Robo Advisors. How to increase trust in Artificial Intelligence compared to traditional financial advisory

Table of Contents

List of Abbreviations

List of Figures

List of Tables

1 Introduction

2 Background Information
2.1 The History of Investment
2.2 The Development of Artificial Intelligence

3 Robo Advisors
3.1 The Rise of Robo Advisors
3.2 Functions of Robo Advisors
3.2.1 Configuration Phase
3.2.2 Matching Phase
3.2.3 Maintenance Phase
3.3 Customer Structure
3.4 Robo Advisors Compared to Traditional Financial Advisory

4 The Meaning of Trust
4.1 Influences on Trust
4.2 Trust in Financial Services
4.3 Trust in Technology
4.3.1 Trust in Automation and Artificial Intelligence
4.3.2 Algorithm Aversion and Algorithm Appreciation

5 Trust in Robo Advisors
5.1 Trust-influencing Mechanisms of Cheng et. al
5.2 Humanized Product Design
5.3 Undistrust
5.4 Building Initial Trust
5.4.1 Reputation and Trust
5.4.2 User Experience and Trust
5.5 Developing Continuous Trust

6 Conclusion

List of References

Appendix

List of Abbreviations

Abbildung in dieser Leseprobe nicht enthalten

List of Figures

Figure 1. Robo-Advisory evolution stages and distribution, adapted from Mouillet et al. (2016)

Figure 2. Comparing robo advisor penetration rate and stock market participation in the US and Germany in 2019, adapted from Fey and Di Dio (2020) and Gallup (2019)

Figure 3. Advisory Process, adapted from Jung, Dorner et al. (2018)

Figure 4. Interdisciplinary trust model, Moin et al. (2015)

Figure 5. Measuring trust in financial services, Ennew & Sekhon, 2007, p. 65

Figure 6. Dimensions of trust in technology, Siau & Wang, 2018, p. 51

Figure 7. Trust-influencing mechanism of robo advisors, Cheng et al. (2019, p. 11)

Figure 8. Trust Cycle, own illustration

List of Tables

Table 1. Four approaches of AI according to Day et al. (2018)

Table 2. Artificial Intelligence Use Cases

Table 3. Customer Types, adapted from Salo (2017)

Table 4. Differences between trust in humans and trust in machines, adapted from McKnight et al. (2011)

Table 5. Initial and continuous trust factors, Siau and Wang (2018, p. 52)

Table 6. Reasons for algorithm aversion, adapted from Dietvorst et al. (2015) and Prahl and van Swol (2017)

Table 7. Results of hypothesis testing, Cheng et al. (2019, pp. 10f)

Table 8. Distinctions of Trust, Distrust, Untrust, Undistrust, Mistrust, and Misdistrust, Cho et al. (2015, p. 28:12)

Table 9. Differences between Robo Advisors and traditional financial advisory, adapted from Lam (2016)

Table 10. Multidisciplinary meanings of trust, Cho et al. (2015, p. 28:3)

Table 11. Trust factors influenced by Trustee, Szeli (2020b, pp. 47f)

Table 12. Trust factors influenced by Trustor, Szeli, (2020b, pp.48f)

Table 13. Trust factors influenced environment, Szeli (2020b, p. 49)

1 Introduction

The decision of how much money to invest, where to invest in, and when to buy and sell has been around for as long as the opportunity to invest itself. To answer these questions, two op­tions are available: either doing the research and coming to conclusions individually, or to seek out advice from an investment advisor.

The financial advice itself can take on many forms, either through a traditional human advisor via oral or written communication, or due to technological developments, through Artificial Intelligence and algorithms, like a robo advisor. Regardless of the option chosen, financial ad­visors have to gather data and find answers within two categories to facilitate the successful investment: one, gaining insights and understanding the client’s needs; and two, combine fit­ting assets into a portfolio (Ghosh & Mahanti, 2014; Jung, Glaser, & Köpplin, 2019). Finan­cial advisors should operate with the maximization of client portfolio returns as the primary goal. They should adapt their strategies to follow the client’s preferences - while staying within a legal and regulatory framework (Castro & Annoni, 2016).

As the global economy has been thriving over the past years, more people found themselves within the segment of society now able to invest the increased wealth into financial assets. Furthermore, within that situation, the demand for financial advice has never been higher: global uncertainty, low and negative interest rates, and longer lifespans are just some of the reasons why people seek advice to invest money safely (Novick et al., 2016). Transparency has become crucial, and to tackle the flood of various products, seeking experienced and knowledgeable advice is the rational consequence (Hakala, 2019).

Nowadays, tech companies have entered our lives in nearly every possible area of application, from smart coffee machines to algorithmic-based music recommendations. Logically, it is not a far stretch that the financial sector will also experience disruption through technology-ori­ented startups. The so-called FinTech’s, short for financial technology, can be independent, newly found startups, or can be implemented by existing financial institutions as a comple­mentary sales channel (Mead, Pollari, Fortnum, Hughes, & Speier, 2016) and span a wide ar­ray of functions, including peer-to-peer lending and crowdfunding, cryptocurrencies and blockchain, and also, robotic investment advice (Novick et al., 2016). It is no surprise that this development will affect traditional financial advisory.

