This thesis is concerned with what AI is capable of in decision-making when involved in organizational decision-making processes or embedded in offered products that per-form decisions. It is also concerned with what is lost and what is gained through its use and which risks businesses face when applying it. It adds value to previous work conducted on challenges and risks by explaining these from a business perspective focusing on the economic implications for organizations. The resulting overview on chances and risks can serve organizations interested in AI investments to augment or automate decision-making in understanding the risk situation and potentials in this field.
In the first chapter AI is introduced in a comprehensible way for non-computer scientists and its relevance for business is outlined. Subsequently, decision processes and how humans and AI tackle them are explained which provides a foundation to under-stand respective strengths and limitations of humans and AI in decision-making. The third chapter explains how AI can be applied in decision-making in businesses processes and products that perform decisions providing benchmark examples. Autonomous driving and recruiting are presented as examples for decision automation and decision augmentation respectively on the basis of which benefits and challenges will be explained. Focusing on these examples aims at making the possible associated effects of using AI in decision-making processes more tangible and understandable for business professionals.
Table of content
Table of abbreviations
Table of figures
List of tables
1 Introduction
2 Artificial Intelligence
2.1 Definition of Artificial Intelligence
2.2 Development of Artificial Intelligence
2.3 Expectations on Artificial Intelligence
2.4 Artificial Intelligence technologies
2.5 Relevance of Artificial Intelligence for business
3 Decision-making
3.1 Human decision-making
3.2 Artificial Intelligence in decision-making
4 Applications of decision-making with Artificial Intelligence in business
4.1 Autonomous driving
4.2 Recruiting
5 Chances and risks
5.1 Chances in terms of cost reduction or increase in revenue
5.1.1 Improved efficiency
5.1.2 Increased decision quality in outcomes
5.2 Risks in terms of cost increase or decrease in revenue
5.2.1 Lack of social acceptance
5.2.2 Legal issues
5.2.3 Ethical issues
5.2.4 Decreased decision quality in outcomes
5.2.5 Technological errors and limitations
5.2.6 Workforce transition
5.3 Trade-offs concerning Artificial Intelligence-powered decision-making
6 Conclusion
Bibliography
Appendix A
A.1 Adopted decision-making tasks by Artificial Intelligence and respective Artificial Intelligence technologies (Own Table)
A.2 Benchmark for Artificial Intelligence in decision-making
Table of abbreviations
Abbildung in dieser Leseprobe nicht enthalten
Table of figures
Figure 1. Possible reasons for the lack of an official definition of Artificial Intelligence
Figure 2. The phases of Artificial Intelligence development with paradigm changes
Figure 3. Gartner Hype Cycle for Artificial Intelligence
Figure 4. The core skills of Artificial Intelligence
Figure 5. The components of Artificial Intelligence
Figure 6. Simon’s decision-making process with feedback loops
Figure 7. Observe Orient Decide Act loop
Figure 8. Decision automation, augmentation, and support
Figure 9. Levels of automation of decisions based on the OODA loop
Figure 10. Recruiting and autonomous vehicles in decision-making with Artificial Intelligence
Figure 11. Operating model of autonomous vehicles
Figure 12. Chances of using AI in decision-making
Figure 13. Risks of using AI in decision-making
Figure 14. Comparative strengths of humans and Artificial Intelligence in decision-making
List of tables
Table 1. Trade-offs concerning AI-powered decision-making
1 Introduction
Artificial Intelligence (AI), which has grown exponentially and gained awareness over the last decade, has become part of humans’ everyday lives (Ertel, 2016, p.12; Joshi, 2020, p.1). In the business world a relevant application of AI is seen in augmenting and automating decision-making (Dejoux & Léon, 2018, 198f; Davenport et al., 2017, p.7)
While the rapid spread of AI suggests many expected benefits of it, its adoption needs to be from on a reasonable understanding of its current strengths and weaknesses and the chances and risks arising from these characteristics (Shrestha et al., 2019, p.67; Bolander, 2019, p.849). These do not only result from limited technological capabilities but also from ethical, legal and social issues (Campolo et al., 2017, p.4). Awareness is a first step to risk mitigation and especially important in consideration that reducing its AI’s limitations and weaknesses is estimated to be a long-term effort (Bolander, 2019, p.849).
