In this research, we investigate the impact of delegating decision making to information technology (IT) on an important human decision bias - the sunk cost effect. To address our research question, we use a unique and very rich dataset containing actual market transaction data for approximately 7,000 pay-per-bid auctions. Thus, unlike previous studies that are primarily laboratory experiments, we investigate the effects of using IT on the proneness to a decision bias in real market transactions. We identify and analyze irrational decision scenarios of auction participants. We find that participants with a higher monetary investment have an increased likelihood of violating the assumption of rationality, due to the sunk cost effect. Interestingly, after controlling for monetary investments, participants who delegate their decision making to IT and, consequently, have comparably lower behavioral investments (e.g., emotional attachment, effort, time) are less prone to the sunk cost effect. In particular, delegation to IT reduces the impact of overall investments on the sunk cost effect by approximately 50%.
"One of philosophy's oldest paradoxes is the apparent contradiction between the great triumphs and the dramatic failures of the human mind. The same organism that routinely solves inferential problems too subtle and complex for the mightiest computers often makes errors in the simplest of judgments about everyday events." (Nisbett and Ross 1980) "We're wondering what a world looks like when there are a billion of these software agents transacting business on our behalf." -Dr. Steve R. White, IBM Research (Chang 2002) During the last decade, the role of information technology (IT) has evolved from being a decision aid to being a decision making artifact. Accordingly, nowadays, IT can not only support decision makers, but also make decisions on behalf of their owners (Chang 2002, Greenwald andBoyan 2001). Examples of these technologies include options for involving automated agents for bidding in online auctions (Adomavicius et al. 2009) or for trading in financial markets (Hendershott et al. 2011). Today, these agents are available at negligible marginal cost and can effectively act on behalf of their owners. As a result, for instance, in 2009 as much as 73% of all equity trading volume in the United States was executed by electronic agents (Mackenzie 2009). Not surprisingly, a significant literature has emerged, analyzing the design of these software agents, their performance in real market situations, and their effect on market outcomes (e.g., Guo et al. 2011, Hinz et al. 2011, Stone and Greenwald 2005). However, despite the widespread usage of these agents, the understanding of how delegating decision making to IT impacts different facets of decision making, especially decision biases, is still significantly lacking.
Considering the economic importance of these decision biases (DellaVigna 2009), it is, nevertheless, critical to analyze the effects that the delegation of decision making to automated software agents has on the occurrence of decision biases.
Studies of decision biases have been featured in the literature for many decades (e.g., Camerer et al. 2004, Kahneman and Tversky 1979, Pope and Schweitzer 2011), including both laboratory and field research (overviews can be found in Camerer et al. 2004, DellaVigna 2009). One important challenge for researchers is to provide ways and means of how these biases can be alleviated or even avoided.
Researchers from the information systems discipline have already made useful contributions in this area.
Several laboratory experiments have shown that decision support systems (DSS) are an effective tool for alleviating some of these decision biases (e.g., Bhandari et al. 2008, Lim et al. 2000, Roy and Lerch 1996). However, none of these studies analyzed the role that automated software agents, which effectively replaces the decision maker for the delegated task and period, might have on the occurrence of decision biases in subsequent human decisions. In addition, there exists no evidence that the laboratory results associated with DSS and decision biases are transferable to real market situations. This handicaps academics as well as practitioners because many scholars are skeptical about the transferability of lab results to the field (Levitt andList 2008, List 2003). Consequently, we investigate whether or not IT can indeed alleviate a decision bias in real market transactions.
One frequently occurring decision bias is called the ‘sunk cost effect’. It has been defined as “a greater tendency to continue an endeavor once an investment in money, effort, or time has been made” (Arkes and Blumer 1985). The sunk cost effect typically occurs in decision situations involving a chain of decisions (e.g., software projects, investments, exploration ventures, auctions) (Kanodia et al. 1989). In many of these situations, it is now possible to delegate parts of the decision making to IT (Chang 2002, Greenwald andBoyan 2001). Therefore, both researchers and practitioners would benefit from a better understanding of the impact of delegation to IT on the sunk cost effect. In this research, we have been fortunate enough to be able to address this issue by analyzing data obtained from a real market setting. In particular, we focus on the following research question: Does the delegation of parts of the actual decision making to IT affect the proneness to the sunk cost effect in a real market situation?
We consider explicitly if the delegation of decision making to IT decreases behavioral investments and if there is a sunk cost effect for these kinds of investments. While there is anecdotal evidence that the delegation of decision making to IT can reduce behavioral investments (Bapna 2003), we do not know of any paper which empirically investigates this issue. In addition, there is no clear picture of the impact of behavioral investments on the sunk cost effect. Some experimental studies have found a positive effect of behavioral investments on the sunk cost effect (e.g., Cu
This content is AI-processed based on open access ArXiv data.
