Sie sind auf Seite 1von 10

Artificial Intelligence on Management Information System

A Literature Analysis of Research on


Artificial Intelligence in Management
Information System (MIS)
Completed Research

Alexandre Moreira Nascimento Maria Alexandra V. C. da Cunha


ICT-UNIFESP (CAPES) EAESP-FGV
alexandre.nascimento@sloan.mit.edu alexandra.cunha@fgv.br
Fernando de Souza Meirelles Eusebio Scornavacca
EAESP-FGV University of Baltimore
fernando.meirelles@fgv.br escornavacca@ubalt.edu
Vinicius Veloso de Melo
ICT-UNIFESP
vinicius.melo@unifesp.br
Abstract
This article presents an overview of Artificial Intelligence (AI) research applied in the Management
Information Systems field by analyzing the databases of the AIS journal basket list representing the major
international journals on MIS. An initial set of 438 published papers was identified based on our search
criteria. From those, 74 were selected because of their appropriate fit with the research questions. They
were organized into domains, type of research, data collection strategy and AI technique. We found: (1) MIS
Quarterly published 42% of the papers; (2) 50% of studies were on Information Systems applications,
followed by 15% on Knowledge Management; and (3) 7 AI techniques were presented. This paper provides
an overview of applied AI in Management, highlights the most studied related topics and techniques and
contributes to a research agenda on the subject.
Keywords
Artificial Intelligence, Information Systems, Literature Analysis.

Introduction
Artificial Intelligence (AI) is related to building software that exhibits intelligence using processes similar
to those used by humans for the same activities (Simon, 1995). The potential uses of Artificial Intelligence
in Management are numerous. For example, using an Artificial Intelligence technique known as Deep
Learning, Evermann et al. (2016) documented surprisingly good results in predicting business processes
without the use of an explicit process model (Evermann, Rehse, & Fettke, 2016). The impact of progress in
Artificial Intelligence is expected to go beyond changing the nature of work. It will potentially change
economic mechanisms and business models themselves (Loebbecke & Picot, 2015). As Loebbecke and Picot
(2015) argue, among those changes in the economic mechanisms are the centralization of production, a
growing harmonization of demand and the erosion of property rights.
However, even with the relevance of the topic for businesses and its potential impacts on MIS, the literature
in the field is fragmented, not convergent and not organized. As a result, a broad portrait of state-of-the-art
of research presented in the existing literature on MIS-AI related research is needed. The main contribution
of such a summary is to facilitate future debate on theoretical and methodological research perspectives
(Scornavacca, Barnes, & Huff, 2006).

Twenty-fourth Americas Conference on Information Systems, New Orleans, 2018 1


Artificial Intelligence on Management Information System

Background
Artificial Intelligence is not a new field of study and application. Its beginning is dated to the 1950s, and it
began as an inquiry into the nature of intelligence (Simon, 1995). Since then, this area of inquiry has made
remarkable progress and influenced our lives significantly, bringing advances in health, safety and
productivity. Its benefits in schools, homes and hospitals are growing at a rapid pace (Stone et al., 2016).
A precise definition of Artificial Intelligence is difficult to finalize. While Stanford University (2016) defines
AI as a science and a set of computational technologies inspired by, but operating differently from, the ways
that people use their nervous systems and bodies to feel, learn, reason and act, Simon (1995) defines it as a
branch of computer science that examines the properties of intelligence by synthesizing intelligence. In an
attempt to understand AI better, Sweeney (2003) reviewed 996 bibliographic reference definitions and
categorized them into 5 classes: (1) human thinking; (2) human behavior; (3) ideal thinking; (4) ideal
behavior; and, (5) animal behavior. Of the 996 references, 987 (99%) were classified as being compatible
with ideal thinking and ideal behavior (Sweeney, 2003); therefore, this is the definition class adopted in
this study.
Furthermore, the goal of AI is difficult to summarize. In fact, it has at least three main goals (Simon, 1995):
(1) the construction of computer programs capable of displaying intelligence; (2) building programs that
exhibit intelligence using processes similar to those used by humans for the same activities; and (3) the
construction of intelligent programs that could supplement or complement human intelligence.
Recently, much attention has been paid to the field due to the many possibilities of automation brought
about by its recent advances as well as its potential consequences for employment. The topic was chosen as
the main theme of the World Economic Forum in Davos in 2016 in the light of studies indicating serious
potential impacts for society. For example, Frey (2016) stated that potentially 47% of jobs in the United
States will be at risk with advances in automation brought about by Artificial Intelligence (Frey et al., 2016).
In this context, automation has changed and will continue to change the nature of the work. This has raised
concerns, as some believe that the path to human-level artificial intelligence will lead to a robot that learns
to perform all the tasks that people are paid to perform (Aleksander, 2017).
The impact of Artificial Intelligence is expected to go beyond changing the nature of work. It will potentially
cause changes in economic mechanisms and business, such as the centralization of production, a growing
harmonization of demand and the erosion of property rights (Loebbecke & Picot, 2015). On the one hand,
it will pose risks to established firms, since they often fail to embrace opportunities brought about by such
transformations or to adapt their business models to new economic mechanisms (Loebbecke & Picot, 2015).
On the other hand, it will bring about many opportunities (Barth & Meirelles, 2011; Hino, Albertin, Cunha,
& Meirelles, 2015) since the potential applications of the AI technology compendium in the Business
Management field are numerous. For example, the application of an AI technique known as Deep Learning
has resulted in state-of-the-art predictions of business processes without the use of an explicit process
model (Evermann et al., 2016). It is reasonable to believe that management research will be considerably
affected by AI advancements.

