Human Factors in Triaging Forensic Items: Casework Pressures and Ambiguity Aversion

Abstract

Triaging is an important step in deciding which items collected from crime scenes to select for forensic analysis, and so subsequent examination and findings often depend on it. This study aims to evaluate the influence of casework pressures and tolerance to ambiguity on triaging of items collected from a crime scene. A realistic pressure manipulation paradigm was developed and was found to be effective in inducing feelings of pressure in participants in an online setting. This pressure manipulation did not affect triaging decisions for both triaging experts (N = 48) and non-experts (N = 98). However, the results revealed inconsistent decisions, even among experts under identical pressure conditions and comparable background demographics. The findings also suggest that ambiguity aversion is an important factor to consider, as it can play a role in early hypotheses (e.g., reaching a decisive or inconclusive impression about a case), among other considerations. This study offers valuable insight for the development of policies for more consistent forensic triaging decisions.

1. Introduction

Pressure on forensic experts can arise from a range of different factors, including financial pressures and constraints [1], time pressure [2], stress [3,4], working on high profile cases [5], or potentially the background and status of the deceased (i.e., individual’s characteristics of education, income and occupation and also perception of their social standing; [6]). For instance, the death of an individual who is well known (such as a celebrity), could result in greater media scrutiny and coverage [7], which in turn may add pressure on the case investigators [8]. However, to date there is a lack of empirical experimental studies that have tested whether casework pressures may influence forensic decision-making. Furthermore, when requests for forensic testing are high, the capability of laboratory staffing and resources often remains limited. These pressures on a resource-limited forensic laboratory can result in backlogs and lengthy waiting times to report the findings (e.g., [9,10]). In addition, it has been suggested that this kind of pressure can be a potential factor impacting expert decision making in forensic science [see 5^th source of bias; 11].

One potential way to minimize backlogs and maximize laboratory efficiency is through triaging, or a testing strategy [12,13]. Kobus et al. [12] argue that there are two main competing demands that play a role in the triaging strategy: effectiveness (the quality of analysis) vs. efficiency (timeliness and costs, from a financial and human resource perspective). The aim is to do the most effective work in the most efficient way. However, often there is a clear trade-off between the two: increasing effectiveness reduces efficiency, and increased efficiency reduces effectiveness.

Items collected from crime scenes can be sent to a laboratory for analysis of biological traces (such as blood and semen), fingermarks, and/or other types of forensic evidence. A triaging strategy would need to be made to select which items should be tested which type of forensic analysis, and in what order the analysis should be performed [14,15]. In the context of this study, a triaging decision refers to the prioritization of items for forensic analysis, after they have been collected from the crime scene. Triaging can be a complex task, particularly if the same item could be tested for multiple forensic tests. For example, a gun could be processed for DNA, fingermarks and/or to be tested ballistically [13], and a mobile phone could be tested for digital data, geolocation and timestamp data, as well as biological traces and marks. Adding to this complexity is a lack of standardization of how triaging is carried out [13], despite being a critical decision point that can determine subsequent forensic analysis. Previous research suggested that potential disagreements in forensic decision-making from task difficulty is not necessarily easy to differentiate from unstandardized methods [16], and also individual-level differences by the evaluators themselves can play a role (e.g., their training and tolerance to uncertainty; see three factors that could affect forensic decision-making in Figure 1; [17]).

Figure 1:

An example of a monetary scenario to measure aversion to ambiguity. In this scenario, the participant has two options: either to choose an option with a 50% chance of winning $400, or another option of unknown chance of winning a higher amount of $500 (where 50% of the probability information was hidden).

Complexity and uncertainty in triaging may influence decision makers who are highly averse to ambiguity — those who dislike events of unknown probabilities [18]. Aversion to ambiguity is a type of bias that has been studied in professional domains, such as healthcare [19–21], military [22] and legal [23] domains. For example, a study identified that physicians who did not tolerate ambiguity did not escalate treatment when it was needed, thus potentially risking patient health, perhaps because the consequences of treatment escalations are typically less known in comparison to continuing the same treatment [20].

Ambiguity might be introduced or enhanced when there is conflicting information, a lack of information, unreliable information, a lack of confidence and other sources of uncertainty [24,25]—all of which are relevant in forensic decision-making. Some recent studies have started to call for more research efforts to understand uncertainty in forensic science [26] alongside experimental studies designed to address the impact of uncertainty (e.g., testing risk-aversion of fingerprint experts compared with non-experts; [27]). Yet, there is still a lack of experimental research on the role of ambiguity aversion on forensic-science decision making.

Experimental research on expert decision making has investigated judgments made at the crime scene (e.g., [28,29]) or in the laboratory (e.g., [30,31]). However, there is a lack of research addressing triaging of forensic traces, after they were collected from crime scenes. Hence, the exploratory research reported here aims to explore the potential influence of pressuring contexts and ambiguity aversion on the triaging of crime scene items. Two behavioral experiments were carried out (Experiment 1 with experts on triaging decisions, and Experiment 2 with novices) to address two main research questions:

1. Do casework pressures influence decisions pertaining to triaging forensic items?

2. Is ambiguity aversion associated with decisions pertaining to triaging forensic items?

Given the complexity and variability of approaches in triaging tasks, a secondary aim of this study was to capture whether triaging decisions for the same forensic casework could vary across practitioners of similar experience and organizational contexts. This could provide initial, preliminary insight on the consistency, or variability, between expert triaging decisions [32] ( see also Between-Expert Reliability [33]), and thus may open up the conversation about what more standardized and evidence-based triaging methods could contribute.

In addition, in this study we included novice data and compared them with experts as this may offer insight into how expertise can play a role in triaging. For instance, previously published research suggested that experts tend to select more relevant cues and make fewer hypotheses [34]. In addition, this comparison might offer insights into whether experts who are more experienced with casework pressures are less influenced by it compared with novices.

2. Method

2.1. Participants

2.1.1. Experiment 1: Triaging Experts

The selection criteria for expert participants were as follows: “(1) an adult forensic examiner who can be involved in the process of prioritizing or triaging items collected from crime scenes; and (2) can be involved in the selection of testing type for triaged items, including testing for biological traces (like blood) and fingermarks. The participant can be working in any relevant sections/ departments (such as crime scene, evidence recovery, or biology) as long as they can be involved in the above two forensic tasks.” Hence, expertise here refers to the triaging tasks, rather than the current role. Participants were also told that fluency in English was necessary for the completion of the tasks.

In total, 51 triaging experts completed the study. Three participants were excluded from the analysis: two reported as being academics with no experience in forensic triaging, and one appeared to be a random response (e.g., failed 2 of 2 embedded attention check questions, spent a few seconds in reading detailed task instructions, kept the study open for more than 20 hours). In the end, 48 practitioners were included, of whom n = 22, 45.8% were males and n = 22, 45.8% were females (did not report, n = 4, 8.3%). The mean age was 42.4 years (SD = 11.3; one participant mistakenly reported their age as 7 years, so their age was not included), and the mean years of experience in triaging crime scene items for forensic testing was 12.4 (SD = 12.3). Participating triaging experts were randomly allocated by Qualtrics to the low pressure condition (n = 27, 56.3%) and high pressure condition (n = 21, 43.8%).

