Chapter 5. The order of examination on skeletal remains and cognitive bias
6.1. Visual assessments and contextual effects
In a manner akin to previous studies regarding the effect of context on the interpretation of forensic evidence (Nakhaeizadeh, Dror et al 2014), the findings from experiments 1 and 2 highlight and extend the findings that visual assessments in forensic anthropology can change as a result of contextual information. Experiment 1 and 2 not only show that visual assessments are mediated by context, but can even change and override the previous assessments made by participants on the same skeletal remains. The literature has repeatedly shown that judgments can be affected by information that suggests the possibility of a particular outcome (e.g. Bieber 2012). Moreover, psychological research has well established that top-down processes can have a fundamental affect on visual perception and bottom-up information (e.g. Balcetis and Dunning 2006. Thus, some researchers suggest general caution with regards to any ‘information’ that is not essential for the given analysis (e.g. Dror et al. 2015). The results of the two experimental studies (experiments 1 and 2) indicate that the ‘suggestive’ contextual information provided to the participants (on previously assessed skeletal remains) caused them to re-evaluate and alter how they perceived the evidence.
The findings emphasise a key argument in favour of ‘shielding’ examiners from case information that is not pertinent for the given analysis, namely because the context changed and possibly also ‘undermined’ the independency of the examiners initial analysis of the skeletal remains. Further, the two experimental studies show a lack of consistency in the interpretation of the skeletal remains when context is given, which has previously not been considered within the domain of forensic anthropology. Although the two studies did not include any professional anthropologists as participants, the inconsistency in the decision-making of the participants when given context still offers some insight into factors that impact human performance.
The findings also highlight the need for further research within reliability issues in terms of addressing the consistency of expert performance and how this relates to potential ‘bias’ within and between experts when context is given in the assessment of skeletal remains (see Dror 2016). Together, studies 1 and 2 thus extend the findings of other studies in other domains (Dror et al. 2005; Dror et al. 2006; Dror & Hampikian 2011) with respect to the impact of context in forensic decision-making, especially when those decisions are complex and difficult. This further exposes a vulnerability to contextual influences in subjective methods used in forensic anthropology.
The results of study 1 and 2 also raise important questions as to the exact nature of the process whereby contextual information overrides previous judgments. Hence, these experiments not only show that context can mediate judgments, but that it can potentially cause people to override and contradict their previous decisions. It highlights the concerns raised by the National Academy report (2009) and The PCAST report (2016) with regards to the issues of potential contextual biases that might result from methods that are reliant upon human judgments. One might argue that participants in experimental study 1 and 2 were not ‘biased’ due to the mere fact that no absolute ‘ground truth’ about the ambiguous skeletal remains used could be established, and therefore no ‘judgment accuracy’ could be measured. Nevertheless, the studies did show that contextual information altered the way in which the skeletal remains were interpreted. For example, the results from experimental study 2, showed an overall change to the previous assessment made by participants (63%) across the decisions made, with the majority of the participants making a decision that confirmed the contextual information given. This demonstrates that exposure to
contextual information could possibly produce cognitive biases.
In forensic anthropology the environmental and contextual influences will vary depending on the legislative practice and the procedures being followed. The practice of providing forensic anthropologists with access to case information will also therefore vary depending on the aforementioned reasons. However, it is not uncommon for forensic scientists in general to have access to case reports, (e.g. pathology reports, police statements, laboratory analysis etc) that could indeed be extremely influential. Although experimental studies 1 and 2 did not include any ‘forensic case reports’, much can still be drawn from the issues that may arise when relying on the context when assessing skeletal remains.
Firstly the findings show that even relatively ‘weak’ context can influence the decision-making outcome, as the context used was not considered to be ‘overpowering’ or ‘emotionally loaded’ compared to other studies (e.g. Nakhaeizadeh et al 2014a and Dror et al 2005). Second, the majority of participants changed their own previous evaluation of the skeletal remains, (a re-evaluation so it fitted with the contextual information given), thereby showing clearly that context does not need to be ‘extreme’ in order for it to have a strong affect on visual methods. Third, this raises concerns with regards to possible sources of contextual biases that might enter in forensic anthropological casework. Many of the cases in forensic anthropology involve the identity of an unknown deceased person, where the police may have an idea of the identity of the remains by searching missing persons databases (Marquez 2015). Consideration should be taken (based on the result from experimental study 1 and 2 of this thesis) of how that information might affect the decision-making process in the establishment of the biological profile, specifically in cases where the context might be suggestive if not overpowering. Knowing what information to obtain, and what questions to ask, has been argued to save considerable time and effort in forensic anthropology (Konigsberg et al. 2009). One of the most poignant dilemma for forensic anthropologists arises when they have to decide whether (and what) information to request, in order to make a positive identification. The results from study 1 and 2 presented in this thesis however, suggest that information does alter the interpretation of skeletal remains, and therefore consideration must be made as to how to identify and counter possible sources of contextual influences. This should
particularly be considered when dealing with incomplete or distorted skeletal remains, (which is not unusual in forensic cases) where the risk of unconsciously relying on police statements and other information might be more likely, as additional information in complex cases is more sought to contribute to the decision-making and inference process.
Finally, the results also highlight the possibility of what is described by Dror (2012) as the ‘bias snowball effect’ (see also what Kassin (2012) refers to as “corroboration inflation”). The results from studies 1 and 2 demonstrated a context effect and a change to a previous interpretation of skeletal remains across sex, ancestry and age at death. As explained in chapter 2 (section 2.3.3), most visual assessments are sex dependent, meaning that if there are cognitive interpretation issues arising during the stage of sex assessment, arguably the interpretation of age at death could be subject to ‘biased’ evaluations as well. This could potentially cause a domino effect that not only causes interpretation issues during the different analysis of the skeletal remains, but may also affect the way in which evidence is presented in court. Preferably, multiple morphological features are to be used in assessing skeletonised remains, although this may not always be possible in forensic contexts. Skeletal remains could be damaged due to poor preservation, and in some cases even burned and fragmented, resulting in highly ambiguous scenarios. Forensic case scenarios become even more challenging when taking into account cognitive processes involved in complex decision-making. The results of study 1 and 2 show that one must be aware of the cognitive limitations involved in each visual assessment in order to avoid a potential biasing snowball effect where one piece of information, potentially biases another element within the identification process.