Academic journal article Human Factors

New Alternative Methods of Analyzing Human Behavior in Cued Target Acquisition

Academic journal article Human Factors

New Alternative Methods of Analyzing Human Behavior in Cued Target Acquisition

Article excerpt

INTRODUCTION

Technology currently allows for the development of more sophisticated automated systems in which a computer can potentially make decisions previously reserved for humans. In many tasks it will be desirable for the human to remain "in the loop" to ratify or confirm the final decision. The interaction of the human participant with the automated system can be subject to "use, misuse, disuse, and abuse" (Parasuraman & Riley, 1997). Our goal is to determine the degree of disuse (failure to use a good cue) and misuse (overreliance on a faulty cue) that occur with a system designed to detect objects embedded in complex natural backgrounds. Some of the basic issues in the use of imperfect automated aids involve the dynamics in the relationship between the human and the automated system. Depending on whether the human perceives the system as a complete partner, as an adviser, or even as an irritant in the task, his or her reliance on the system and response to the system's cues are affected (see Mosier & Skitka, 1996, for a review).

We define here four possible extreme modes of behavior in response to a fallible cue:

Reliant. The observer relies exclusively on the cue rather than on his or her own abilities. In this case the accuracy el" the observer's performance would be correlated with the cue's accuracy. The observer is susceptible to two types of errors: false alarms, labeled "errors of commission," in which cued nontargets are identified as targets; and misses, or "errors of omission," in which uncued targets are not detected (Mosier & Skitka, 1996; Skitka. Mosier, & Burdick, 1999). Reliant or complacent behavior is similar to cognitive tunneling (Yeh, Wickens, & Seagull, 1999) if the observer does not take cue reliability into account.

Skeptical. The observer does not trust the cue, ignores it, and prefers to rely on his or her own normal search-and-detect ability. In this case the observer's performance level and decisions are not affected by the presence of the cuing system. Dzindolet, Pierce, Beck, and Dawe (2002) found that automated system errors often led to skepticism and disuse of the system by the operators.

Optimal. The observer takes the advice of the cue with caution, allowing it to assist his or her decision in correct identification while rejecting incorrect advice.

Confused. In this case skepticism about the correctness of the cue will lead not to the observer's replacement of the cue's advice with his or her own expertise but, rather, to confusion and an overall degradation of the human-system performance. Despite the assertion of Yeh et al. (1999) that cuing a true target always helps in its detection, the cuing system could potentially lower detection performance levels (Nicoll, 1992).

There is some evidence that observers' reliance on cues is at least partly dependent on two factors: cuing reliability and task difficulty. Studies have found that observers were more willing to use cues when the cues were more reliable (e.g., Maltz & Meyer, 2001 ; McFadden, Giesbrecht, & Gula, 1998). However, McFadden et al. found that observers failed to adjust their behavior dynamically as the cuing system's reliability changed. This would indicate that once an initial adaptation has been made, observer reliance on a cue is not a simple function of the cue's reliability--or, at least, is not very sensitive to changes in cue reliability. In addition, McFadden ct al. found that as the task became more difficult or complicated, observers followed the advice of the cue more.

Nicoll (1992) found that cuing provided advantages in the first 30 s of the visual search process. However, detection of uncued targets was substantially worse with a cuing system than without it. He also found a speed-accuracy tradeoff in that as the number of cues increased, the observers' performance speed increased but their accuracy deteriorated. …

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.