Knowledge matters: Restrictiveness and performance with decision support

a r t i c l e i n f o We study 56 subjects of varying knowledge in a preferential choice task, aided by one of two Decision Support Systems (DSS) of different restrictiveness: an eliminative tool (ELIM) and a parametric search tool (PS). Using a novel measure for performance based on Data Envelopment Analysis (DEA), we find that the gains due to effort are greater with the less restrictive DSS. Surprisingly knowledge has a negative effect on performance, an effect exacerbated with the less restrictive DSS. We interpret our results in terms of knowledge-effort substitution and the nature of knowledge relative to the restrictiveness of the DSS. Decision support tools are ubiquitous. They are commonly found both as stand-alone applications and as embedded tools in larger systems. They have been applied across a wide range of domains including consumer purchases in e-commerce [32], production planning (e.g., [22,23]), revenue yield management (e.g., [20,21,57]) and corporate recovery [3]. Experimental research into the behavioral aspects of effective decision support has a long tradition, from the well-known Minnesota experiments [24] to the studies by Todd and Benbasat [7,62–65]. Yet, important gaps in the literature remain. Much of the prior research has adopted a behavioral decision theory perspective focusing on factors such as effort and accuracy [48], economic incentives [65], perceived usefulness [22,23], decisional guidance [55], naturalness [35], forms of feedback [37,51] and descriptions of the decision tool and its prior performance [22]. Performance however is a function of both individual user characteristics and technology design in the context of a specific task [19]. In comparing technology designs, a key factor is the range of decision strategies supported, which is well reflected in the relative degree of system restrictiveness (i.e., a more restrictive system supports fewer strategies). Almost axiomatically, user domain knowledge is a critical individual characteristic in exploring performance with a DSS. Despite the volume of research in decision support, rather surprisingly, user domain knowledge has not commonly been explicitly featured as a central construct (for exceptions see [5,39,61]). Only in the comparisons of expert versus novice behavior in expert systems research does user domain knowledge take the centre stage [3,41]. This is often by necessity, since many of the tasks in the expert systems context are non-normative – there is no objective correct decision even ex post – and thus the focus is on decision processes and expertise. In contrast, knowledge …