Interactive Decision Aids in E-Commerce

This introductory chapter describes the fundamentals for later analysis, modeling and discussion of choice tasks and behavior. Figure 2.1 depicts the basic elements of the choice process which are relevant for the present work. On the left hand side, we see the general problem the decision makers are faced with: the choice task. Generally speaking, a choice task defines the problem of choosing the preferred out of a discrete and finite set of alternatives. The decision makers’ preferences determine what the preferred alternative is. Thus, in Sect. 2.1, we define both choice tasks and preferences.

In this chapter, we present an empirical study which investigates the influence of context-based complexity on decision processes.1 To determine context-based complexity accurately, we measure each subject’s preferences individually with two advanced techniques from marketing research: choice-based conjoint analysis (CBC, Haaijer and Wedel 2007) and pairwise-comparison-based preference measurement (PCPM, Scholz et al. 2010), rather than relying on less precise estimates of preferences. Furthermore, we use eye tracking to trace the process of information acquisition precisely. Our results show that low context-based complexity leads to less information acquisition and more alternative-wise search. Moreover, people search information attribute-wise in the first stage of the decision process, then eliminate alternatives, and search alternative-wise in the last stage. We also found evidence that in situations of low context-based complexity, people switch earlier to alternative-wise processing. In essence, our findings suggest that people select decision strategies from an adaptive toolbox in situations of varying complexity, starting with lexicographic and elimination by aspect type rules and ending with strategies that imply an alternative-wise search.

In this chapter, we present a new approach for the design of choice task experiments that analyze the final respondent’s choice but not the decision process.1 The approach creates choice tasks with a one-to-one correspondence between decision strategies and the observed choices. Thus, a decision strategy used is unambiguously deduced from an observed choice. Furthermore, the approach systematically manipulates the characteristics of choice tasks and takes into account measurement errors concerning the preferences of the decision makers. We use this approach to generate respondent-specific choice tasks with either low or high complexity and study their influence on the use of compensatory and non-compensatory decision strategies. We provide results for the same three measurements of context-based complexity, namely the attribute range, the attractiveness difference, and the correlation of attribute vectors, which we considered in the previous study in Chap. 3. Furthermore, we study two measurements of task-based complexity, namely the number of alternatives and the number of attributes. We find that an increase in context-based complexity and number of alternatives lead to an increased use of non-compensatory strategies and a decreased use of compensatory decision strategies. In contrast, the number of attributes does not influence strategy usage. Furthermore, we observe interaction effects between the attribute range and the correlation of attribute vectors. The proposed approach does not rely on particular decision strategies or hypotheses to be tested and is immediately applicable to a wider range of decision environments. It contributes to research attempts that create designs that maximally discriminate between different models (see Sect. 2.3 and Myung and Pitt 2009).

In the previous chapter, we pointed out that although users evaluate IIMT very positively, only a limited number of IIMT are offered. We assume that the main reason for this is that online sellers do not know which IIMT they should offer and what they should exactly look like. We address these two aspects in this chapter. Firstly, we review literature on IDA in the field of human interaction. We discuss several drawbacks of current approaches as well as the resulting requirements for the design of IIMT. Secondly, we break down observed decision-making behavior into typical steps decision makers apply in their decision processes. These steps indicate which IIMT would offer appropriate decision support. Based on these findings, we implement an IIMT-prototype, called INTACMATO in an iterative approach where two qualitative usability studies and implementation phases alternate.

In this chapter, we analyze to what extent INTACMATO meets the requirement of low effort. We take an existing effort-accuracy model, adapt and extend it to INTACMATO and show analytically the savings of effort achieved.

The main purpose of the empirical study which we present in this chapter is to evaluate the IIMT-prototype (INTACMATO). INTACMATO is evaluated across the criteria: perceived ease of use (effort), perceived usefulness, shopping enjoyment, confidence, and satisfaction. The results show that – compared to a control group of more than 30 students who just saw a product-comparison matrix without any IIMTs – the web store with INTACMATO was evaluated more positively across all five evaluation criteria.

In the present work, we addressed two research questions: (1) How does the complexity of a choice task influence decision-making behavior? (2) How can we consider knowledge about decision-making behavior for the design of IIMT? By answering the first research question, we contributed to current theory on decision-making behavior, while the main contribution of addressing the second research question is the development of INTACMATO, an IIMT-prototype for supporting choice decisions.