The majority of participants were crime scene examiners (n = 34, 70.8%), some of whom reported taking other roles in addition to being a crime scene examiner, such as also working as forensic biologists and/or fingerprint examiners (n = 8, 16.7%). The remaining participants had other roles, such as working in forensic biology/DNA, or as a case manager (see Table 1, for details). One-third (n = 16, 33.3%) of triaging experts reported that they had a supervisory/ managerial duty in addition to their current forensic role. Participants were working in North America (USA (n = 13, 27.1%) and Canada (n = 10, 20.8%)), Europe (Belgium (n = 7, 14.6%), Sweden (n = 3, 6.3%), Austria (n = 1, 2.1%), France (n = 1, 2.1%), Moldova (n = 1, 2.1%), Netherlands (n = 1, 2.1%), Romania (n = 1, 2.1%), Switzerland (n = 1, 2.1%), UK (n = 1, 2.1%)) and Asia (Thailand (n = 3, 6.3%), India (n = 1, 2.1%), Turkey (n = 1, 2.1%), and UAE (n = 3, 6.3%)).

Table 1:

Demographics of triaging experts sampled.

Education Level Completed (n, %)
High school diploma/A-levels or equivalent	5 (10.4%)
Technical/community college	4 (8.3%)
Undergraduate degree (BA/BSc/other)	14 (29.2%)
Graduate degree (MA/MSc/MPhil/other)	18 (37.5%)
Doctorate degree (PhD/other)	6 (12.5%)
Other*	1 (2.1%)
Current Role (n, %)
Crime scene investigator	26 (54.2%)
Forensic biology/DNA examiner	5 (10.4%)
Forensic evidence recovery	1 (2.1%)
Multi-role (CSI, biology and/or fingerprints)	8 (16.7)
Forensic pathology	2 (4.2%)
Forensic case manager**	2 (4.2%)
Other***	3 (6.3%)
Not reported	1 (2.1%)
Testify in court? (n, %)
No	9 (18.8%)
Yes, I have testified in court more than 100 times	5 (10.4%)
Yes, I have testified in court between 25–100 times	14 (29.2%)
Yes, I have testified in court less than 25 times	20 (41.7%)
Size of provider (n, %)
100 or more	13 (27.1%)
25–99	12 (25.0%)
10–24	12 (25.0%)
9 or less	8 (16.7%)
I am not sure	3 (6.3%)

^*One participant reported “MS and CAS” which is interpreted as Master’s in Science and Certificate of Advanced Study

^**One participant reported being a forensic case manager and the other as a director

^***The participants reported generic terms, like “forensic investigator”, which could not be categorized into the other roles.

The expert practitioners were recruited in two main ways. First, forensic science organizations, such as the American Academy of Forensic Sciences, UK College of Policing, and the European Network of Forensic Science Institutes, were contacted to distribute the survey to their contact database. Second, personal contacts of managers or other representatives of forensic services providers were contacted for them to distribute or post the survey to staff members that met the inclusion criteria.

2.1.2. Experiment 2: Novices

The selection criteria for novice participants were as follows: “adults with a minimum age of 18 and a minimum Prolific approval rating of 95%.” Data were collected from N = 102 novice participants via Prolific, and novices could come from any country. Four participants were excluded, leaving a final sample of N = 98: one did not provide the correct Prolific completion code needed to verify participation, and three offered text responses in a non-English language. Non-expert participants were randomly allocated by Qualtrics to the low-pressure condition (n = 48, 49%) and high pressure condition (n = 50, 51%) (see Table 2 for the other demographics).

Table 2:

Demographics of novice participants.

Sex
Male (n, %)	49 (50%)
Female (n, %)	47 (48%)
Prefer not to say (n, %)	2 (2%)
Age in Years
Mean (SD)	26.4 (9.3)
Range	18–60
Education Level Completed
High school diploma/A-levels or equivalent	32 (32.7%)
Technical/community college	12 (12.2%)
Undergraduate degree (BA/BSc/other)	37 (37.8%)
Graduate degree (MA/MSc/MPhil/other)	17 (17.3%)
Doctorate degree (PhD/other)	0 (0%)

2.2. Overall Design

This is a 2 × 2 between-subject design (i.e., pressure context: high vs. low and expertise: novices vs. experts). All subjects did the same tasks on ambiguity aversion at the end of the experiment.

2.3. Casework Pressure Manipulation

2.3.1. Pressure Manipulation Design

Participants were assigned the role of a forensic scientist, and randomly allocated to a high- or low-pressure scenario (see Appendix 1). Each imaginary scenario contained the same four characters (inspired by the storytelling research in [35]: Mr. Peter Smith (the deceased), Olivia (the participant’s manager), Adnan (the participant’s colleague) and Michael (the lead police investigator on the case).

The story in these two scenarios was centered around the deceased, Mr. Smith. Differences in the socioeconomic status of the deceased were used as the basis to justify casework pressures. In the high-pressure condition, the deceased was a CEO of an international oil and gas company, with a master’s degree in Business Administration from Alpha University. In the low-pressure condition, the deceased was portrayed as a gas station attendant, following a six-month period of unemployment because of the pandemic. In this condition, the deceased’s highest degree was Grade 10 from Alpha High School.

In addition to the written information about the deceased, a portrait image of Mr. Peter Smith was shown to the participants. The portrait image of the deceased was selected so that it contains features similar to the deceased in the real crime scene images used in this study, such as having white skin and grey hair. This portrait image was manipulated to portray a CEO-looking person (e.g., wearing a tie) or a gas station attendant (e.g., wearing a uniform). The portrait images of the other characters (i.e., manager, colleague, and police investigator) were identical in the high- and low- pressure scenarios. The portrait image of Mr. Smith and all other characters in the two scenarios were taken from www.thispersondoesnotexist.com. The website provides images that have been generated by combining thousands of images together. Hence, the resulting image looks like a real person, but is not an image of a real individual, in order to adhere to ethics requirements.

Each scenario entailed eight stages, consisting of tasks that resembled a real-world 9am-5pm workday within a forensic science setting (see Table 3). To ensure that participants were engaged in the scenarios, they were asked to make decisions throughout the eight stages. In stage I, participants were asked to imagine that they work in Crime and Forensic Laboratory Services and were instructed to engage in the tasks as if they were completing them in real life. Next, the tasks during the morning part of the workday were presented, starting with Stage II, in which participants were asked to complete a general skills test. This test was portrayed by the manager, Olivia, to form part of the yearly performance evaluation of the participants. The test consisted of eight questions (four general knowledge questions and four mathematical questions). In the high-pressure condition, participants were asked to complete the questions within a time limit and with feedback given. In the low-pressure condition, participants were asked to complete relatively easy questions, with no time limit and no feedback (see Appendix 1 for details). The aim of this stage was to induce stress feelings in participants by using the method detailed in [36]. This method combines two stress elements, which are common in professional workplaces e.g., [37,38]: social evaluative threats (when one is negatively being judged by others) and uncontrollability (when nothing can be done to control a situation).