Research Methodology
Many aspects were considered in selecting the literature review method for performing the current study.
The literature review has an established tradition of use as a tool for understanding state-of-the art research
in fields related to a technology (Scornavacca et al., 2006). In addition, it helps to identify critical knowledge
gaps by highlighting the discrepancy between what is currently known, what needs to be known and what
motivates other researchers to close that gap (Webster & Watson, 2002). In addition, it helps to make sense
of existing studies and supports readers in the identification of new directions in the research field (Jones
& Gatrell, 2014). Finally, it can help to create a foundation for advancing knowledge (Webster & Watson,
2002).
The state-of-the-art review was the selected literature review approach for this study because our aim is to
investigate the current state of knowledge, underscore areas for further research, point out research
limitations and offer new perspectives on issues while remaining grounded in a comprehensive
understanding of the current literature (Booth, Sutton, & Papaioannou, 2016). As an organizational review
framework, the concept-centric approach was adopted, as suggested by Webster and Watson (2002). A

Twenty-fourth Americas Conference on Information Systems, New Orleans, 2018 2


Artificial Intelligence on Management Information System

formal systematic literature review protocol was followed to support this state-of-the-art review, following
the recommendations of Levy and Ellis (2006), Vom Brocke et al. (2009), Webster and Watson (2002) and
Mathiassen, Chiasson and Germonprez (2012) for extracting, analyzing, interpreting and reporting the
literature-based findings.
The first step was to define a search strategy, which we structured through the selection of appropriated
search terms from the journals included in the search process. This review was based on published research
papers available through December of 2017. The major terms used to perform the database search, derived
from the research questions of this study, were: "Artificial Intelligence", "Neural Network", "Cognitive
Computing", "Deep Learning", "Machine Learning", "Fuzzy Logic", “Learning Algorithm", “Genetic
Algorithm", "Evolutionary Algorithm", "Expert Systems", “Time Series Forecasting”, “Genetic
Programming” and “Symbolic Regression”. These terms include the relevant synonyms identified in
exploratory research conducted with Artificial Intelligence researchers to compound them. Boolean
Operation uses of the conjunction “OR” were applied for concatenating the synonyms. Due to the limitation
of characters in the search field, it was not possible to build only one string. In these cases, the search string
was split into substrings, or an exclusive database retrieve, per synonym, was performed. Hence, the main
string used was: ("Artificial Intelligence" OR "Neural Network" OR "Cognitive Computing" OR "Deep
Learning" OR "Machine Learning" OR "Fuzzy Logic" OR "Learning Algorithm" OR "Genetic Algorithm" OR
"Evolutionary Algorithm" OR "Expert Systems" OR “Time Series Forecasting” OR “Genetic Programming”
OR “Symbolic Regression”).
The systematic search was performed in the databases indexing the selected journals using the Association
for Information Systems’ (AIS) journal basket list (Lowry et al., 2013), consisting of the following journals:
European Journal of Information Systems, Information Systems Journal, Information Systems Research,
Journal of AIS, Journal of Information Technology, Journal of MIS, Journal of Strategic Information
Systems and MIS Quarterly.
Then, all abstracts, titles and keywords were scrutinized to check their fitness with the goals of this research.
Papers with a scope not related to this research were excluded. Further, papers that only mentioned AI
search terms but that did not have AI playing a pivotal role in their research were excluded. After careful
examination, only 74 papers considered pertinent to the topic were selected as samples for a full reading,
detailed analysis and classification. The considerable drop from 438 to 74 papers occurred for many
reasons, such as misuse of the terminology or correct use of the terminology in the context of an example
within a paper that did not actually focus on the topic.
In order to focus the study (Scornavacca et al., 2006) and support the research goal of presenting a clear
picture of the state of the art in literature in AI in MIS field, the following research questions (RQ) were
posed, based on the works of Scornavacca et al. (2004,2006): [RQ1] What was the contextual domain of
the study? (e.g. Finance, Marketing, etc); [RQ2] What was the research method (empirical, non-empirical)?
[RQ3]; What was the data collection method (primary, secondary)?; [RQ4] What was the role (tool, object)
of AI play in the paper? [RQ5]; Which AI subfields were studied or used? (e.g., General AI, Machine
Learning, Big Data, Expert Systems, etc.); [RQ6] What AI techniques were studied or used? (Artificial
Neural Network, Fuzzy Logic, Genetic Algorithm, etc); [RQ7] What were the research agendas suggested by
the papers’ authors?
These questions helped to present a clear picture of the saturation [RQ1-5], lack and the coverage of the
studies over the different domains [RQ1], the research approach [RQ2] and data collection strategy used
[RQ3], AI’s role in the study [RQ4], AI subfields used or examined [RQ5] and AI techniques used or
examined [RQ6]. Together with [RQ7], these questions helped to delineate a research agenda. Prior
literature analyses demonstrated that these questions allow researchers to synthesize research fields in the
Information Systems discipline (Scornavacca et al., 2006).
The sample papers were coded over categories 1 to 7 [CT1-6] to answer questions 1 to 6 [RQ1-6], and then
the papers were analyzed. Different strategies were used to create the codes for each of the categories. For
[CT1-Context Domain], codifications were derived by the context domain of the study according to what
was reported by their authors, as suggested by Scornavacca et al. (2004). For [CT2 Research Method] and
[CT3 Data Collection], codes were straight and similar to those suggested by Scornavacca et al. (2006). For
[CT4 Role], coding was based on the coding techniques followed by Scornavacca et al. (2006); therefore,
the code “object” was used when the paper had AI as the object of the research (i.e., when the technique was

