Conference on Information Systems, AMCIS 2014.
Ranjan, J. (2009). Business intelligence: Concepts, components, techniques and benefits. Journal of Theoretical and Applied Information Technology, 9(1), 60-70.
Shaikh, A. A., & Karjaluoto, H. (2015). Making the most of information technology & systems usage: A literature review, framework and future research agenda. Computers in Human Behavior, 49, 541-566.
Stoffregen, T. A. (2003). Affordances as properties of the animal-environment system. Ecological psychology, 15(2), 115-134.
Strauss, A., and Corbin, J. (1998). Basics of qualitative research: Procedures and techniques for developing grounded theory.
Strong, D. M., Johnson, S. A., Tulu, B., Trudel, J., Volkoff, O., Pelletier, L. R., Bar-On, I., and Garber, L. 2014. “A Theory of Organization-EHR Affordance Actualization,” Journal of the Association for Information Systems (15:2), pp. 53–85.
Trieu, V. H. (2017). Getting value from Business Intelligence systems: A review and research agenda. Decision Support Systems, 93, 111-124.
Volkoff, O., & Strong, D. M. (2013). Critical Realism and Affordances: Theorizing IT-associated Organizational Change Processes. Mis Quarterly, 37(3).
Volkoff, O., & Strong, D. M. (2018). Affordance theory and how to use it in IS research. The Routledge Companion to Management Information Systems, 514
Watson, H. J., Goodhue, D. L., & Wixom, B. H. (2002). The benefits of data warehousing: why some organizations realize exceptional payoffs. Information & Management, 39(6), 491-502.
Weick, K. E. (1993). The collapse of sensemaking in organizations: The Mann Gulch disaster. Administrative science quarterly, 628-652
Weick, K. E. (1995). Sensemaking in organizations (Vol. 3). Sage.