This thesis is concerned with what AI is capable of in decision-making when involved in organizational decision-making processes or embedded in offered products that perform decisions, what is lost and what is gained through its use and which risks businesses face when applying it. It adds value to previous work conducted on challenges and risks by explaining these from a business perspective focusing on the economic implications for organizations. The resulting overview on chances and risks can serve organizations interested in AI investments to augment or automate decision-making in understanding the risk situation and potentials in this field.
In the first chapter AI is introduced in a comprehensible way for non-computer scientists and its relevance for business is outlined. Subsequently, decision processes and how humans and AI tackle them are explained which provides a foundation to understand respective strengths and limitations of humans and AI in decision-making. The third chapter explains how AI can be applied in decision-making in businesses processes and products that perform decisions providing benchmark examples. Autonomous driving and recruiting are presented as examples for decision automation and decision augmentation respectively on the basis of which benefits and challenges will be explained. Focusing on these examples aims at making the possible associated effects of using AI in decision-making processes more tangible and understandable for business professionals.
2 Artificial Intelligence
AI is a broadly discussed topic with viewpoints ranging from its potential to transform businesses and humanity to big concerns about threats for jobs and even mankind (Wirth, 2018, p.435f.; Brock & von Wangenheim, 2019, p.1; Gentsch, 2019, p.1; Ramge, 2018, p.10f.). In the business world this diversity in opinions leads to confusion about AI which becomes a “conscious unknown” meaning that humans know that they do not know enough about it (Burgess, 2018, p.1f). In fact, a survey, conducted in 2017, supports this exclamation through its findings that only 17% of 1,500 senior executives were familiar with the concept of AI and its applications at their companies (Davenport et al., 2017, p.3). In a more recent survey of 11,000 businesses, a similar phenomenon can be observed. While 72% of the respondents see AI as important, only 31% feel ready to address it (Deloitte Insights, 2018, p.5). However, to unlock AI’s potential and understand its chances and risks, it is necessary to understand the basics of AI presented in this chapter (Ramge, 2018, p.8).
2.1 Definition of Artificial Intelligence
There are countless attempts to define the term AI, which have a different focus and different facets depending on their technical and historical origin (Gentsch, 2019, p.17). More than 60 years ago, McCarthy, the father of AI, defined the AI problem along with Minsky, Shannon and Rochester as “that of making a machine behave in ways that could be called intelligent like if a human were so behaving” (McCarthy et al., 1955, p.12). The complication of this definition is that there is no uniform opinion about what intelligence means neither for humans nor for computer programs (Bolander, 2019, p.850). However, it is important to generate an understanding of intelligence as the term AI contains an explicit reference to it (High-Level Expert Group on Artificial Intelligence, 2019, p.1). To achieve this, Legg & Hutter analyzed over 70 definitions of intelligence and found that all implicitly contain the ability to interact, learn and adapt in order to a specific goal (2007, pp.1,9). For the focus of potential practical AI applications, to which decision-making belongs, it is suggested to refer to intelligence as the capability of solving hard problems which requires the above-mentioned abilities (Wang, 2008, p.365).
AI is an umbrella term comprising different concepts situated at different stages in terms of value creation (Markiewicz & Zheng, 2018, p.4; Sicular et al., 2019, p.3). It comprises a heterogeneous and broad set of tools, techniques and algorithms used in many different applications and technologies which has given rise to a variety of subfields (Jarrahi, 2018, p.1; Elliot et al., 2019, p.1). But AI is not only a universal field regarding its high number of methods and technologies, but also because it is a multidisciplinary field (Russell & Norvig, 2016, pp.1, 5). Philosophy, mathematics, economics, neuroscience, psychology, computer engineering, cybernetic and linguistics are some of the disciplines that contributed ideas, perspectives and methods to AI (Russell & Norvig, 2016, pp.5-17).
