Hartley Stephanie, Winburn Allysha Powanda, Dror Itiel E
Department of Anthropology, University of West Florida, Pensacola, Florida, USA.
SNA International, Alexandria, Virginia, USA.
J Forensic Sci. 2022 Jan;67(1):68-79. doi: 10.1111/1556-4029.14931. Epub 2021 Nov 2.
Subjective decisions make human cognitive processes more susceptible to bias and error. Specifically, research indicates that additional context biases forensic anthropologists' morphological analyses. To address whether metric analyses are also subject to bias, we conducted a pilot study in which 52 experienced osteologists measured a difficult-to-classify human femur, with or without additional contextual information. Using a metric sectioning-point sex-estimation method, participants provided a sex estimate for individual skeletal element(s) and, when given multiple elements, the combined skeletal assemblage. Control group participants (n = 24) measured only the femur. In addition to the femur, bias group participants (n = 28) either measured a female humerus and viewed a female-biasing photograph (n = 14) or measured a male humerus and viewed a male-biasing photograph (n = 14). We explored whether the experts in the different groups would differ in: (1) femoral measurements; (2) femoral sex-estimation conclusions; and (3) final sex-estimation conclusions for the skeletal assemblage. Although the femoral measurements and femoral sex estimates were comparable across groups, the overall sex estimates in the female-biased group were impacted by contextual information-differing from both the control and male-biased groups (p < 0.001). Our results demonstrate that cognitive bias can occur even in metric sex-estimation conclusions. Specifically, this occurred when the metric data and single-element sex estimates were synthesized into an overall estimate. Thus, our results suggest that metric methods are most vulnerable to bias when data are synthesized into an overall conclusion, highlighting the need for bias countermeasures and comprehensive statistical frameworks for synthesizing metric data to mitigate the effects of cognitive bias.
主观决策会使人类认知过程更容易受到偏差和错误的影响。具体而言，研究表明，额外的背景信息会影响法医人类学家的形态学分析。为了探究计量分析是否也会受到偏差影响，我们进行了一项初步研究，让52位经验丰富的骨学家对一根难以分类的人类股骨进行测量，测量时分别提供或不提供额外的背景信息。使用一种计量分段点性别估计方法，参与者针对单个骨骼元素给出性别估计，若有多个元素，则对整个骨骼组合给出性别估计。对照组参与者（n = 24）仅测量股骨。除股骨外，偏差组参与者（n = 28）要么测量一根女性肱骨并观看一张偏向女性的照片（n = 14），要么测量一根男性肱骨并观看一张偏向男性的照片（n = 14）。我们探究了不同组的专家在以下方面是否存在差异：（1）股骨测量；（2）股骨性别估计结论；（3）骨骼组合的最终性别估计结论。尽管各组之间的股骨测量和股骨性别估计相当，但偏向女性的组中的总体性别估计受到背景信息的影响，与对照组和偏向男性的组均不同（p < 0.001）。我们的结果表明，即使在计量性别估计结论中也可能出现认知偏差。具体而言，当将计量数据和单元素性别估计综合成一个总体估计时就会出现这种情况。因此，我们的结果表明，当数据综合成一个总体结论时，计量方法最容易受到偏差影响，这凸显了采取偏差应对措施以及建立综合统计框架来综合计量数据以减轻认知偏差影响的必要性。