Mainly robo advisors are seen as one of the most disruptive technologies in the financial sec­tor. What used to be a people’s business and strived through human connections and relation­ships turned digital: a robo advisor can replace all functions of traditional financial advisors at a lower cost point and while being available 24/7. Based on financial theory, the offer inves­tors personalized portfolios - all through pressing buttons on a phone screen. Whilst promis­ing to streamline financial investment and to make it accessible to everybody, regardless of wealth, customer adoption compared to the global financial service market has been low.

Disruptive technologies offer a lot innovative and smart features, but customers might be hes­itant to try the solutions. People rely on the experience of others to build trust, and the little experience of early adopters might not be enough to influence trust to a large extent. Trust is an important factor for all services or technologies, but especially in unprecedent areas such as fully automated financial advice. Beldad, Jong, and Steehouder (2010) demonstrated that a lack of trust hinders the technology adoption, while Prahl and van Swol (2017) go a step fur­ther and describe product utilization as a behavioral measure of trust.

Trust will play a large role in the robo advisors’ journey to become one of the most accepted investment tools. As robo advisors are still a rather new product, scientific research is limited. This thesis analyses how trust-building factors and the benefits compared to traditional finan­cial advisory increase trust in robo advisors.

It will be limited through the focus of a company’s perspective regarding what factors could be manipulated to increase trust. Also, it will focus on trust in the context of product utiliza­tion, or customer adoption, in the sense of Prahl and van Swol (2017). Therefore, initial trust will be discussed in more detail than continuous trust.

The thesis will be based on a literature review methodology and will assess the theoretical background of trust through analyzing and comparing previously done research on the matter. Additionally, a quantitative study focusing on trust-building factors in robo advisors has been used as a basis to form conclusions regarding the increase of trust. Industry insights, journal articles and conference papers build the foundation of this thesis. They were identified through the usage of scientific search engines, but also through backward and forward refer­encing searches. This approach provided a multitude of applicable literature from the fields of artificial intelligence and trust.

After this introduction, the second chapter will provide some background information on the history of investment as well as development of artificial intelligence, as these two develop­ments were essential for the emerging of robo advisors. Chapter three will focus on robo advi­sors, their functions and costumers and will compare features with traditional financial advi­sors. Afterwards, chapter four analyzes trust in a multidisciplinary setting, and with a focus on financial services and artificial intelligence separately, before the two aspects are evaluated on their influences towards robo advisor trust in chapter five. Finally, chapter six will conclude the previous findings and will relate options for increasing trust to the comparison with tradi­tional financial advisory.

2 Background Information

The following two chapters will portray the history of investment and the development of arti­ficial intelligence, which both paved the way for the creation of robo advisory services.

2.1 The History of Investment

Humans have been investing their assets since the Mesopotamian Empire, and just like socie­ties as a whole, investment methods have changed dramatically since 3,000 BC and have adapted to the current needs of the people. Investment managers were first employed by the Ancient Greeks, being responsible for the upkeeping of the power elite’s agricultural invest­ments (Reamer & Downing, 2017). The Industrial Revolution and its economic boom have made investments, previously a privilege of the rich and famous, a topic of importance for all. As assets available overgrew the spending of most, many people started looking into ways to invest their savings (Reamer & Downing, 2017).

As stock markets gained popularity in the post-World War I epoch, the American dream with its rags to riches approach and the newly, easy access to credit paved the way for many (first­time) investors. Governmental requests of providing more market transparency and the pre­vention of fraudulent selling were never ratified; similarly, the public did not evaluate the risk of stock market abuse and asymmetric, unreliable information about securities. In the 1920s, 20 million shareholders tried to generate wealth on the stock market - but 50% of new securi­ties offered became worthless (SEC, n.d.).

The attitude towards investing changed tremendously after the October 1929 stock market crash. With the plunge of markets, confidence decreased rapidly. No winner was coming out of the crash, as both investors and banks lost significant amounts of money. Moreover, in or­der to allow the economy to recover, grow, and prosper again, measures that restore trust in the capital market had to be established (SEC, n.d.).

One goal of the Securities Exchange Act, passed by Congress in 1934, was to reestablish faith in the market. It created the Securities and Exchange Commission (SEC) intending to estab­lish rules of honest trading and provide investors and markets with reliable information. The two fundamental values - companies offering securities and brokers must be transparent and truthful about risks (SEC, n.d.) - paved the road for the 1961 ban of insider trading; the first step against unfair dealing and therefore a start towards democratization of investments (Reamer & Downing, 2017).

