Forensic image comparison, wherein questioned images are compared to known images to determine similarities and differences in image content, has been the subject of judicial commentary in courts in Canada, the US, and the UK for several years. Some courts have endorsed this method of analysis, while other courts have restricted it or disallowed it. For details on some of these earlier decisions, see articles previously posted on this website.
A recent decision of the Supreme Court of Minnesota is worthy of note, not because it is ground-breaking but because it is not that common that the topic of image comparison makes it to a top appellate court. In State v. Hassan, 977 N.W.2d 633, 2022 WL 2709444, the Court was asked to determine whether the state had produced sufficient evidence at trial to merit a conviction for first degree murder. In doing so, the Court considered forensic image comparison evidence that was led by the state.
Facts
Two people fired numerous rounds into a parked motor vehicle, killing one occupant and injuring two others. Both alleged shooters were co-indicted and one pled guilty to second degree murder on the first day of trial. Hassan was thereafter tried alone and was alleged to be the second shooter. CCTV from two nearby business establishments captured the shooting as well as images of the shooters. There had been a separate shooting earlier that evening at a mall and the victim of that shooting was taken to the hospital. It was suspected that the second shooting was in retaliation for the mall shooting. Gang activity was allegedly involved. Several people associated with the first shooting victim arrived at the hospital and were recorded on CCTV and police body camera recordings. One of the visitors was Hassan, who identified himself to the police at the hospital as the cousin of the mall shooting victim.
The jury was shown video of Hassan arriving at the hospital and him being consoled by several other visitors. A police officer pointed out similarities between the clothing worn by Hassan at the hospital and that of the unidentified second shooter at the earlier vehicle shooting. A forensic video analyst testified regarding the analyst’s extensive examination of the questioned images of the vehicle shooting (questioned images) and the hospital images (known images). The analyst’s conclusion was that the hospital images of Hassan were consistent with the images of the second shooter at the vehicle shooting scene. The analyst noted that certain features of Hassan’s clothing that were observable in the hospital CCTV (pattern of jacket, small white Nike logo on pants, and a possible design on shoes) were not visible in the questioned images. The analyst testified that due to resolution and lighting issues, it would not be expected that those type of details would be visible even if present. Finally, the analyst noted that the decreased image resolution at the vehicle shooting scene caused the analyst to be unable to confirm or eliminate Hassan as the second shooter.
The Ruling
There was significant circumstantial evidence that supported the state’s theory that Hassan was the second shooter. Collectively, the Court was satisfied that the conviction was sound. While the forensic image comparison evidence was not pivotal to this conclusion, it was one of thirteen pieces of evidence that the Court relied upon in upholding the conviction. The Court accepted that there was consistency in the clothing worn by Hassan at the hospital and the second shooter at the vehicle scene, thereby adopting the analyst’s opinion.
Analysis
The Supreme Court did not undertake an evaluation of the soundness of forensic image comparison or the analyst’s methodology or conclusions. An opposing analyst was not called by the defense and the Court accepted the comparison evidence largely uncriticized. No legal analysis on this point was undertaken by the Court. Bearing in mind this contextualization, this decision, while not robust, is helpful because it represents a superficial endorsement of the forensic image comparison process. However, it raises unstated issues that are worthy of discussion.
The evidence of the analyst is reflective of a status quo that I believe is less than ideal. Assuming the Court correctly summarized the evidence of the analyst, several issues are readily apparent. First, the analyst stated that the clothing worn by Hassan at the hospital was consistent with the clothing worn by the second shooter at the vehicle scene. This statement was in accordance with the analyst’s training and so my criticism is not of this analyst. However, I believe that this approach lacks scientific and legal rigor and should no longer be used. In image comparison, analysts evaluate and interpret images and point out similarities and differences. That is as it should be. However, I believe it to be inappropriate to then take the next step and pronounce the presence of consistency. That is an argument to be made by counsel and not the proper subject of an analyst’s opinion. The problem with the word consistent is that it is a word that often conveys more certainty than an analyst may intend or can support. What a jury hears when an analyst says consistent is that they are the same or that they match. A more informative approach is to use a levels of support approach wherein the analyst evaluates the similarities and differences along a continuum and then assesses where on that continuum the comparison best falls. An opinion that “there is moderate support for the proposition that the questioned clothing matches the known clothing” is more informative and accurate than to say they are consistent. There are different levels of support continuums used in forensic science and a search will give readers some examples of those that are in current use.
I had occasion to discuss this topic with the Forensic Science Regulator in the UK in 2019 and she was adamantly opposed to UK analysts using the phrase consistent for the reasons stated above. It follows that I support that view. Analysts should strive to be informative and not overdrive the evidence under consideration. A levels of support approach, which is also not perfect and has its own critics, offers a more rational and nuanced approach to image comparison that the older consistent approach that was taught in North America up until more recently.
A second aspect of the analyst’s opinion that bears mention, assuming the Court summarized it correctly, is that the analyst was unable to confirm or eliminate Hassan as the second shooter due to poor image resolution. This statement raises two issues that I find troubling. If the analyst can neither confirm nor eliminate Hassan as being the second shooter, then there is an insufficient basis upon which to pronounce that the images are consistent. Though there is a basis for both statements to stand together, these opinions are arguably in conflict. Counsel should have challenged the analyst on this point. Additionally, given that the analyst said that due to resolution and lighting issues, the analyst would not expect these specified clothing and footwear details to be visible even if present, the analyst must also acknowledge that another reason they were not visible is because they were not there, irrespective of resolution issues. The option that they are not the same clothing and footwear must also be highlighted. The analyst may well have said this – I can only go by what the Court said of the analyst’s evidence.
A third and final point that emanates from this case is the incorrect language that is often used by analysts, counsel, and the court. This comparison was only about the clothing worn by Hassan at the hospital and the clothing worn by the second shooter. It was never about the person wearing the clothing. It is therefore imprecise and misleading for anyone to use image comparison in a case like this to make evaluative statements about the person wearing the clothing based only on the analyst’s clothing opinion. It is tempting to conflate the clothing with the person, but it is scientifically and legally wrong to do so. It is for this reason that when giving evidence about text messages, emails, and internet searches, a digital forensics expert rarely knows who is doing the typing and therefore comments only on the generic “person” who did the typing, without attempting to identify them from the typing alone. This is an important point because when questioning forensic video analysis students in moot court settings, I frequently point out that they have said nothing about the person wearing the clothing and so it is inappropriate to offer an opinion as to whether the defendant is the person observed in the questioned images. It is easy to be lulled into this false identification narrative and analysts must guard against conflating clothing with the person wearing it if they are not also evaluating the person wearing the clothing.
Forensic image comparison evidence has been an important component of forensic video analysis for many years. Training in this area is evolving but issues remain, as noted in this article. If these issues are not addressed, attorneys who are alive to these issues may well raise them in court and this could result in judicial decisions that criticize and even reject expert opinion evidence on the basis that it is not sound. That would have a decidedly unpleasant rippling effect on image comparison evidence.