Setting the Threshold for the Admissibility of Computer Generated Evidence — A Brief Review of Recent US and Canadian Decisions

© 2004, Deeth Williams Wall LLP. All Rights Reserved. By: Colin Adams, Student at Law (July 8, 2004)

Questions regarding the admissibility of computer generated evidence have re-emerged following two recent cases in the US and Canada. In a lengthy analysis, the Connecticut Supreme Court set down a detailed six-part test for the admission of computer generated evidence. In contrast, an Ontario decision discussed the criteria for the admission of computer generated evidence in broad, generalized terms. An earlier Canadian decision from the Saskatchewan Court of Appeal made a number of observations similar to those of the Connecticut case. In reviewing these cases, the divergent approaches raise admissibility and reliability concerns, which future Canadian courts may need to consider when reviewing the admissibility of computer generated evidence.

State v. Swinton – Connecticut Supreme Court ¹

In Swinton, the Connecticut Supreme Court was charged with the task of ruling on the defendant's appeal after he was found guilty of murder in March 2001. In upholding the conviction, the Court spent the majority of the decision considering the admissibility of two pieces of computer generated evidence, which were crucial to the State's case. The first piece of evidence considered enhanced photos of bite marks using software entitled, "Lucis". The second piece involved images of the defendant's teeth overlaid upon photos of the bite mark using Adobe's Photoshop software. In appealing the application of these pieces of evidence, the defendant argued that the State failed to present foundation testimony on the adequacy of these two programs for the task of matching his dentition with the bite marks on the victim given that the computer enhanced and computer generated exhibits were introduced by experts with only an elementary familiarity with the programs.

To resolve the issue of establishing the necessary foundation for the presentation of computer generated evidence at trial, the Court developed a new test. The Court acknowledged that the old test in the Connecticut case of American Oil Co. v. Valenti, which required "testimony by a person with some degree of computer expertise, who has sufficient knowledge to be examined and cross-examined about the functioning of a computer", was of limited guidance given that the reliability of the evidence could not be extrinsically established in this case. As such, the Court created a new six-part test:

In establishing this revised test, the Court was cognizant that computer generated evidence was becoming more commonplace and expanding on a daily basis. To that end, it was important to develop a standard that is applicable today and would not become useless tomorrow. The Court also stated that it disagreed with commentary which suggested that the application of an identical standard to all types of computer generated evidence would cause mischief. Instead, the Court saw no harm in applying a single standard.

Guided by the principles of the new test, the Court ruled that the computer enhanced photos of the first expert were properly admitted whereas the computer generated overlays of the second expert were not. The Court found that the computer enhanced photos submitted and created by the first expert were developed using reliable software, proper procedures, and standard equipment in the field. In addition, although the first expert could not testify on the algorithms of the program to illustrate its proper operation, the Court determined that the first expert had sufficient knowledge and expertise with the program and in the creation of the evidence to be considered a qualified operator of the software. In contrast, the reliability of the computer generated overlays were undermined by its very genesis. The Court seized on the fact that the second expert testifying did not personally perform the computer manipulations to generate the evidence. Instead, the second expert testified that he only watched a colleague, with a deep knowledge of Photoshop, perform the computer manipulations. This factor proved to be the undoing for the second expert's evidence. The Court went on to rule that the second expert could not confirm whether the computer processes used to create the overlays were accepted in the field of odontology as standard, whether proper procedures were followed in connection with input and output of the information, whether Photoshop was reliable for this sort of forensic application or whether the equipment was programmed and operated correctly.

The decision to exclude the computer generated overlays may suggest that the Court's initial interpretation of the test may be unduly restrictive. Essentially, once the Court determined that the second expert did not perform the computer manipulations, the evidence was thrown out entirely. Based on the Court's interpretation, it would appear that only the expert testifying can prepare and deliver the computer generated evidence, otherwise the evidence runs the risk of being jettisoned on the basis of its assumed unreliability. Seen in that light, the test seems to require computer expertise on the part of the expert, which may not always be realistic. Moreover, it also suggests that a supervisory role in the preparation of computer generated evidence is not sufficient to determine if proper procedures and a reliable program were used, the equipment was programmed and operated correctly or the resulting evidence is properly identified as the output in question. Further, there may be a danger that the Court's test will stifle innovation. The Court's focus on accepting the status quo appears to potentially threaten new and emerging methods for the generation of computer evidence. New and better methods might fall victim to the test simply because the procedures or software programs are not standardized. Consequently, the test may actually inhibit the search for truth by disqualifying potentially useful or helpful evidence.

