自动化使用决策：在目标检测任务中控制意图和评估误差

Beck Hall R, Dzindolet Mary T, Pierce Linda G

Psychology Department, Appalachian State University, Boone, NC 28608, USA.

Hum Factors. 2007 Jun;49(3):429-37. doi: 10.1518/001872007X200076.

BACKGROUND

It was proposed that misuse and disuse often occur because operators (a) cannot determine if automation or a nonautomated alternative maximizes the likelihood of task success (appraisal errors) or (b) know the utilities of the options but disregard this information when deciding to use automation (intent errors).

OBJECTIVE

This investigation assessed the effectiveness of performance feedback, a procedure developed to attenuate appraisal errors, and scenario training, an intervention designed to decrease intent errors.

METHODS

Operators given feedback were told how many errors they and an automated device made on a target detection task. Scenario training took operators through the thought processes of optimal decision makers after the utilities of the automated and nonautomated alternatives had been determined. Following 200 training trials, participants chose between relying on their observations or an automated device.

RESULTS

There was little misuse, but disuse rates were high (84%) among operators receiving neither feedback nor scenario training. Operators paired with a more accurate machine and given feedback made approximately twice as many errors as the automated device. Nevertheless, intent errors were commonplace; 55% of the operators who received feedback without scenario training did not rely on automation. Feedback effectiveness was enhanced when used in conjunction with scenario training; the disuse rate decreased to 29%.

CONCLUSION

A combination of feedback and scenario training was more effective in mitigating disuse than either intervention used in isolation.

APPLICATION

An important application of these findings is that operator training programs should incorporate techniques to control intent and appraisal errors.

背景

有人提出，误用和不用的情况经常发生，原因在于操作人员：（a）无法确定自动化操作或非自动化操作哪一种能最大程度提高任务成功的可能性（评估错误）；或者（b）了解各种选择的效用，但在决定使用自动化操作时忽略了这些信息（意图错误）。

目的

本研究评估了绩效反馈（一种为减少评估错误而设计的程序）和情景训练（一种旨在减少意图错误的干预措施）的有效性。

方法

接受反馈的操作人员被告知他们自己以及自动化设备在目标检测任务中所犯错误的数量。情景训练让操作人员在确定了自动化操作和非自动化操作的效用之后，经历最优决策者的思维过程。在进行200次训练试验后，参与者要在依靠自身观察还是依靠自动化设备之间做出选择。

结果

几乎没有出现误用的情况，但在既没有接受反馈也没有接受情景训练的操作人员中，不用的比例很高（84%）。与更精确的机器配合并接受反馈的操作人员所犯错误数量大约是自动化设备的两倍。然而，意图错误却很常见；在没有接受情景训练只接受了反馈的操作人员中，55%的人没有依赖自动化操作。当反馈与情景训练结合使用时，其有效性得到增强；不用的比例降至29%。