Twenty-fourth Americas Conference on Information Systems, New Orleans, 2018 3


Artificial Intelligence on Management Information System

investigated or a component of the investigated object) and the code “tool” was used when AI was a support
tool for the method used to investigate other objects. For [CT5-Subfields] and [CT6-Techniques], similar to
[CT1], the codes were based on what was reported by their authors (Scornavacca Jr & Barnes, 2004) and,
due to the wide range of techniques, subfields and misuses of terms, the terminologies were adapted and
normalized according to field references (Brownlee, 2013; Gani, Siddiqa, Shamshirband, & Hanum, 2016;
Hall et al., 2009; Khan, Baharudin, Lee, & Khan, 2010; Russell & Norvig, 2016; Witten, Frank, Hall, & Pal,
2016).

Findings
The oldest papers (2) found, and the most recent ones (7), were published in 1984 and 2017, respectively.
No publications were found in 2002, 2009 or 2010. The largest number of papers were published in 1995
(7) and 2017 (7). The average number of papers published per year along those 34 years was 2.4, while the
average in the last 10 years was 2.6. The number of papers published per year can be seen in Figure 1.

Figure 1 – Distribution of the publications per decade (left) and per year (right)
The highest number of publications of the four decades was dated to the 1990s, followed by 2017 (Figure
1). The publications found were not homogeneously distributed over the 8 journals in the AIS’ basket list.
In fact, 60% of all the papers were published by MIS Quarterly (42%) and Information Systems Journal
(18%). Another 25% of the publications were found to be concentrated in the Journal of Management
Information Systems (14%) and JSIS (11%). Finally, only 3% of the publications were found in the Journal
of the Association for Information systems (JAIS). These data are presented in Table 1.
The publication coverage over the decades differs significantly among the journals. MIS Quarterly and
JMIS are the only journals that published articles on the topic over the 4 decades (1980s, 1990s, 2000s and
2010s). JSIS, Information Systems Journal and Information Systems Research had their first publications
in applied AI on management during the 1990s. The European Journal of Information Systems and the
Journal of Information Technology had publications during the 1990s and during the current decade
(2010s). Finally, the Journal of the Association for Information Systems started to publish such studies in
the current decade (2010s).
A considerable range of contextual domains was evident in the papers. Based on the purpose (described as
goal, aim and objective) stated in each paper, the papers were coded under the following contextual
domains: Information Systems, Finance, Human Resources, Knowledge Management, Legal, Marketing
and Operations. Instead of aiming to interpret the research goals, these classifications were developed to
reflect the primary domain of research, according to the authors (Scornavacca et al., 2006). Aiming to make
the domains more concise, the following adjustments were done: Internal Communication and
Organizational Behavior studies were coded as Human Resources; Competitive Intelligence was coded as
Marketing; Auditing was coded as Finance; Supply Chain, Production Management and Enterprise Process
Management were coded as Operations. Most of the studies (50%) were concentrated in applications to
Information Systems. In fact, studies in Information System, Knowledge Management and Finance
represent 76% of the total papers. Table 2 illustrates the distribution of the considered papers over the 7
categories.