Table 3:

The eight stages in pressure manipulation.

Working Hour	Manipulation Stage	High Pressure Condition	Low Pressure Condition
N/A	Stage I: General Introduction	• Participants are asked to imagine that they work at Crime and Forensic Laboratory Services.
9.02 am	Stage II: Yearly Performance Evaluation (General Skills Test)	• Participants are asked to answer eight mathematical/general knowledge questions under time pressure and with feedback provided	• Participants are asked to answer eight comparable mathematical/general knowledge questions but without time pressure or feedback
9.50 am	Stage III: Forensic Case Briefing	• Case brief, containing elements of high socioeconomic status of the deceased, is presented	• Case brief containing elements of low socioeconomic status of the deceased, is presented
10.23 am	Stage IV: Email from Manager	• Participants are asked to respond to an urgent email from their manager, with perceived time pressure	• Participants are asked to respond to a routine email from their manager, without perceived time pressure
11.04 am	Stage V: Performance Reminder from Colleague	• Participants are asked to respond to their colleague who asks about performance in the general skills test, with perceived time pressure	• Participants are asked to respond to their colleague who asks about performance in the general skills test, without perceived time pressure
12.45 pm	Stage VI: News Headlines	• Participants are asked to decide whether they want to share a headline from an international news outlet about the current case	• Participants are asked to decide whether they want to share a headline from a local news outlet about the current case
2.45 pm	Stage VII: Phone Call from Police Investigator	• Participants are asked to respond to a phone call from the lead police investigator, with perceived time pressure	• Participants are asked to respond to a phone call from the lead police investigator, without perceived time pressure
4.31 pm	Stage VIII: Meeting with Top Management	• Participants are asked to make choices in response to an urgent meeting with top management, with perceived time pressure	• Participants are asked to make choices in response to a routine meeting with top management, without perceived time pressure

In Stage III, a short briefing about the case of Mr. Smith was introduced to participants. The case details were the same in the high- and low- pressure conditions except for the socioeconomic details of the deceased, i.e., workplace, education and portrait image. Next, in Stage IV, participants were asked to respond to an email from Olivia. In the high-pressure condition, the email contained pressurizing messages, such as ‘urgent’ and ‘important’, whilst the low-pressure condition did not (e.g., see email urgency bias as a source of pressure and stress [39]). In addition, a countdown timer was shown in the high-pressure condition only. The pressuring deadline was shown in four of the remaining stages throughout situations that could require a limited time to address them, such as a verbal conversation (i.e., Stages IV, V, VII and VIII; see Appendix 1). It should be clarified that the deadline was a perceived one, since participants could submit their responses, even after the time had expired.

In Stage V, the participant was asked to respond to a conversation with Adnan, the colleague, who asked about the participant’s performance in the general knowledge test. This task was included to reinforce any negative feedback that may have been received in the high-pressure scenario, thus potentially increasing the social-evaluative threat and stress [40,41]. Next, during lunchtime, the participant is asked to decide whether to share news headlines within their network that detailed the investigation (Stage VI). The news headlines were covered by an international news outlet (i.e., BBC) in the high-pressure condition, and by a local news outlet in the low-pressure condition. In Stage VII, the participant was informed about a phone call from Michael, the lead police investigator in the hypothetical scenario. The call, in the low-pressure scenario, did not contain any pressuring messages. In contrast, the high-pressure scenario call contained stressful terms, such as ‘quickly’ and ‘monitoring.’ At the end of the workday, the participant was asked to choose from several options in response to a meeting notification from top management (Stage VIII). The meeting in the high-pressure condition was set as urgent and had to take place on the same day, whereas the low-pressure meeting was routine and could wait for the next workday.

2.3.2. Pressure Manipulation Check

Two self-reported measures were used to determine that the pressure manipulation was effective by asking about the engagement in the task and the stress level:

Looking back at the scenario of your 9am-5pm working day, to what extent do you agree with the following statements:

I was engaged in the scenario, and I felt I was part of the forensic team working on Mr. Smith’s case. (Strongly Disagree, 0% - Strongly Agree, 100%)

I felt I was facing pressures working on the forensic case of Mr. Smith. These may include internal pressures (originating within Crime and Forensic Laboratory Services), or external pressures (originating from outside the laboratory), or a combination of both. (Strongly Disagree, 0% - Strongly Agree, 100%)

To minimize the possibility of hindsight [42]), the two questions were placed at the end of the forensic tasks. In addition, there was an opportunity for participants to express any other comments in an open-text format.

2.3.3. Piloting Pressure Manipulation Stimuli

Two studies were conducted with non-experts, using Prolific Academic, to pilot the pressure manipulation stimuli (see Table 4). In the first pilot study, pressure and engagement level measures were inserted straight after the pressure manipulation, while in the second pilot, these two measures were placed after the stage of viewing crime scene images and making triaging decisions, as would be conducted in the actual experiments. The aim of pilots was to develop and test the pressure manipulation, before launching the study with busy practitioners who are difficult to recruit.

Table 4:

Descriptive statistics of novice participants sampled.

	Pilot (1)	Pilot (2)
N	40	39*
Condition
High Pressure (n, %)	20 (50%)	18 (46%)
Low Pressure (n, %)	20 (50%)	21 (54%)
Sex
Male (n, %)	20 (50%)	20 (51%)
Female (n, %)	20 (50%)	19 (49%)
Age in Years
Mean (SD)	25.63 (8.82)	24.28 (5.44)
Range	19 – 63	18 – 45

^*One participant was excluded as their text responses were in a non-English language.

The data presented in Table 5 indicate the pressure manipulation that was piloted was found to be effective. In both pilots, participants in the high-pressure condition felt with significantly higher-pressure level than participants in the low-pressure condition, while the engagement levels remained equal in the high pressure and low pressure conditions.

Table 5:

Perceived engagement and pressure levels.

	High Pressure	Low Pressure	Student’s t-test
	% Mean (SD)	% Mean (SD)
Pilot (1)
Engagement Level	81.40 (20.18)	77.00 (19.75)	t(38) = −.697, p = .49, d = −.22
Pressure Level	70.55 (32.01)	40.95 (32.25)	t(38) = −2.91, p = .006, d = −.92
Pilot (2)
Engagement Level	79.33 (24.87)	79.52(19.16)	t(37) = −.027, p = .98, d = −.01
Pressure Level	68.61 (28.18)	50.24 (29.90)	t(37) = 1.96, p = .057, d = −.63*

^*Approaching significance.

2.4. Forensic Case Vignette

A forensic case vignette was made available for the participants to view, as they would have in typical casework (see Appendix 2). The vignette was the same in the high- and low-pressure conditions. It consisted of a case brief and high-quality crime scene images, taken from a real case. The brief clarified that the deceased, Mr. Smith, was found dead on the street, below the building where his one-bedroom apartment is located. The case brief also stated that there were no witnesses at the time of the crime.

