Academic journal article Human Factors

The Reduction of Uncertainty and Troubleshooting Performance

Academic journal article Human Factors

The Reduction of Uncertainty and Troubleshooting Performance

Article excerpt

INTRODUCTION

As Rouse (1978a) reported, with increasing technological sophistication and automation of tasks and jobs in all work domains, human requirements have become less operation oriented and more supervisorial in nature. This job evolution has required workers to take on problem-solving and troubleshooting roles that were once not required of them (Rasmussen & Rouse, 1981).

Over the past 15 years, human problem solving has been examined extensively. Studies have included troubleshooting skills training (Fath, Mitchell, & Govindaraj, 1990; Su & Govindaraj, 1986), problem-solving strategies (Rasmussen, 1981), cognitive styles (Henneman & Rouse, 1984), behavioral and organizational characteristics (Sage, 1981), problem-solving models (Rouse, 1978b; Rouse & Hunt, 1984; Rouse, Rouse, & Pellegrino, 1980; Shrager, Hogg, & Huberman, 1988), and task variables affecting troubleshooting (Allen, Teague, & Carter, 1996; Morris & Rouse, 1985a).

Although much has been learned about human behavior and performance in problem-solving and decision-making situations, researchers have begun to investigate another aspect of human troubleshooting that has previously received only indirect attention: uncertainty (Allen et al., 1996). Schrenk (1969) stated that the most important aspect of problem solving is the "resolution of uncertainty" (p. 545). "Uncertainty," as described by Dougherty and Fragolla (1988), "never enhances human performance" (p. 149), and yet it is an ever-present factor in the person-environment relationship. Reduction of uncertainty has been studied in investigations concerned with the development of problem-solving skills and system knowledge (Govindaraj & Su, 1988; Hunt & Rouse, 1981; Morris & Rouse, 1985b); the use of statistical versus heuristic rules (Kahneman, Slovic, & Tversky, 1982; Trope & Ginossar, 1987), and the unknown results of decisions (Wickens, 1992).

Uncertainty in problem solving has also been addressed using algorithms (e.g., Bayesian statistics), which model problem-solving behavior based on the building and modification of a troubleshooter's subjective probability of involvement distribution (Bond, 1966; McKenzie, 1994; Mills, 1971; Shigemasu, 1976; Stolurow, Bergum, Hodgson, & Silva, 1955). This research has been conducted in situations in which a test of a component was 100% reliable and information gained from diagnostic tests could be faithfully used to modify the probability of involvement distributions. However, the effects of uncertainty on human troubleshooting performance have not been assessed in terms of intermittent faults, which are one type of uncertainty that trouble-shooters confront on a regular basis. Intermittency takes uncertainty to another level. Not only are involvement/noninvolvement questions more difficult to answer, but the believability of a particular test is reduced as well.

Intermittent Faults

Many people can identify with the consumer who has taken a product in for repairs, only to find that the symptoms disappear when the technician is present. Such unpredictable behavior is common. Sometimes faulty components may give a reliable account of their status, whereas at other times they may give false readings. This intermittency can create a number of problems for troubleshooters. The fact that a test of a component may indicate that it is not involved when in fact it is involved lengthens the time it takes to determine that the component is part of the problem. Because the true status of a component may not be revealed with a single test, the troubleshooter is forced to make a number of tests in order to obtain the information needed to reach a determination. In fact, Gorman (1986) has shown that even the prospect of unreliable components in a problem-solving situation can interfere with performance and result in increased solution times and errors. For the two experiments discussed in this paper, intermittency is defined as the rate at which a component or output, when tested, reports that it is working when it is actually not working. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.