Twenty-fourth Americas Conference on Information Systems, New Orleans, 2018 4


Artificial Intelligence on Management Information System

Papers Papers
Journal Application
# % # %
MIS Quarterly 31 42% Information Systems 37 50%
Information Systems Journal 13 18% Knowledge Management 11 15%
Journal of Management Information Systems 10 14% Finance 8 11%
Journal of Strategic Information Systems 8 11% Marketing 8 11%
Information Systems Research 7 9% Operations 6 8%
Journal of Information Technology 2 3% Human Resources 2 3%
Journal of the Association for Information Systems 2 3% Legal 2 3%
European Journal of Information Systems 1 1% Total 74 100%
Total 74 100%
Table 2 - Distribution of the
Table 1 – Distribution of the publications per journal publications per contextual domain
With consideration to their research methods, the papers are almost evenly distributed between conceptual
(32, 43%) and empirical (35, 46%) methods. In addition, 11% (8) of the papers presented both research
approaches. For their the data collection methods, 64% (47) of the papers used primary data types while
36% (26) of the papers were based on secondary data. In 61% (45) of the papers, AI was the object of the
study, while in 39% (29) of the papers, AI-related techniques were used to perform an investigation.
A considerable range of AI subfields and techniques was evident in the papers. Some of the studies reported
the investigated subfield(s), some reported the employed technique(s) and some reported both the
subfield(s) and the technique(s) in the research. Therefore, a study could have more than one subfield and
technique. Nine AI subfields (Figure 2a) were found in the sample: ES-Expert Systems (33, 37%), AI-
General Artificial Intelligence (14, 16%), KBS-Knowledge based system (12, 13%), DSS-Decision Support
System (8, 8%), BD-Big Data (7, 7,5%), DM-Data Mining (7, 7,5%), PM-Predictive Models (5, 9%), ML-
Machine Learning (2, 2%) and CIS-Collaborative Information Systems (1, 1%). Also, seven AI related
techniques were identified from the sample papers (Figure 2b): NN-Artificial Neural Networks (7, 35%),
RBM-Regression Based Models (6, 30%), GA-Genetic Algorithms (2, 10%), FL-Fuzzy Logic (2, 10%), CAS-
Complex Adaptive Systems (1, 5%), AHP - Analytic Hierarchy Process (1, 5%) and CA-Cluster Analysis (1,
5%).
An analysis was performed to understand the distribution of each identified subfield over the period of time
in which the papers were studied. This analysis denominated the “stickiness” of AI subfields over the years.
Gantt charts (Figure 2c) were built using the year of the first and the last published papers for each subfield.
The chart in Figure 2c helps to visualize which AI subfields had shorter and longer lifetimes in the IS
journals as well as when they showed up and disappeared. As can be seen, Expert Systems and Decision
Support System were the topics with the longest range, covering the whole period. They were followed by
the general term “Artificial Intelligence.” Although this might suggest the diminishment of AI subfields in
the IS field, a further analysis suggests that some of them had potentially evolved and been “re-branded”,
such as “data mining” evolving into to “big data”. This chart also reveals how recent the usage of some
subfields in IS research is, pointing out windows of opportunity for more publications.
The same analysis was done to understand the stickiness of the reported AI techniques. The chart in Figure
2d shows the AI techniques over the time period of the papers (some with longer and other with shorter
lifetimes in the IS journals), as well as when they showed up and disappeared. As can be seen, Neural
Networks research has had the longest reported lifespan, followed by Regression-Based Models. Some
techniques, such as Fuzzy Logic and Complex Adaptive Systems, had a short lifetimes in the IS journals.
The prevalence of studies using AI as an object (46 studies) compared to as a tool (28) can be seen in Figure
2e-f. An analysis according to the subfield and technique used and the role of AI in the study identified a
heterogeneous distribution, with a considerable proportion of studies investigating the subfield of Expert
System as an object (28 studies).
For future studies, the reviewed articles suggest validating and operationalizing the bigness of data
construct by considering the impact of the inclusion of variables as well as considering the impact of specific
decisions. Also, it is suggested to investigate how reliable the data outcomes from data analytics systems
are (Ghasemaghaei, Ebrahimi, & Hassanein, 2017) and to evaluate how significantly types of IT systems
may relate to business agility along with the role of Business Intelligence and communication technologies
(Carlo, Lyytinen, & Rose, 2012). Investigating the relationship between IS and social theories is
recommended by Carlo et al. (2012) and Park, El Sawy, and Fiss (2017). It has also been suggested to

Twenty-fourth Americas Conference on Information Systems, New Orleans, 2018 5


Artificial Intelligence on Management Information System

consider conducting studies using cost-sensitive regression problems with (and comparing) different cost
functions, as well as examining the cognitive and psychological aspects of using data mining methods
(Bansal, Sinha, & Zhao, 2008).