The crime scene images were carefully selected from a database of old cases, so that the case was ambiguous in nature. The selected case contains elements to suggest a homicide (e.g., blood-like stains and a knife), and other elements that suggest a suicide (e.g., no disarray in the crime scene and possible wrist cuts to Mr. Smith’s right hand). In the actual closed case, the forensic experts reported an expert opinion that it was a suicide. The crime scene images showed only three locations: around the body, the toilet and the bedroom of the apartment. This was for two reasons: (1) to increase the ambiguity of the case, and (2) to make the study design efficient so that busy forensic experts could participate. The scene images are not displayed in Appendix 2 to maintain anonymity of the case, but they can be requested from the corresponding author upon reasonable request.

There was a total of 35 items to be triaged. These items were located in/around the body (n = 8), in the toilet (n = 19), and in the bedroom (n = 8). In essence, all possible items that could be reasonably viewed from the crime scene images were listed to be prioritized by the participant. Some of the items were labeled as ‘swabs’ (e.g., bloodstain swab) or ‘recovered marks’ (e.g., recovered marks from the doorknob). These two terms were explained to novice participants, to minimize ambiguity. Notably, some of the items appear not to be relevant to the case (such as toothbrush, books, etc), while other items appear to be relevant to the case (such as bloodstains, knife, etc). Since the images come from a real case, the ground truth is not known. This is different from previous research on crime scene decision-making (e.g., [43]), in which items could be classified as either ‘crime scene’ traces or ‘foils’ because the ground truth was known.

2.5. Forensic Science Decisions

2.5.1. First Impression on the Crime Scene

Participants were able to view crime scene images with no time limit, and they were then asked to briefly describe their impression of what had occurred. On the next Qualtrics slide, participants were asked to make one of three impression choices: homicide, suicide or inconclusive. In addition, participants were asked to rate their level of confidence that their impression was correct from 0% (not confident at all) to 100% (extremely confident). In the context of this study, first impression and confidence about it are considered decisions pertaining to triaging decisions, since there is evidence from previous research that the initial hypotheses can also play a role in subsequent decisions (e.g., [43,44]). Lastly, participants were given an opportunity to offer any other comments about the decisions they made.

2.5.2. Forensic Triaging Strategy

Participants were asked to develop a forensic triaging strategy. In this study, a triaging strategy refers to prioritizing items, after they were collected from the crime scene, to be sent (or not) for forensic analysis. There is no correct or incorrect triaging strategy [45]; rather, we were interested in observing any meaningful changes in the prioritization of crime scene items, as a result of induced casework pressure, or aversion to ambiguity. There were four options from which participants could choose:

• BIOLOGY ONLY: This choice meant that the participant wanted to screen for any biological traces, such as blood, semen, skin cells (touch DNA), saliva, urine, hair, and/or feces.

• FINGERPRINTS ONLY: This choice meant that the participant wanted to screen for either visible or latent fingermarks through any available fingerprint development or collection methods available, such as photography, alternative light sources or chemical treatment.

• BIOLOGY AND FINGERPRINTS: This choice meant that the participant wanted to screen for both biological and fingermarks.

• NO EXAMINATION: This choice meant that the participant did not want to send the item for forensic analysis to screen for biological or fingermarks at this stage.

We decided to limit the triaging to fingerprints and/ or biology/ DNA for two reasons. First, these two types of forensic testing are common and considered strong identification traces (versus e.g., toolmarks). Second, to simplify the design. Since this is an exploratory study on complex triaging decisions, we hope that future research can explore other types of testing (e.g., chemistry, toolmarks, firearms). As a first step in investigating this issue, we wanted to keep it relatively simple.

For analysis and interpretation purposes, the above options were grouped into two broad categories:

• single-targeted triaging strategy (testing for biological traces only or fingermarks only)

• multi-targeted triaging strategy (testing the item for both biological traces and fingermarks).

Each of these two broad testing strategies has advantages and disadvantages in practice. For instance, a multi-targeted testing strategy might increase the chances of finding evidence that is valuable for the case, but it naturally requires additional resources and costs. It should be noticed that, in this study, some of the items can only be triaged through a single-targeted study (e.g., blood swabs or recovered fingermarks).

The participants also had the option not to triage any of the items for any type of test. However, it might be easier to think about or interpret the total number of items that experts examined, than those not examined. Hence, a new variable was added, which is the total number of examined items. This new variable complements the ‘no examination’ variable.

Lastly, participants were asked about their confidence levels, in that their triaging strategy would have probative value to the case of Mr. Smith. They were further asked if they wanted to offer any comments about how they decided to formulate their triaging strategy.

2.6. Aversion to Ambiguity

The degree of an individual’s aversion to ambiguity was measured using a modified paradigm of Saposnik et al. [20] (see Appendix 3 for details). Here, participants were asked to choose one of two monetary visual options. The first option consisted of a 50% chance of winning $400 and a 50% chance of winning $0. This is the standard option, which remained the same in all the monetary scenarios. In the second option, 50% of the probability information was hidden, so it consisted of a 50% unknown probability or ambiguous chances of winning an amount (see example in Figure 1).

There were nine different winning amounts, ranging from $200 to $1,000. In this paradigm, an ambiguity neutral person would interpret the ambiguous option centred around 50% winning chance, as representing a 50% success chance. Since the standard option offered $400 with a 50% chance, the ambiguity neutral decision maker would be indifferent between the standard option and the ambiguous option if it offered $400. In addition, there were two ambiguous options that were dummy choices (i.e., 30% of unknown probability and 80% of unknown probability of winning $600). These dummy choices were inserted to minimize the possibility of survey fatigue, if all choices were at 50% ambiguity level. The monetary scenarios were randomized via Qualtrics.

2.7. Procedure

The study consisted of five steps. In Step 1, participants were randomly allocated to either the high- or low- pressure condition. In Step 2, they viewed the crime scene images and made interpretations of what they thought happened at the crime scene (i.e., the first impression). Next, participants were asked to triage each of the 35 items collected from the scene (Step 3). After deciding whether or not to send these items for forensic analysis, participants were asked about their confidence in their testing strategy, and to rate retrospectively the perceived engagement and pressure levels in working on the scenarios. In Step 4, participants were asked to make choices on the monetary scenarios, and in Step 5, they were asked to provide demographical information (see Figure 2). Two attention check screeners were included in the study to check whether participants read the instructions [46,47]. Furthermore, participants were asked not to discuss the study with anyone, in order to not impact the findings. This research was approved by the Yale Institutional Review Board (Project ID: HIC 0910005795), and all participants signed an informed consent form.

Figure 2:

Graphic timeline of the experimental procedure.

2.8. Statistical Approach

Two human factors are tested for their effect on forensic triaging decisions: casework pressures and ambiguity aversion. Chi square was used to test differences in the reported first impressions (categorical decisions). For the continuous, dependent variables (i.e., triaging by testing type, triaging by location, confidence levels on triaging strategy, and also on the first impression), a series of 2 × 2 ANOVAs were run to test if they differed by condition (High vs. Low Pressure) and/or by expertise (Experts vs. Novices). Bonferroni post hoc analyses were conducted to test differences among the forensic experts’ decisions under high vs. low pressures. This statistical approach is a similar approach to previously published studies, which involved cohorts of both experts and non-expert participants (e.g., [16,29]).

