That was it, right? – ‘very likely’ in percentage terms
Given the legal questions we often face, especially regarding the quantification of “very likely” in percentages, I would like to explain why I, as an expert, have decided to avoid such numerical assessments. The probability of identity or non-identity is deliberately formulated as a verbal predicate because biostatistical calculations are problematic. There is a lack of sufficiently extensive and validated databases that consider various ethnic origins and age groups. Moreover, the reference images were not created under standardized conditions. Therefore, the “actual” expression of features is often not discernible, and calculations based on the “apparent” expression could lead to erroneous conclusions.
Human cognition struggles with the interpretation of probabilities, as our minds often cannot objectively assess them. A probability of 99% seems high, but the question arises whether this is sufficient for a conviction when there is little other evidence. Judges and jurors are often overwhelmed in such cases, as they are not experts in probabilities. Thus, percentages in forensic reports can indeed be misleading. The examples provided illustrate that we should be cautious when interpreting probabilities.
Consider, for instance, the passenger traffic at a large U.S. airport with approximately 2.5 million passengers per day. If the probability of a serious accident is 99.9%, this means there’s a 0.1% chance an accident will occur. With 2.5 million passengers, this could potentially mean 2,500 accidents, which is significant. In another example, an accused might be deemed the perpetrator with a 95% probability according to an expert report. This also implies a 5% probability that he is not. Additional evidence like found clothing or vehicle access could increase the probability. However, this aspect falls within the court’s jurisdiction, which considers further factors to make an informed decision. Experts should never be tempted to excessively increase the probability based on additional circumstantial evidence and cognitive intentions.
Ultimately, experts and courts should be cautious in interpreting probabilities and consider all available information to gain a comprehensive picture of the situation. Experts must critically reflect and minimize cognitive biases to ensure objective probability assessments.
Moving to the identification of individuals from images: The frequency of a feature typically relies on experiential data and is assessed cautiously, in favor of the person in question if in doubt. Depending on the frequency of a feature within the population, it is classified as less, moderately, or very characteristic. Particularly striking features like lines, scars, and clothing are referred to as individual characteristics.
Schwarzfischer suggests nine predicate classes for the outcome of a photo analysis, from “identity practically proven” (>99.72%) to “non-identity practically proven” (<0.28%). Intermediate predicates are possible and should be explained in more detail. Mathematical frequency statements about feature expression are difficult because corresponding studies are lacking. It’s important that not the single feature, but the sum of features determines the identification rate. In individuals with a foreign ethnic background, many features may have different frequency distributions.
In evaluation, it’s crucial to closely examine marginal differences, especially if a search with the perpetrator’s image was conducted prior to the procedure, also known as the principle of pre-selection. Suspects resemble the perpetrator’s image, so it’s essential to look for exclusion criteria.
Limitations of the expert report include providing comparison photos that authentically represent the suspect. If relatives are alternative subjects, measures like personal presentations or current comparison photos are necessary. Distinguishing between monozygotic twins can be difficult, whereas dizygotic twins do not show greater similarity than other siblings.
In image forensics, various sources of error and cognitive biases that affect evaluation must be considered: image quality, poor resolution, unfavorable lighting conditions, or infrared images. Human perception is susceptible to biases like confirmation bias and anchoring. Misidentified individuals can be based on similar photographs, and manipulations can highlight or obscure features.
To minimize these errors, experts should proceed methodically and critically, be aware of their limitations, and apply various strategies: thorough training, standardized methods, collaboration with experts, awareness of one’s own cognitive biases, and careful documentation. Through these approaches, experts can contribute to the justice of the legal system.
In forensic image identification, scientific precision and methodological accuracy are as important as the human dimension of the legal system. Experts carry immense responsibility, not only through professional expertise but also through ethical stance. Continuous self-reflection and critical engagement with scientific foundations are essential. Modern technologies offer new possibilities, but human review remains indispensable to correct erroneous machine analyses.
The use of modern technologies keeps forensic image identification a human discipline that requires experience and understanding of human complexity. Empathy and sensitivity in dealing with those affected are equally important. A forensic report can have far-reaching consequences from exonerating the innocent to convicting the guilty. This responsibility requires professional competence and moral integrity.
An appeal for scientific rigor: Forensic image identification operates in the tension between science and justice. The scientific method demands objectivity, traceability, and constant self-criticism. In justice, these findings are contextualized with human fates. Experts must make scientifically grounded statements and communicate them understandably.
Technological advancement in forensic image identification promises increased accuracy and efficiency, but the role of the human expert remains indispensable. Technologies support the work, but the final evaluation must remain human to consider ethical implications. The combination of technological progress and human expertise offers a promising future perspective. Through continuous improvement and commitment to scientific integrity, experts significantly contribute to truth-finding and justice in the legal system.
This reflective approach shows that the path to truth leads not only through precise technical analyses but also through an understanding of human complexity and ethical responsibility.