Figure 2 – AI subfields (left) and techniques (right)


Some studies suggest the following: designing and studying performance on datafication research (Markus,
2017) and studying technologies applied to data analysis while developing intelligent databases to provide
management with easy-to-use technology for model analysis (Y. Liu, Mezei, Kostakos, & Li, 2017).
Moreover, some authors recommend longitudinal studies on the use, acceptance and costs of big data
analytics systems (Loebbecke & Picot, 2015) and knowledge-based systems (Carlo et al., 2012), as well as
the identification of key emerging issues with their use (Loebbecke & Picot, 2015). Taking the knowledge
environment into consideration, future studies could delineate the role and impact of routines for how firms
leverage their knowledge base and performance (Carlo et al., 2012). Considering the macro environment, it
is suggested that researchers study how IS can integrate interactions between the administrative, political
and sustainability spheres of engagement (Corbett & Mellouli, 2017).

Discussion
Considering the wide range of potential applications of AI (Stone et al., 2016), insufficiently investigated
research domains were found. For example, no studies were found on Strategy. Also, in some areas with

Twenty-fourth Americas Conference on Information Systems, New Orleans, 2018 6


Artificial Intelligence on Management Information System

considerable potential (Stone et al., 2016) and in areas where solutions have already ben developed, such
as Customer Relationship Management and Corporative Education, for example, no studies were found.
Few studies were found in areas with considerable potential, such as Internal Communication (1),
Enterprise Process Management (1), Organizational Behavior (1) and Auditing and Supply Chain (1). A few
more studies, but still an insufficient number, were found in Production Management (2) and Law (2).
Hence, within the scope of the present research, many domains seemed to be under-represented, and it is
recommended that such applied domains be included in future research agendas.
After AI publications peaked during the 1990s, there were few publications in following decade. In fact, the
number of publication was reduced by almost half. Maybe this is because MIS researchers’ attention drifted
away from AI to the emergence of the Internet (and mobile Internet) (Scornavacca et al., 2006) and its
impact on businesses. However, the numbers of related publications in the present decade suggest that this
topic is re-capturing the attention of IS researchers. Evidence suggesting that AI research in Management
is becoming a more established field of research is supported by the even distribution of the papers’ research
methods (conceptual and empirical), suggesting a transition (Scornavacca et al., 2006) toward the growth
of real applications and deployments as the objects of research studies. The suggestion is also supported by
the fact that more deployments of AI solutions are becoming available for empirical studies using primary
data collection.
However, there are indications that AI subfields and AI techniques are being insufficiently covered by the
MIS research. In fact, the numbers of studies addressing subfields such as Big Data, Predictive Modeling
and Machine Learning are low, especially considering the current hype around and expectations of those
AI subfields (Andrews et al., 2017; Herschel, 2017; Prentice, 2017; Walker, Cearley, & Burke, 2016). Also,
the frequency and the number of studies, their “stickiness” over time and the date of the last paper found
indicate that many AI techniques, such as Genetic Algorithms, Fuzzy Logic and Cluster Analysis, are being
under covered. Further, the last study using Neural Network techniques was dated to 2006. This is
especially surprising considering the recent advancements in these techniques (Mnih et al., 2015),
particularly deep learning neural networks (Bengio & others, 2009; LeCun, Bengio, & Hinton, 2015;
Schmidhuber, 2015). These indicators of underutilization suggest that the IS research field has only been
scratching the surface of the potential outcomes and impacts of AI techniques in business management. In
other words, it appears that these subfields and techniques have not been researched and reported to the
extent of their rising importance to management. Hence, a future research agenda should include those
under covered AI subfields and techniques.
Perhaps a reason for the number of papers in which AI is examined as tool rather as an object is explained
by Shmueli and Koppius (2011). They pointed out that the slight use of predictive analytics in IS could be
explained by the very little focus given, to date, to statistical education in the predictive analytics (important
to machine learning, data mining, and related fields). In their view, this could explain the little use of it so
far in the IS field.
Our review of the articles, however, reveals encouraging results regarding AI research on decision making
(Arnold, Clark, Collier, Leech, & Sutton, 2006; Ghasemaghaei et al., 2017; Meyer et al., 2014; Park et al.,
2017; Ye & Johnson, 1995), generative capacity (Avital & Te’Eni, 2009), supply chain management (Nissen
& Sengupta, 2006), financial time series forecasting (Walczak, 2001), market predictions (Bansal et al.,
2008; Qiu & Kumar, 2017), innovation (Carlo et al., 2012), database design (Storey & Goldstein, 1993),
behavioral research (Y. Liu et al., 2017), fraud detection (Wong, Ray, Stephens, & Lewis, 2012) and helping
sustainable development goals (Corbett & Mellouli, 2017).
Models and frameworks for AI use have been proposed, for example, in predictive analytics (Balachandran
et al., 1999; Shmueli & Koppius, 2011), decision support (Clark, Jones, & Armstrong, 2007; Glezer, 2003;
Steiger, 1998), supply chain management (Malhotra, Gosain, & Sawy, 2005) and social and business
transformation (Loebbecke & Picot, 2015). Other frameworks that have been tested and discussed include
knowledge management (Fowler, 2000), innovation implementation (Osei-Bryson, Dong, & Ngwenyama,
2008) and decision support systems (Arnott, 2006). Studies also point out aspects to be considered when
adopting AI, such as the need for a high commitment to the area (King, 1984), limited empirical research
investigating big data (Günther, Rezazade Mehrizi, Huysman, & Feldberg, 2017), the generalizability of
results (Yoon, Guimaraes, & O’Neal, 1995), responsibilities when developing collaborative IS (K. Liu, Sun,
Dix, & Narasipuram, 2001), the impact of its adoption in end-user jobs (Yoon, Guimaraes, & Clevenson,