To estimate the individual’s ambiguity attitude, we calculated the proportion of ambiguous responses by the total trial count of nine in the economic tasks. This measure ranged from 0 (strong aversion to ambiguity) to 1 (strong tolerance to ambiguity). The current paradigm in measuring ambiguity aversion attitudes is novel and relatively shorter than typical economic ambiguity paradigms [22,48], so we first evaluated and compared the attitudes of both triaging experts and novices on the economic ambiguity task. Then, we tested whether the level of ambiguity aversion is associated with the forensic triaging decisions and confidence levels using Pearson’s correlations.

The study was not pre-registered. However, all raw data have been made publicly available in Open Science Framework and can be accessed at: https://osf.io/fgwbv/.

3. Results

3.1. Overall

Two human factors pertinent to the triaging process were explored: casework pressures and ambiguity aversion (Experiments 1 and 2). The between-expert reliability, qualitative responses on text-entry questions, and demographic information were also considered.

3.2. Casework Pressures

3.2.1. Pressure Manipulation Check

The pressure manipulation was effective for the triaging experts (Experiment 1). On average, expert participants in the high pressure condition reported significantly higher pressure levels (M = 57.95, SD = 34.87) than participants in the lower pressure condition (M = 25.48, SD = 25.66), t(46) = −3.72, p < .001, Cohen’s d = −1.08. Importantly, participants in both pressure conditions were equally engaged in the scenarios (t(46) = 1.02, p = .158, Cohen’s d = .296). Average engagement levels were reasonably high – above 70% – in both the low pressure (M = 78.74, SD = 25.22) and high pressure (M = 70.52, SD = 30.83) conditions.

The pressure manipulation was also effective for novice participants in Experiment 2. On average, novice participants in the high pressure condition reported significantly higher pressure levels (M = 65.66, SD = 27.30) than participants in the lower pressure condition (M = 47.65, SD = 26.69), t(96) = −3.30, p = .001, Cohen’s d = −.667) . Participants in both pressure conditions were equally engaged in the scenarios (t(96) = .324, p = .747). Average engagement levels were high – above 80% – in both the low pressure (M = 82.46, SD = 17.71) and high pressure (M = 81.36, SD = 15.82) conditions.

3.2.2. First Impression

There was no statistically significant association between induced pressure and first impression, among the triaging experts (χ²(2) = .636, p = .728) or among novices (χ²(2) = 1.081, p = .582) (see Figure 3). Bonferroni post hoc analyses of 2 × 2 ANOVA revealed no difference in the confidence level of experts and novices across the pressure conditions (p > .05, see Figure 4, and Appendix 4 for details on the statistical analysis). In addition, confidence levels remained not significantly different (p > .05), after inconclusive impressions were excluded.

Figure 3:

First impression by condition and expertise.

Figure 4:

Confidence levels by condition and expertise.

3.2.3. Triaging Decisions

All triaging decisions, across testing type and location, were comparable in the high and low pressure conditions for both the experts and novices (p > .05; see Figures 5 and 6). Although experts in the high-pressure (M = 6.19, SE = 0.32) condition collected more traces from the body compared with the low-pressure condition (M = 5.26, SE = 0.28, p = .026; See Figure 6), this finding did not survive correction for multiple comparisons.

Figure 5:

Triaging decisions by testing type, condition and expertise.

Figure 6:

Triaging decisions by location, condition and expertise.

3.3. Ambiguity Aversion

3.3.1. General Attitude on Economic Ambiguous Decision-Making

We evaluated and compared the attitudes of both triaging experts and novices on the economic ambiguity task. This comparison was performed for two reasons. First, because the current paradigm in measuring ambiguity aversion attitudes is novel and relatively shorter than typical economic ambiguity paradigms [22,48], and so it would be useful to evaluate whether the measure is effective for this study. Second, because the economic ambiguity tasks are general tasks, not specific to forensic practice, we are able to directly compare attitudes of non-experts with experts.

There was substantial variability across expert and novice participants in the degree of ambiguity aversion, as shown in Figure 7. Interestingly, experts were slightly more ambiguity averse than novices (two-tailed t-test t(144) = 1.72, p = .088 (approaching significance), d = 0.30; see also Figure 4.1 in Appendix 4).

Figure 7:

Proportions of ambiguous choice made by the experts and novices across all the nine monetary trials. An ambiguity neutral person would be indifferent at $400.

12 out of 48 expert participants (25%) and 10 out of 98 novice participants (10%) were strongly averse to ambiguity (i.e., they chose the non-ambiguous option in all the nine trials), while the rest of participants varied in their responses.

3.3.2. Economic Ambiguity Tasks across Pressure Condition

There is mixed empirical evidence on whether ambiguity aversion is a stable construct or not (i.e., whether it changes with time, conditions, other factors; e.g., [49,50]). To test this, the mean ambiguity proportions of ambiguity aversion were compared in the high and low pressure for both experts and novices (see Figure 8). Experts in the high pressure condition were more ambiguity averse (M = 0.29, SD = 0.32) compared with the low pressure group (M = 0.42, SD = 0.31), F(1, 142) = 3.52, p_expertise = .0626 (approaching significance). However, ambiguity attitudes were comparable across the pressure conditions for novices (p > .05, see Figure 8).

Figure 8:

Ambiguity attitude of experts and novices across pressure conditions.

3.3.3. Triaging Decisions

We sought to explore whether a tendency of experts to report inconclusive first impressions could be related to their ambiguity aversion attitudes. Reporting inconclusives, as opposed to reaching a decisive outcome (like, a ‘match’ or ‘non-match’ for fingerprints) can have implications in practice (e.g., see [51,52]). In order to test this question, we combined ‘homicide’ and ‘suicide’ impressions in a new variable called ‘decisive’ impression. We hypothesized that inconclusive impressions would be associated with greater aversion to ambiguity. Consistent with our hypothesis, participants who reported inconclusive impressions tended to be more ambiguity averse (M = 0.28, SD = 0.25) than those who did not (M = 0.41, SD = 0.34, at t(46) = 1.47, p = .074 (approaching significance), d = 0.44; see Figure 9).

Figure 9:

Inconclusive vs. definitive impressions (i.e., homicide and suicide combined) by the experts and the proportions of ambiguous choices made by forensic experts.

Furthermore, exploratory findings suggest that experts who are more tolerant to ambiguity appear to lean towards a single-targeted triaging strategy (i.e., biology-only or fingerprints-only), as opposed to multi-targeted triaging strategy (i.e., both biology and fingerprints). Specifically, when uncorrected for multiple comparisons, the triaging experts who were more tolerant to ambiguity tended to test more for fingerprints (r = 0.38, p = .008), and, for fingerprints OR biology (r = 0.28, approaching significance at p = .052). Levels of ambiguity aversion were not correlated with any other dependent variables (p > .05, see also Figure 4.2 in Appendix 4).

