Reliability in assessment centres depends on general and exercise performance, but not on dimensions
本研究通过大规模数据（2917名候选人）使用贝叶斯概化理论，发现评估中心的可靠性主要取决于练习表现而非维度，为测量结构提供了新视角。
Abstract This study contributes to the literature on assessment centre (AC) measurement structure by evaluating whether dimension, exercise or mixed‐model theoretical perspectives are supported by reliability outcomes. In a large‐scale study ( N candidates = 2917) utilizing Bayesian generalizability theory, we tested reliability estimates configured to conform to dimension, exercise or mixed‐model perspectives. Our findings reveal that reliability outcomes for AC ratings greatly depend on the measurement intentions of the researcher. When this intent aligned with the traditional dimension perspective, we found evidence that reliability was unacceptably low (mean reliability = .38, SD = .15). However, when the intent aligned with the exercise perspective, we found evidence that reliability exceeded acceptable criteria (mean reliability = .91, SD = .09). The addition of dimension‐ to exercise‐related effects to reflect a mixed‐model perspective did not make an appreciable difference to reliability.