Twenty-fourth Americas Conference on Information Systems, New Orleans, 2018 7


Artificial Intelligence on Management Information System

1996) and on productivity (Ko & Osei-Bryson, 2004), human requirements to deal with the techniques,
such as monitor and override automation (Markus, 2017), and analytical skills (Shmueli & Koppius, 2011).
King (1984) observed, over 30 years ago, that AI seemed to be still valid and should be seriously considered
by the IS community. The author observed, however, that there seemed to be gap between exaggerated
claims about the topic and the work still unfulfilled. In fact, although the present research found very good
studies, the claims made by vendors and the media seems to be ahead of what can be supported by the
findings. For example, the fields in which state-of-the-art applications can excel are very narrow, such as
playing games and recognizing images (Delcker, 2018). As Noel Sharkey, emeritus professor of Artificial
Intelligence and Robotics at the University of Sheffield, recently pointed out, “the wrong idea of what
robotics can do and where AI is at the moment is very, very dangerous” (Delcker, 2018). Finally, King (1984)
advised that AI-related topics should be treated with care because they require much greater commitment
than is initially considered by most researchers. Fordyce (1985), however, presented a more optimistic view,
noting the positive results in AI related domains up until 1985. From his point of view, the reason for any
misalignment is that the history of research work on the topic is well documented while business-related
work in AI is not. As a result, according to him, most of the work on business applications is not reported
in the literature. Further, he explained, the pace of AI developments in business is set by different
mechanisms.
The misalignment between perceptions of the stage of maturity of AI related technologies seems to remain
unaddressed. As a result, there seems to be a detachment between the social and the technical dimensions
of some AI related artifacts. For example, while many headlines in the media point out that we are about to
have millions of autonomous cars deployed (Garret, 2017), practical examples suggest that these artifacts
are not ready yet. For example, an Uber self-driving car killed a pedestrian (Wakabayashi, 2018), a Tesla
auto pilot vehicle killed its driver in an accident (BBC, 2018) and Google has officially reported 272 failures
and 13 near misses for its self-driving car (Harris, 2016). To consider another example, autonomous
security robots are being deployed in public spaces (JIJI, 2017), but a security robot ran over a child in a
mall in 2016 (Wakefield, 2016).
These examples, although not related to AI applied to management, show that the use of AI requires
caution, since failures can occur. For this reason, it is very important to research the uses of AI in
management, not only in the technical respect, but also in the human and social respects, as pointed out by
Markus (2017) and Shmueli and Koppius (2011).
Maybe the most serious impact of this misalignment can be seen in the legal domain (Clarke, 2016; K. Liu
et al., 2001). For example, regulators are approving autonomous artifacts, such as driverless cars, for use in
public spaces (DMV California, n.d.) while, at the same time, legislators are discussing whether to grant
personhood to AI-related technological artifacts with autonomy (Delcker, 2018). See, for instance, the EU
Parliament Report on Civil Law Rules on Robotics (Committee on Legal Affairs, 2017). According to Noel
Sharkey, “By seeking legal personhood for robots, manufacturers were merely trying to absolve themselves
of responsibility for the actions of their machines” (Delcker, 2018). In other words, under this idea,
autonomous AI-related artifacts would be accountable and liable for their decisions, freeing the companies
that created them from liability. What could seem as absurd at first glance is, in fact, a very complex topic
with profound impacts on the social dimension of the development of AI-related technologies. Finally, when
this discussion meets the business context, the question is raised of who (“or what”) would be liable for a
catastrophic management decision performed by an algorithm. This discussion, and its intersection with
the moral, legal and ethical aspects of the use of AI, is very relevant to the MIS research agenda.

