Ray Jessica M, Ray Roger D
University of Central Florida, Orlando, Florida, USA.
Behav Res Methods. 2008 Aug;40(3):673-93. doi: 10.3758/brm.40.3.673.
Problems in training behavioral observers to a high degree of interindividual accuracy and intraindividual stability are fundamental concerns in descriptive research, as well as in provisions of behavioral intervention services. This article presents design characteristics of and results from three formative evaluations of an adaptive computerized expert system that shapes observation and recording skills and maximizes both individual coding accuracy and stability. The system, called Train-to-Code, allows instructors or trainers to import their own video source files and to code those videos using any appropriate descriptive behavioral-coding scheme. This generates customized expert reference data that automate subsequent training on the basis of an operant response-shaping instructional design model. Successful training relies on transitions through alternative levels of prompting and feedback designed to optimize ongoing performance until stable expert-equivalent levels of interobserver accuracy are maintained without prompting or feedback.
在将行为观察者训练到高度的个体间准确性和个体内稳定性方面存在的问题，是描述性研究以及行为干预服务提供中的基本关注点。本文介绍了一个自适应计算机专家系统的三次形成性评估的设计特点和结果，该系统塑造观察和记录技能，并使个体编码准确性和稳定性最大化。这个名为“编码训练”的系统允许教师或培训师导入他们自己的视频源文件，并使用任何适当的描述性行为编码方案对这些视频进行编码。这会生成定制的专家参考数据，这些数据基于操作性反应塑造教学设计模型自动进行后续训练。成功的训练依赖于通过不同层次的提示和反馈进行过渡，这些提示和反馈旨在优化持续表现，直到在没有提示或反馈的情况下保持观察者间准确性的稳定专家等效水平。