3.4. Qualitative Responses

We coded the responses to open-ended questions in themes, similar to an approach by Lidén and Almazrouei [16]. The qualitative responses can offer further insight beyond the quantitative data, such as any potential differences in justification between the low and high pressure experts may provide possible insight into differences of behavior. Some of the open-ended questions were optional (e.g., “Would you like to add any comments about the monetary scenarios that you just completed?”), and are used here as supplementary evidence on why experts made their decisions. Here, we discuss the data from the expert participants. Further details on the open-ended questions are available in Appendix 5.

The two questions related to triaging decisions (i.e., first impression and testing strategy) show minor differences between the high- and low-pressure conditions. For instance, for the open-ended question titled, “What is your impression about what happened to the deceased in the crime scene”, 48% (10 of 21 responded) of triaging experts in the low pressure condition indicated that the case was inconclusive, and they needed more forensic information, whereas 30% (8 of 27 responded) of experts in the high pressure condition indicated the same theme. In addition, 30% (8 of 27) of practitioners in the low-pressure condition indicated the importance of the injuries on the deceased’s wrist, while only 14% (3 of 21) of practitioners in the high-pressure condition indicated the same theme.

Further, the open-ended question, “Would you like to add any comments about how you decided on your forensic testing strategy?” allowed for practitioners to offer subjective reasons behind their chosen testing strategies. Of those who answered the question, 40% (3 of 8) practitioners in the low pressure condition recommended further forensic analyses, while 0% (0 of 9) practitioners in the high-pressure condition indicated the need for further analysis. It is also worth noting that 2 practitioners, one from each condition, considered the cost and efficiency of the forensic tests in their decisions, while 1 practitioner in the low-pressure condition disregarded cost, and only focused on the value of the forensic testing in their decision.

3.5. Between-Expert Reliability

We were also interested to explore whether experts would be consistent in their decision-making when accounting for the pressure manipulation (i.e., between-expert reliability). We explored between-expert reliability in two ways: by selecting three triaging experts of comparable demographics as an illustration example, and by showing the variability of triaging decisions by all experts.

Hence, we selected three experts, all in the same high pressure condition who all made a suicide impression. In addition, we selected participants who operate in the same working context (i.e., country) and have similar years of experience in the triaging of forensic traces (see Table 4). This is a similar approach to a previous study that compared CSI decision-making among sub-sample of two individuals with identical conditions and comparable background demographics [16].

As can be seen from Table 6, the three experts were not consistent in the triaging strategy for forensic items, even under the same experimental conditions. First, there was a wide variability on how many items were triaged in total (i.e., a range of 13 to 32 items). Second, there were noticeable differences on the type of testing (e.g., Participant A triaged for biological testing two times more than Participant B). However, the targeted locations of triaged items seem consistent among the three participants, with the highest ratio of items being collected from the bathroom.

Table 6:

Triaging decisions made by three expert participants in the same pressure condition.

	Participant A	Participant B	Participant C
Pressure Condition	High	High	High
Reported Impression	Suicide	Suicide	Suicide
Current Role	CSI (Multi-Role)	CSI (Multi-Role)	CSI
Years of Experience	6	9	9
Country*	X	X	X
Total Tested
	32	22	13
Type of Testing
Biology	22	11	6
Fingerprints	6	6	6
Biology AND Fingerprints	4	5	1
Location
Body (n=8, %)	7, 0.88	4, 0.50	3, 0.38
Bathroom (n =19, %)	18, 0.95	14, 0.74	8, 0.42
Bedroom (n = 8, %)	7, 0.88	4, 0.50	2, 0.25

^*All the three participants work in the same country, X, which is kept anonymous.

Selecting three experts who share a similar background, in a manner akin to previously published research [16] offers insight into the degree of variability that can exist between expert, even when they have similar backgrounds . It is important to consider this finding in the context of the variability observed across all the experts, and Figure 10 shows the broad variance of total items triaged by all experts in the low pressure (range: 12–35; SD = 5.95) and high pressure conditions (range: 12–35; SD = 6.78).

Figure 10:

Box plot showing the broad range of total items triaged by experts in the low and high pressure conditions.

3.6. Moderating Factors

Statistical analyses were conducted to check whether any of the demographical information moderated triaging decision-making (i.e., the crime scene impression, confidence level of that impression, total triaged items for testing, and confidence levels on the triaging strategy). These analyses were done for both triaging experts (Experiment 1) and novices (Experiment 2).

None of the demographical information moderated the decisions made (all at p > .05; see Tables 1 and 2 for all the demographical information). There were two exceptions. First, the role of the experts affected the initial scene impressions, χ² (14) = 38.75, p < .001, V = .635, but not the other decisions as outlined before. Hence, we conducted a secondary analyses to explore whether the role of the experts played a role in triaging decision-making. To do so, we analyzed data of a subset of experts who identified working as crime scene investigators or CSIs (N = 34; n = 15, high pressure and n = 19, low pressure), thus enabling comparison of a more coherent group of experts in terms of their role. The sample of this subset of CSIs is comparable to previously published studies (e.g., N = 30 in [28] and N = 37 in [16]). The exploratory analyses for CSIs showed no significant findings across the conditions (all at p > .05). That is, CSIs behaved similarly in the high and low pressure conditions, in terms of their impressions about the scene, confidence levels about the impression, total items triaged, type of testing for triaged items, location of triaged items, and confidence level about the triaging strategy. In addition, ambiguity aversion of CSIs did not predict any of these decisions (p > .05). Finally, regardless of the pressure condition, CSIs (N = 34) did not differ from those who did not identify as CSIs (N = 14) in the mean of total examined items (p > .05), further demonstrating that the current role was not a main factor in this study.

The statistical analyses on moderating factors also revealed that the total number of triaged items seemed to slightly differ by the region that the experts operate in, F(4, 43) = 2.99, p = .029, partial η² = .218. Figure 4.3 in Appendix 4 shows that, on average, triaging experts from Asia triaged more items for testing compared with other countries/ regions. To clarify, for testing for any potential variations among the countries that experts operate in, we grouped some European (n = 10; Austria, France, Moldova, Netherlands, Romania, Sweden, Switzerland and UK) as “Other European” and Asian (n = 8; India, Thailand, Turkey and UAE) countries. This enabled comparing these grouped countries of low participant number with Belgium (n = 7), Canada (n = 10), and USA (n = 13).

4. Discussion

4.1. Casework Pressures

To the best of our knowledge, this is the first study in forensic science that utilizes storytelling scenarios in a behavioral experiment. Our successful pressure manipulation suggests that a combination of realistic algorithmic generated images, engaging tasks, and/or perceived deadlines can generate feelings of pressure in participants, even in online environments. It is increasingly recognized that online research offers several advantages, such as accessing a more diverse cohort of participants [53] and overcoming geographical barriers [54].

We found that casework pressures, as induced in this exploratory study, were not sufficient to influence triaging decisions made by experts and novices. On average, the triaging decisions were comparable between the high- and low- pressure conditions by testing type and by location (see Figures 5 and 6). We offer four possible explanations for these null findings. First, working under high pressuring cases may not have practical impact on which items are prioritized and which ones are not. Indeed, the qualitative data may support this explanation in that almost all triaging experts did not raise clear-cut connections between the pressure manipulation and their decisions about the impression of the scene and/or how they triaged samples. For example, the experts appeared to focus on the materials that may have value in the case (e.g., bloodstains for identification and items to link deceased with the scene, etc.; see Appendix 5). Going along this explanation, it might be the case that feelings of pressure may have more influence on decisions made at the crime scene where the environment is chaotic and time-pressured [2] instead of on triaging decisions after items were collected from the crime scene. Hence, it is worth exploring the effect of casework pressures for crime scene decisions (e.g., searching for or collection of forensic traces).