Conclusion
The purpose of this paper was to present a broad picture of the AI literature in MIS. A sample of 74 papers
from the AIS journal basket list was reviewed. Most of the studies (50%) applied AI to IS, followed by
Knowledge Management (15%) and Finance (11%). The even distribution of the papers’ research methods
suggests a transition to more mature research related to AI in management. The high number of papers
based on primary data is more evidence of this transition. There are indications, however, of research under
coverage of some AI subfields (e.g., Big Data, Predictive Modeling and Machine Learning) and AI
techniques (e.g., Genetic Algorithms, Fuzzy Logic, and recent Neural Networks such as Deep Learning).
Also, some subfields seem to have evolved and “re-branded” (such as Data Mining to Big Data). Finally, the

Twenty-fourth Americas Conference on Information Systems, New Orleans, 2018 8


Artificial Intelligence on Management Information System

number of papers in which AI is used as tool is overshadowed by those examining AI as an object. A possible
reason for this is pointed out by Shmueli and Koppius (2011): there are few predictive analytics on IS
education. The reviewed papers also suggested the need for future research, such as longitudinal studies on
AI applications in different business areas, as well as research considering different variables, their impacts
and relationships with other subjects than those already studied, and for studies that consider the macro
environment, taking administrative, political and social environments into consideration.
The review of the articles reveals encouraging results for the use of AI in decision making, generative
capacity, supply chain management, financial time series forecasting, market prediction, innovation,
database design, behavioral research, fraud detection and helping management to set sustainable
development goals. Models and frameworks for AI use are proposed, for example, in predictive analytics,
decision support and supply chain management (Malhotra, Gosain, & Sawy, 2005), covering both societal
and business transformation. Other frameworks tested and discussed include knowledge management,
innovation implementation and decision support systems.
Some research opportunities can be derived from the current research’s limitations. Only 8 journals were
considered, which limits the conclusions and the state-of-the-art research to an important but small
fraction of the IS scientific literature. This study did not propose a theoretical model or framework to
leverage future studies. Also, it did not consider conference papers, which would help to draw more exact
conclusions about the state of the research since innovative topics usually appear first at conferences.
Finally, there are many opportunities to be considered for a future research agenda. The relevant topics
identified that must be more deeply investigated include the misalignment of perceptions on the stage of
maturity of AI related technologies between IS researchers and practitioners and the potential detachment
of the social from the technical dimensions of some AI related artifacts, with its profound consequences to
society (human life, businesses, law makers, regulators and liability). The presented results suggest that
there is a wide-open space in the IS research field on the potential outcomes and impacts of AI-related
technology as applied to management. They also suggest that most of the subfields and techniques have not
been researched to their full potential, opening a window of opportunity for researchers. In addition, many
application domains are not covered by the discussed studies (such as Customer Relationship Management,
Corporative Training, Strategy, etc.), making these areas potential targets for future studies.

REFERENCES
Due to publication page limits some references are not include herein. For the full set of references contact
the authors or go to https://www.researchgate.net/publication/325126898_REFERENCES_-
_A_Literature_Analysis_of_Research_on_Artificial_Intelligence_in_Management_Information_Syste
m_MIS.
Aleksander, I. (2017). Partners of humans: a realistic assessment of the role of robots in the foreseeable
future. Journal of Information Technology, 32(1), 1–9. https://doi.org/10.1057/s41265-016-0032-4
Andrews, W., Karamouzis, F., Brant, K. F., Revang, M., Reynolds, M., & Hare, J. (2017). Predicts 2017 :
Artificial Intelligence, (November 2016).
Arnold, V., Clark, N., Collier, P. A., Leech, S. A., & Sutton, S. G. (2006). The differential use and effect of
knowledge-based system explanations in novice and expert judgment decisions. Mis Quarterly, 79–97.
Arnott, D. (2006). Cognitive biases and decision support systems development: a design science approach.
Information Systems Journal, 16(1), 55–78. https://doi.org/10.1111/j.1365-2575.2006.00208.x
Avital, M., & Te’Eni, D. (2009). From generative fit to generative capacity: exploring an emerging
dimension of information systems design and task performance. Information Systems Journal, 19(4),
345–367.
Balachandran, K., Buzydlowski, J., Dworman, G., Kimbrough, S. O., Shafer, T., & Vachula, W. J. (1999).
MOTC: An Interactive Aid for Multidimensional Hypothesis Generatio. Journal of Management
Information Systems, 16(1), 17–36.
Bansal, G., Sinha, A. P., & Zhao, H. (2008). Tuning data mining methods for cost-sensitive regression: a
study in loan charge-off forecasting. Journal of Management Information Systems, 25(3), 315–336.
Barth, N. L., & Meirelles, F. de S. (2011). Access to information: assessment of the use of automated
interaction technologies in call centers. Revista de Administração de Empresas, 51(1), 27–42.
BBC. (2018). Tesla in fatal California crash was on Autopilot. BBC News, p. March 31th. Retrieved from