The meaning of AI has always adapted to the continously evolving technical possibilities (Bitkom & DFKI, 2017, p.28; Elliot et al., 2019, p.1). E. Rich emphasizes this dynamic by defining AI as “(…)study of how to make computers do things at which, at the moment, people are better.“ (E. Rich, 1985, p.117 quoted from E. Rich, 1983). Her definition illustrates AI as a man-machine competition across time and therefore AI’s performance is always seen in comparison to humans (Gentsch, 2019, p.18). Today many technologies and techniques that used to be closely related to AI such as chess computers or navigation systems have become a matter of course and are consequently not considered AI anymore (Elliot et al., 2019, p.8; Ramge, 2018, p.10). This phenomenon, characterized by a spreading opinion that the behavior of a specific AI system cannot be seen as real intelligence, is known as AI effect (Haenlein & Kaplan, 2019, p.2). As a result of AI’s dynamic nature, it is suggested that AI should be defined as “the computer science problems we have not yet solved” (Kurzweil, 1985, p.264).
AI can be classified into various approaches with different goals. Russell and Norvig distinguish between four AI approaches, human thinking, human behavior, rational behavior and rational thinking (2016, p.1f) The first two approaches measure success on how well human performance is imitated, whereas the rational approaches uses the ideal rationality as a measure of success (Russell & Norvig, 2016, p.1). The ideal rationality is a component of intelligence and refers to “the ability to choose the best action to take in order to achieve a certain goal, given certain criteria to be optimized and the available resources” (High-Level Expert Group on Artificial Intelligence, 2019, p.1).
Abbildung in dieser Leseprobe nicht enthalten
Figure 1. Possible reasons for the lack of an official definition of Artificial Intelligence (Author’s own figure)
Concluding, AI is a dynamic multi-disciplinary field and very heterogenous in terms of its research objectives, technologies and methods. These characteristics together with the different understandings of the term intelligence contribute to the lack of an official definition as shown in Figure 1. The missing definition might also be a reason for the confusion about AI that Burgess outlines (2018, p.1f). However, a further characterization of AI can be given based on three common features of the definitions in order to achieve more clarity:
1) AI is closely related to computer science (Bitkom & DFKI, 2017, p.29; Panesar, 2019, p.1).
2) A common goal of AI can be seen in the ability to handle complex problems independently (Kirste & Schürholz, 2019, p.21; Brynjolfsson & Mcafee, 2014, p.91). In so doing the overarching objective is to serve humans (Krüger & Lischka, 2018, p.23).
3) AI performs tasks that formerly could only be carried out by humans (Brynjolfsson & Mcafee, 2014, p.91).
2.2 Development of Artificial Intelligence
Regarding the development of AI, it is important to understand the differencea between symbolic and connectionist AI as they have different limitations, especially regarding explainability. According to Bitkom & DFKI, this development of AI can be divided in four phases (2017, p.29). In the first two phases a lot of knowledge was programmened manually into the AI system and the symbolic paradigm dominated (Bitkom & DFKI, 2017, p.30; Steels, 2007, p.23). This paradigm follows a top–down approach by trying to directly simulate the highest levels of human cognition, namely its symbolic reasoning which allows to do logical inference or plan a sequence of actions to achieve a certain goal (Gentsch, 2019, p.30; Bolander, 2019, p.853). Symbolic AI is not able to learn and adapt flexibly (Haenlein & Kaplan, 2019, p.5). An example for this paradigm used in intelligent decision-making support consists in expert systems, which are based on ‘if then’ statements as a set of rules (Burgess, 2018, p.11; Tweedale et al., 2008, p.4).
[...]
-
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X.