Second, the current pressure paradigm may not have exerted enough pressure on participants to influence their decisions. Even though there was a significant difference in pressure levels between the high- and low- pressure groups, the mean reported pressure levels were relatively low across the board, with the highest mean pressure level being 65.66% in the high pressure, novice group. Furthermore, the pressure paradigm was more casework general as opposed to triage-specific, which may also go some way toward explaining the null effect. Future research could explore triage-specific pressures (e.g., making explicit experimental conditions of limited vs. unlimited resources; e.g., [16]).

Third, it could be argued that the manipulation does not measure the impact of casework pressures per se. The manipulation includes other factors, such as information about the socioeconomic status of the deceased. All these elements, however, are irrelevant to the triaging items for forensic analysis, and should not affect triaging decisions. In addition, the manipulation check question was specifically about casework pressures. Hence, at a minimum, the manipulation addresses whether task-irrelevant information, including casework pressures, may have influence on forensic triaging strategies (see [55] for a discussion). Future research could study the causal effect of contextual socioeconomic information on forensic science decisions, noting that other professional domains, like medicine (e.g., [56]) explored these relationships to improve expert decision-making.

The fourth possible explanation for the non-significant finding is that experts (and novices) were inconsistent in their triaging decisions to the degree that the pressure manipulation did not reveal a pattern, or a change in triaging, across conditions. This explanation is supported when looking at the lack of reliability of decisions made by triaging experts. As shown in Table 6, the three experts can be considered to be directly comparable: they are in the same pressure manipulation condition, have the same role, work in the same country, similar experience in triaging, etc. Yet, their triaging decisions were different (e.g., total triaged items ranged from 13 items to 32 items; see also Figure 10 for the whole expert pool). This finding has a practical implication; it highlights the possibility that the outcome of a case may depend on which expert did the triaging, for example, if one expert handled the case, then the suspect may be identified, but if another expert handled the same case the outcome may have been different (see, “Eeny, Meeny, Miny, Moe, With Which Expert Should We Go?”; [57]).

The lack of consistency between experts in triaging decisions raises interesting questions. All the triaging experts who met the inclusion criteria viewed the same photos and were given the same triaging choices, yet there was a wide range in what was triaged and for what type of testing. As discussed previously, it is acknowledged that triaging tasks are relatively complex, and there are significant variations in resources and organisational contexts that may affect the triaging task (see organizational factors, or level 5, as part of human factors in forensic decision-making; [11]). These findings indicate that there is value in exploring ways to develop more transparent and reproducible triaging methods. For example, previous studies conducted inter-laboratory studies to measure baseline assessments of existing methods for bloodstain pattern analyses [58,59] and even for more novel, quantifiable methods in duct tape physical fits [60], yet similar studies are lacking for triaging tasks. Importantly, understanding feedback loops and the rationale behind subjective assessments can help refine and develop more consensus-based methods [60]. Developing a collaborative approach to finding a path forward that incorporates both explicit and implicit forms of knowledge [61] that brings both established procedures and creates space for expertise to the triaging task appears to be a valuable next step. It may also be important to consider how triaging decisions are currently documented and whether there is space for enhancing the transparency of the steps and rationale taken in each situation that led to those decisions See, Forensic Disclosure; [62,63]) as well as considering current training programmes.

Some previous studies compared experts and novices with forensic recognition tasks (distinct from pattern-comparison tasks) that are similar to triaging decisions, such as the number of traces to be collected [29], or visually checked [34] from crime scenes. While this was not the main research question addressed in this current study, we note that it is possible to identify a clearly different behavior of the experts to that of non-experts (see Figures 3–6, and supplemental analysis in Appendix 4). Here, we address decision points in terms of first impression and overall triaging decisions.

Most of the triaging experts thought that the deceased killed himself (i.e., suicide), whereas most of the non-experts thought that someone else killed the deceased (i.e., homicide). Interestingly, in the real case—where the scene images for the current study were taken from—the experts ultimately reported a suicide. Hence, it is reasonable for the purposes of this discussion to work on the understanding that ‘suicide’ is the ‘correct’ categorical decision. One plausible explanation is that non-experts were less flexible in updating their initial hypothesis of homicide. For example, the initial images portrayed a potential weapon (knife) and blood, thus suggesting homicide. It is possible that non-experts had a tunnel vision about homicide, and discounted other evidence that suggested alternative scenarios, such as possible wrist injuries for suicide (e.g., see a discussion on the topic of alternative hypotheses; [64]). The qualitative data may also support this in that 11 of 48 (23%) reported on possible injuries to the wrist, whereas only 1 of 98 (1%) novices did so. Novices may not have reported, or even noted, the wrist injuries due to the lack of expertise. Another influencing and biasing factor maybe base-rate [11], in the sense that experts know that suicide is more common than homicide.

When comparing the items triaged by novices with experts, a general trend can be observed in that experts are more selective in the items that they want to get tested than non-experts (see Total Examined in Figure 5). This is not a surprising finding and concurs with previous research findings, which demonstrated that experts look for specific cues that may have probative value as evidence [34], such as selecting traces that they think are crime-related [29].

4.2. Ambiguity Aversion

The exploratory evidence from this study suggests that experts who tend to dislike ambiguity seemed not able to make a conclusive impression about what happened to the deceased. Though not strong (p = .074, approaching significance, see Figure 9), this is still an important finding that requires attention for considering the possible practical implications. Previous research suggested that interpreting the information about what happened at the crime scene may unduly influence the practitioner’s thinking and may miss the evidence that do not fit their hypothesis, and look for traces that confirm their hypotheses (e.g., [43,65]). In essence, early interpretations can cascade and affect subsequent decisions [44]. In the current study, all the experts viewed the same crime scene images, yet some experts reported a categorical impression of ‘inconclusive’ as to what happened to the body, whereas other experts, given the same evidence, felt confident enough to make a conclusive impression (‘homicide’ or ‘suicide’). This might mean that the trait of the individual experts, like their tolerance to ambiguity, is an important human factor in reaching a threshold of making an impression, as opposed to solely being based on the data of the forensic case materials, like the crime scene images. However, we must emphasise that this finding is about first impressions, which are important, but final decisions are made only after practitioners conduct a thorough investigation. Future research could build on this finding to consider whether tailored training on risk management and educational interventions could minimize “knowledge-to-action” gaps, as is the case in the medical domain [20].