One of the products that tried to simplify the act of investing was the Exchange Traded Fund (ETFs) in the early 1990s. Similar to the mutual index funds created in the 1970s (Bogle, 2016), ETFs allow a passive management investment option that follows selected indexes consisting of stocks, bonds, or commodities, such as the Standard & Poor’s 500 Index (J. Chen, 2020). Passive fund management allows investors to match the risk and return of the selected index with low input by merely buying and holding assets with the underlining strat­egy that eventually, the market will yield positive returns. Thereby, passive management con­trasts active investment with its goal of capturing short-term price fluctuations and beating the market (J. Chen, 2019; H. Chen, Noronha, & Singal, 2006).

The principal underlining of ETFs remains in the phenomena that ETFs merely replicate mar­ket performance, rather than trying to outperform it. Today, ETFs are one of the most popular investment options, and most financial advisors will offer ETFs. Within the 30 years of its ex­istence, ETFs have taken over the world: in December 2019, 6,970 ETFs have a combined global 6.35 trillion USD assets under management (ETFGI, 2020).

With the motion of investment development, investment services have also evolved. Ancient Greeks were the first to use investment managers, and the automated teller machine (ATM) was the first mass-market financial technology. In the U.S., asset management emerged in the 1970s and is currently experiencing disruption through FinTechs (Gold & Kursh, 2017).

2.2 The Development of Artificial Intelligence

Human development is characterized by the never-ending pursuit to liberate people from manual and mental labor. So far, industrial revolutions have significantly contributed by cre­ating machines to take over heavily physical labor, thereby fostering social and economic growth. The possibility of creating intelligent, human-like machines capable of replacing mental labor became the long-sought-after ambition (Zhongzhi, 2011).

John McCarthy (2006) used the term Artificial Intelligence (AI) for the first time in 1995 at the Dartmouth Summer Research Conference. He worked together with psychologists, mathe­maticians, computer scientists, and information theorists to study the central concept that any feature of intelligence can be replicated and simulated by a machine (Zhongzhi, 2011). Russell and Norvig (2003) analyzed different definitions of AI. They concluded that defini­tions are mainly based on two aspects: Firstly, AI should think and act like humans, and sec­ondly, AI should think and act rationally. Day, Lin, and Chen adapted this model in 2018 (see Table 1).

Table 1. Four approaches of AI according to Day et al. (2018)

Abbildung in dieser Leseprobe nicht enthalten

Alan Turing (1950) developed the Turing Test, highlighting four dimensions systems should possess to be deemed as thinking: natural language processing, knowledge representation, au­tomated reasoning, and machine learning (Day et al., 2018). According to his paper, a ma­chine can be called intelligent (meaning to think and act humanly) if a third person cannot dis­tinguish a machine from a human in an interrogation scenario. Therefore, his understanding of machine intelligence follows the imitation of human behavior.

In 2010, Nilsson defined the intelligence of AI as possessing the capabilities to “function ap­propriately and with foresight” (Nilsson, 2010, p. 13); therefore, taking the meaning of intelli­gence from merely interpreting data rationally to a next level. The broad definition allows AI to adapt to specific needs of various industries and functions. Therefore, AI applications can differ significantly within fields.

One of the most recent definitions of AI comes from Haenlein and Kaplan (2019). In their re­search, AI covers “a system’s ability to interpret external data correctly, to learn from such data, and to use those learnings to achieve specific goals and tasks through flexible adapta­tion” (Haenlein & Kaplan, 2019, p. 5). AI can be organized based on intelligence type into an­alytical (cognitive intelligence), human-inspired (emotional intelligence), and humanized (so­cial intelligence) AI. Additionally, it can be grouped depending on the maturity and evolu­tionary stage into Artificial Narrow, General, and Super Intelligence (Haenlein & Kaplan, 2019). Similarly, Borona (2016) classified AI into weak (act as if they are intelligent by fol­lowing a task) and strong (represent human minds). This further proofs that while many re­searches define AI and its concepts differently, the essential understandings remain the same.

Ever since McCarthy developed the idea of AI, experts forecasted the maturity of Artificial General Intelligence to be reached within a few years. However, even in 2020, it is projected that the goal of reaching a system indistinguishable from humans in the fields of cognitive, emotional, and social intelligence will take more than ten years (Gartner, 2019). Nevertheless, it is evident that the most significant development of AI has been accomplished in the past years, and development options are endless. AI technology spans a wide array of functions such as machine learning, natural language processing, robotics, and pattern recognition (2016; Kumar, 2018; Zhongzhi, 2011). These diverse functions make AI a valuable solution in many branches such as mathematics, linguistics, psychology, or engineering - and finance (Hakala, 2019; Nilsson, 2010).

The research on AI has mainly focused on five aspects (see Table 2). Firstly, the field of rea­soning characterizes the solving of problems through logical deduction—secondly, knowledge as in possessing knowledge about the world. Thirdly, planning, which includes setting and achieving of goals. Fourthly, communication through understanding written and spoken language and fifthly, perception, meaning the ability to deduct information from sen­sory input like images or sound (Hakala, 2019; Kelnar, 2016). The financial sector benefits already from those five dimensions, and the full potential of AI is not even exhausted.

[...]