Twenty-fourth Americas Conference on Information Systems, New Orleans, 2018 9


Artificial Intelligence on Management Information System

http://www.bbc.com/news/world-us-canada-43604440
Bengio, Y., & others. (2009). Learning deep architectures for AI. Foundations and Trends®in Machine
Learning, 2(1), 1–127.
Booth, A., Sutton, A., & Papaioannou, D. (2016). Systematic approaches to a successful literature review.
London, UK: Sage.
Brownlee, J. (2013). A tour of machine learning algorithms. Machine Learning Mastery.
Carlo, J. L., Lyytinen, K., & Rose, G. M. (2012). A knowledge-based model of radical innovation in small
software firms. MIS Quarterly, 36(3), 865–895.
Clark, T. D., Jones, M. C., & Armstrong, C. P. (2007). The dynamic structure of management support
systems: theory development, research focus, and direction. Mis Quarterly, 31(3), 579–615.
Clarke, R. (2016). Big data, big risks. Information Systems Journal, 26(1), 77–90.
Committee on Legal Affairs. (2017). REPORT with recommendations to the Commission on Civil Law Rules
on Robotics. Retrieved from http://www.europarl.europa.eu/sides/getDoc.do?pubRef=-
//EP//TEXT+REPORT+A8-2017-0005+0+DOC+XML+V0//EN
Corbett, J., & Mellouli, S. (2017). Winning the SDG battle in cities: how an integrated information
ecosystem can contribute to the achievement of the 2030 sustainable development goals. Information
Systems Journal.
Delcker, J. (2018, April). Europe divided over robot “personhood.” POLITICO. Politico SPRL. Retrieved
from https://www.politico.eu/article/europe-divided-over-robot-ai-artificial-intelligence-personhood/
DMV California. (n.d.). Driverless Testing of Autonomous Vehicles. Retrieved from
https://www.dmv.ca.gov/portal/dmv/detail/vr/autonomous/auto
Evermann, J., Rehse, J.-R., & Fettke, P. (2016). A deep learning approach for predicting process behaviour
at runtime. In International Conference on Business Process Management (pp. 327–338).
Fordyce, K. (1985). To the Editor: [Editor’s Comment: Decision Support Systems, Artificial Intelligence,
and Expert Systems]. MIS Quarterly, 9(1), 1–2. Retrieved from http://www.jstor.org/stable/249268
Fowler, A. (2000). The role of AI-based technology in support of the knowledge management value activity
cycle. The Journal of Strategic Information Systems, 9(2–3), 107–128.
Frey, C. B., Garlick, R., Friedlander, G., Mcdonald, G., Wilkie, M., & Lai, A. (2016). TECHNOLOGY AT
WORK v2.0.
Gani, A., Siddiqa, A., Shamshirband, S., & Hanum, F. (2016). A survey on indexing techniques for big data:
taxonomy and performance evaluation. Knowledge and Information Systems, 46(2), 241–284.
Garret, O. (2017, March). 10 Million Self-Driving Cars Will Hit The Road By 2020 -- Here’s How To Profit.
Forbes. Forbes Magazine. Retrieved from https://www.forbes.com/sites/oliviergarret/2017/03/03/10-
million-self-driving-cars-will-hit-the-road-by-2020-heres-how-to-profit/#5f741bf57e50
Ghasemaghaei, M., Ebrahimi, S., & Hassanein, K. (2017). Data analytics competency for improving firm
decision making performance. The Journal of Strategic Information Systems.
Glezer, C. (2003). A conceptual model of an interorganizational intelligent meeting-scheduler (IIMS). The
Journal of Strategic Information Systems, 12(1), 47–70.
Günther, W. A., Rezazade Mehrizi, M. H., Huysman, M., & Feldberg, F. (2017). Debating big data: A
literature review on realizing value from big data. The Journal of Strategic Information Systems, 26(3),
191–209. https://doi.org/10.1016/J.JSIS.2017.07.003
Hall, M., Frank, E., Holmes, G., Pfahringer, B., Reutemann, P., & Witten, I. H. (2009). The WEKA Data
Mining Software: An Update. SIGKDD Explor. Newsl., 11(1), 10–18.
https://doi.org/10.1145/1656274.1656278
Harris, M. (2016, January). Google reports self-driving car mistakes: 272 failures and 13 near misses. The
Guardian. Guardian News and Media. Retrieved from
https://www.theguardian.com/technology/2016/jan/12/google-self-driving-cars-mistakes-data-reports
Herschel, G. (2017). Develop Your Artificial Intelligence Strategy Expecting These Three Trends to Shape
Its Future, (April).
Hino, M. C., Albertin, L. F., Cunha, M. A., & Meirelles, F. (2015). The role of information and
communication technology in the transformation of consolidated business model: A study of taxi
cooperatives in Brazil. In CONF-IRM (p. 22).
JIJI, A. F. P. (2017, December). Robots to be “scattered” about Haneda airport to help visitors to 2020
Tokyo Olympics. The Japan Times. The Japan Times. Retrieved from
https://www.japantimes.co.jp/news/2017/12/13/national/robots-scattered-haneda-airport-help-
visitors-2020-tokyo-olympics/#.WufGTIjwbIU

Twenty-fourth Americas Conference on Information Systems, New Orleans, 2018 10

Das könnte Ihnen auch gefallen