Having said this, it is important to note that reaching an inconclusive conclusion (e.g., in fingerprint assessments) carries different consequences to interpreting the crime scene as inconclusive (i.e., reaching an inconclusive first impression). If forensic experts reported an inconclusive conclusion—when it is possible to make conclusive decisions from the data offered from fingerprints, DNA or other traces—there is a risk of not convicting perpetrators, or miscarriages of justice (see a discussion in [51]) and more recent reports on how to score inconclusive as correct/incorrect decisions [52,66–68]. Having an inconclusive first impression about the crime scene can have different consequences, as discussed earlier. In addition, being conservative can be appropriate and should be allowed, especially in difficult and ambiguous forensic traces [69], or in early stages of an investigation when multiple pieces of evidence still need processing that lend weight to a particular hypothesis.

The exploratory findings from this study also suggest that experts who are not averse to ambiguity lean towards single-targeted triaging strategy, i.e., testing for fingermarks only or for biology only. That is, the more comfortable experts are with ambiguous situations, the more they tend to target a specific type of test, as opposed to nonspecific multi-targeted testing (with non-significant findings). This finding aligns with our expectation. It was predicted that experts who are highly averse would lean towards multi-targeted strategy, since it can be considered a maximizing strategy—when individuals are unsure, they would select to triage for both biology and fingerprints, rather than risking to test for one type of forensic test versus the other. Arguably, processing the same evidence, like a gun, for multiple types of forensic testing is a complex task and may help in explaining the emerging trend: “..decisions must be made about the order of swabbing for DNA, dusting or fuming for prints and firing the gun for ballistic testing. For example, the heat from ballistic testing of the gun may destroy the DNA on the gun or swabbing for DNA may eliminate fingerprints.” ([13], p.10). However, we must emphasize that the current findings are exploratory and must be considered with caution. For example, for some items to be prioritized in the current study, it would only make sense to apply a single targeted strategy (e.g., bloodstain swabs for biology only), which some novices may not have been aware of. Hence, there needs to be further studies that can unpack this complexity, perhaps by focusing on items which may go through multiple types of forensic testing.

Furthermore, a preliminary trend appeared within the triaging experts who were more ambiguity averse than non-experts, such as reporting more inconclusive impressions (Figure 9). These observations are consistent with some previous research reporting that fingerprint experts were more risk averse than the general population [27]. In practice, these findings may reflect the amount of information needed to reach a conclusive decision, such as identification or exclusion, or not reach a conclusive decision (e.g., [27]), impression or interpretation (e.g., [29]), such as the decision threshold [70]. Decision thresholds are dynamic, and can be affected by factors, like motivations to close the case [71] or time pressures [72]. Relatedly, a possible shift emerged in the experts’ tolerance to ambiguity; experts appeared to be more averse to ambiguity under high pressure compared with low pressure (see Figure 8). It should be noted that this is a between-subject observation, not within-subject, so the baseline ambiguity aversion levels are different to start with.

More broadly, it was interesting to observe the variability in the ambiguity aversion trend within the sample investigated (see Figure 7). Finding variability in such an efficient design (of nine counted trials) enables investigating the behavior of busy experts, such as medical doctors or forensic experts. Typically, investigating ambiguity aversion bias in behavioral economics studies entails large number of trials, sometimes more than 200 trials (e.g., [48]). Having long studies may hinder experts from participating, thus the current study supports evidence from previous studies on professionals that ambiguity aversion can be at least estimated with a small number of trials [19,20,73].

4.3. Moderating Factors

We conducted further analyses to understand whether there were any trends between any of the demographic factors and the triaging decisions. Nearly all the demographic factors did not seem to moderate the decisions (first impression, confidence levels, ambiguity aversion levels and total triaged samples). The two exceptions were that the role of experts influenced initial impression, and the total triaged items differing among the regions/countries grouped. The former exception was further explored through examining decisions of CSIs only (N = 34), and also examining their triaging decisions with non-CSIs, and no differences were found in the decisions of this more homogenous group of experts. The latter was not explored further because the differences are small and inter-country variations are not a main question in this study. Previous research also found no differences in collecting bloodstains among organizations of different countries that CSI experts came from [16]. However, 62% (n = 23) of the participants in this study reported that there are no rules or recommendations from the organizations on what traces to select, and it is up to the CSIs’ own judgment about the case [16]. This highlights the importance of ensuring transparency in how decisions are made and what information was known (or unknown) during the decision-making process [63].

4.4. Limitations

This exploratory study carries several limitations and so findings need to be considered with appropriate caution. First, this study relied on decisions based on (real) crime scene images. Having realistic scene images could add an advantage compared to mock scenes; for example, in contrast to mock scenes, the conditions of the images would remain the same even when several participants examined them before [16]. However, participants may have missed the feeling of experiencing an actual crime scene [29] and images may limit the understanding of a 3-D scene, though it was interesting that most participants were able to reach a conclusive impression from these images (i.e., homicide or suicide). One might also argue that, since the ground truth of the ambiguous scene is unknown, then accuracy of triaging decisions may not be established. This argument is acknowledged; however, of interest here is the observable change of participant behavior given the exact same crime scene (e.g., see, [16]), rather than accuracy. Establishing factors contributing to achieving more consistent decisions can be considered a key step forward for enhancing the triaging process (see discussion on consistency; [32]).

Second, the findings reported here are limited to the top-down (i.e., pressure manipulation) and bottom-up information (i.e., the case scenario) used for this study. In terms of the top down manipulation, naturally, the online pressure manipulation would be less effective than real-world pressures and stressors, such as making crime scene decisions in a limited time and with the context of incomplete information or unverified information (e.g., [2]). Future research could potentially take the storytelling approach in this study a step further for more realistic and engaging experiences, such as utilizing virtual reality or augmented reality tools (e.g., [74]), or simulations. In terms of bottom-up information, it is recognized that every crime is different [75]. Whilst insights from the data could be generalizable to different investigation contexts, it is imperative to bring further empirical data to address the sensitivity of each case and crime scene items within it.

Third, as discussed earlier, we used a simplified paradigm to estimate the ambiguity aversion levels, and the sample size can be considered not large for such a novel design in understanding ambiguity aversion in the context of forensic science. This indicates that the reported findings are exploratory in nature, rather than offering causational relationships. With future research work and samples, we can better understand the relationship between ambiguity aversion and forensic decision-making. Further, the current study used monetary scenarios to measure ambiguity aversion, in a similar manner to previous studies within professional domains such as medicine [20]. Future research may explore the potential of creating scenarios that are more relevant to forensic science casework.

Nevertheless, this study explored important and complex questions and pave the way for future experimental studies and data generation to gain deeper understanding of human factors relating to triaging; thereby, contributing to the development of interventions or strategies for more transparent and consistent early forensic decisions.

5. Conclusion

This is the first study in forensic science that experimentally studied two human factors pertaining to forensic triaging: casework pressures and ambiguity aversion. While the pressure manipulation for this study was found to be effective in inducing feelings of pressure, it did not have a practical influence on triaging decisions. When considering the lack of reliability in triaging, the findings may reflect a foundational issue of inconsistency [32]. The findings also suggest that ambiguity aversion is an important factor to consider, as it can play a role in early decisions, like making a decisive or inconclusive impression about a case. Early on decisions can “cascade” and influence subsequent decisions in the forensic science process [44,76]. This study is a key step forward in understanding and improving triaging decisions in forensic science.