The effects of continued use of intelligent decision aids upon auditor procedural knowledge
Student: Micheal Axelsen
Supervisor: Professor Peter Green, Dr Fiona Rohde
ABSTRACT
This research proposal builds upon the theory of technology dominance (Sutton & Arnold 1998), which has as one of its propositions that the continued use of intelligent decision aids may have the effect of deskilling auditors over time. A theoretical contribution is made through a consideration of this effect through the operation of the anchoring and adjustment heuristic (Epley & Gilovich, 2006; Kowalczyk & Wolfe, 1998; Tversky & Kahnemann, 1974) and cognitive load theory (Mascha & Smedley, 2007; Sweller, 1988). The anchoring and adjustment heuristic is a technique used by people in judgment tasks to remove cognitive burden. In making a judgment, the assessor ‘anchors’ upon the first value provided in making an estimate, and then ‘adjusts’ this estimate until a ‘reasonable’ estimate is reached. This heuristic has the effect of a systematic adjustment bias in the final estimate made. Cognitive load theory finds that an expert uses different and more efficient problem-solving strategies as a result of their past experiences in comparison to the novice. The expert draws upon their experience with past problems to develop their problem-solving strategies. Theoretically the argument is developed that the professional auditor’s ability to develop efficient problem-solving strategies is reduced as a result of their use of the anchoring and adjustment heuristics encouraged by the continued use of intelligent decision aids.
It is proposed that this integrated theory be empirically tested through a series of semi-structured interviews with audit professionals and a survey of public sector auditors designed to test the developed theoretical model. This investigation will consider the role of the continued use of intelligent decision aids and any deskilling effect such use may have upon auditor ‘know-how’, or procedural knowledge.
The contributions of this proposed research are several. Firstly, a theoretical contribution is made through extension and reconciliation of the theory of technology dominance with the anchoring and adjustment heuristic and cognitive load theory. Secondly, a practical contribution is made by extension of the testing of the theory to the field rather than experimentally. A third practical contribution is made through an empirical test of the theory of technology dominance in the context of procedural knowledge (auditor ‘know-how’), which has not previously been tested.
Anchoring & adjustment is one of three heuristics that are used to make estimates (Tversky & Kahnemann 1974) Estimates are made by starting from an initial value that is adjusted to yield the final answer; initial value may be suggested by problem’s formulation, or be the result of a partial computation The adjustments made are usually insufficient (Slovic & Lichtenstein 1971), resulting in adjustment bias Support found for anchoring in IDA use (Kowalczyk & Wolfe 1998) and audit judgment (Joyce & Biddle 1981; Presutti 1995) This theory suggests that experience will be reduced if the IDA design encourages excessive anchoring, which cognitive load suggests decreases knowledge
Overview of Cognitive Load Theory Experts and novices in a problem domain adopt differing problem solving strategies (Sweller 1988): Novices identified end goals and worked backwards from those goals through identified sub-goals using means-end analysis Experts eliminated the backward-looking phase (Sweller, 1988). Experts have acquired schemas as a result of their experience with past problems, whereas novices are forced to use generalised problem-solving strategies as they do not yet possess these schemas (Sweller, 1988). Experts and novices in a problem domain adopt differing problem solving strategies (Sweller 1988): Novices identified end goals and worked backwards from those goals through identified sub-goals using means-end analysis Experts eliminated the backward-looking phase (Sweller, 1988). Experts have acquired schemas as a result of their experience with past problems, whereas novices are forced to use generalised problem-solving strategies as they do not yet possess these schemas (Sweller, 1988).
Key Concepts Intelligent decision aid: software-intensive systems that integrate the expertise of one or more experts in a given decision domain (Arnold, Collier, Leech & Sutton 2004) Deskilling: a decline of auditors’ skills and abilities (Arnold & Sutton 1998) Declarative knowledge: definitions, examples and rules that are stored into long-term memory (Anderson 1993; McCall, Arnold & Sutton 2008) Procedural knowledge: the ability to apply and extend declarative knowledge and is acquired through experience - a key antecedent to expertise (Anderson 1993; McCall, Arnold & Sutton (2008) Reliance: the degree to which a user applies and incorporates the recommendations of the IDA during judgment formulation (Arnold & Sutton 1998; Hampton 2005) System restrictiveness: the degree to which and the manner in which a Decision Support System restricts users' decision-making processes to a particular subset of possible processes (Silver 1988) – made up of: Physical (Chu & Elam 1990; Speier & Morris 2003) Structural - embedded internally within the system (Anson et al 1995; Desanctis & Poole 1994) Process (Kim et al 2002; Wheeler & Valacich 1996) Propositions Proposition One: At low to moderate level of experience, there is a negative relationship between task experience and reliance on a decision aid Proposition Two: Positive relationship between task complexity and reliance on a decision aid Proposition Three: Positive relationship between decision aid familiarity and reliance on the decision aid Proposition Four: Positive relationship between cognitive fit and reliance on the decision aid Further, the theory, relating to the susceptibility of a user to dominance by technology, sets out two propositions: Proposition Five: When the expertise of the user and intelligent decision aid are mismatched, there is a negative relationship between the user’s expertise level and the risk of poor decision making Proposition Six: When the expertise level of the user and intelligent decision aid are matched, there is a positive relationship between reliance on the aid and improved decision making And finally, the theory of technology dominance considers the long-term effects of dominance by intelligent technologies through the following two propositions: Proposition Seven: There is a positive relationship between continued use of an intelligent decision aid and the de-skilling of knowledge workers’ abilities for the domain in which the aid is used Proposition Eight: There is a negative relationship between the broad-based, long term use of an intelligent decision aid in a given problem domain and the growth in knowledge and advancement of the domain
Background Arnold & Sutton (1998) developed this theory in response to the limited success of IDAs in the audit domain, and to see how IDAs might be more effectively designed for audits Propositions 1, 2, 3 & 4 generally hold (Arnold, Collier, Leech & Sutton 2004; Masselli, Ricketts, Arnold, & Sutton 2002; Noga & Arnold 2003) Propositions 5 & 6 have support (Arnold et al 2006; Arnold et al 2004) Proposition 7 investigated for declarative knowledge (Dowling, Leech & Moroney 2006; Mascha & Smedley 2007; McCall et al 2008) Proposition 8 not investigated Proposition One: At low to moderate level of experience, there is a negative relationship between task experience and reliance on a decision aid Proposition Two: Positive relationship between task complexity and reliance on a decision aid Proposition Three: Positive relationship between decision aid familiarity and reliance on the decision aid Proposition Four: Positive relationship between cognitive fit and reliance on the decision aid Further, the theory, relating to the susceptibility of a user to dominance by technology, sets out two propositions: Proposition Five: When the expertise of the user and intelligent decision aid are mismatched, there is a negative relationship between the user’s expertise level and the risk of poor decision making Proposition Six: When the expertise level of the user and intelligent decision aid are matched, there is a positive relationship between reliance on the aid and improved decision making And finally, the theory of technology dominance considers the long-term effects of dominance by intelligent technologies through the following two propositions: Proposition Seven: There is a positive relationship between continued use of an intelligent decision aid and the de-skilling of knowledge workers’ abilities for the domain in which the aid is used Proposition Eight: There is a negative relationship between the broad-based, long term use of an intelligent decision aid in a given problem domain and the growth in knowledge and advancement of the domain