R. v. Jamieson – Ontario Superior Court of Justice²

In R. v. Jamieson, the Ontario Superior Court of Justice was faced with the particularly disturbing issue of an alleged physical assault of a child with significant medical problems by the nurse who was hired to care for the child. In order to determine the source of a string of bruises and fractures inflicted on their child's body, the parents purchased and installed a "nanny" camera, which was hidden from view and focused on the child's crib. When the videotape was reviewed by the child's father the following day, he viewed what he considered to be an assault on his child and called the police, along with the nurse's employer. Due to the darkness of the images on the original videotape, an audio-video technician made a copy. The contrast and brightness of the copy were adjusted with the aid of computer. In addition, a time scale was added to the computer enhanced copy of the videotape. The nurse was charged with aggravated assault, contrary to s. 268(1) of the Criminal Code. In an evidentiary ruling released on April 30, 2004, the defendant applied to the Court for an order seeking the exclusion of videotapes, both original and enhanced, and any evidence of experts based upon the contents of the videotapes. The defendant's request was rejected and the videotape evidence was allowed.

In considering the admissibility of the computer enhanced evidence, the Court canvassed the issue with broad brush strokes to outline general criteria. In setting the backdrop for the analysis, the Court focused on the use of videotape, and technology in general, in the evidentiary process. Referring to the Supreme Court of Canada decision in R. v. Nikolovski, Cory J. noted that, "[T]he admission of new types of evidence is often resisted at first and yet, later accepted as commonplace and essential to the task of truth finding." ³ In another Supreme Court decision referred to by the Court, L'Heureux-Dube J. in R. v. L. (D.O.) had stated that "[I]t would seem contrary to the judgments of our Court to disallow evidence available through technological advances such as videotaping, that may benefit the truth seeking process." ⁴ With the background established, the Court then turned its attention to the question of admissibility. The Court first addressed the admissibility of normal videotapes. Returning to Cory J. in Nikolovski, it was noted that, "[O]nce it has been established that a videotape has not been altered or changed, and that it depicts the scene of a crime, then it becomes admissible and relevant evidence." With respect to "enhanced" videotapes, such as the one in Jamieson, the Court sought the guidance of Trafford J. in R. v. Brown. ⁵ In that case, the enhancements of the videotape consisted of adjusting the brightness and contrast as well as the assistance of a computer to adjust the pixel values of the facial features of the defendants.

The videotapes are admissible if they accurately and fairly present the information they purport to convey...If a change in the image is not so substantial as to be misleading, it may be admissible in the discretion of the trial judge. Exclusion is the proper result if the image is distorted in a material particular. The test of admissibility may be compendiously described as one of substantial accuracy. In this case, the Crown must – prove on a balance of probabilities – the substantial accuracy of the original tape and the enhancements to it...

Taken collectively, the cases noted in Jamieson suggest that new technology and its use should not be feared, but rather, encouraged and employed in the evidentiary process when it aids in the search for the truth. Consequently, Courts may admit computer generated evidence where it is substantially accurate. The evidence must either remain unaltered or, if altered, the enhancements cannot be so substantial as to be misleading. In Jamieson, the Court felt that the computerized enhancements to adjust the brightness and the addition of a time scale did not undermine the accurate presentation of events, and were therefore admissible.

In comparison to the test in Swinton, the broad criteria for admissibility espoused in Jamieson may be too general. In Swinton, the Court recognized that computer generated evidence, through a want of skill, inadequate instruments or materials, or through intentional or skillful manipulation, may not only be inaccurate, it may also be dangerously misleading. The extent of this important conceptual underpinning in Jamieson is not entirely clear. Although the Court does appear to be aware of the problem of digital manipulation insofar as the relationship between the original data and any computer generated enhancements must be shown, it is silent on the process of demonstrating, or the requirements that may be necessary, to determine the accuracy of the enhancements. For instance, what would happen if the computer generated enhancements provide additional unexpected evidence or opinion evidence as in Swinton? In cases such as those, can it really be said that the computer generated evidence accurately and fairly presents the evidence it purports to convey? To be fair, the Court may not have had these questions in mind given that the evidence was videotape based and not derived from a traditional computer context as in Swinton. Nevertheless, this may be an important consideration for Canadian law in the coming years, particularly if computers continue to be used in increasingly sophisticated ways for evidentiary purposes.

University of Regina v. Pettick – Saskatchewan Court of Appeal ⁶

Pettick is an earlier Canadian decision from the Saskatchewan Court of Appeal discussing the admissibility of computer generated evidence. Although the case dates from 1991, the decision is noteworthy for a couple of reasons. First, unlike Jamieson, Pettick is Canadian precedent that deals more specifically with traditional computer generated evidence, as opposed to videotape based evidence, and offers a more specific discussion of the criteria for the admission of computer generated evidence similar to Swinton. Second, unlike Swinton, the criteria discussed in Pettick are general enough to allow for a measure of flexibility in the assessment of computer generated evidence.

In Pettick, the Saskatchewan Court of Appeal considered liability issues arising from the construction of a gymnasium roof which began to crack. In doing so, one of the issues considered was whether an expert's opinion on the status of the roof should have been accepted since it was based on a computer generated report which was neither prepared nor understood by the expert delivering the testimony. In setting down the dissenting opinion, MacLeod J. discussed the fundamental elements for the admissibility of computer generated evidence. Surprisingly, these factors were not discussed in the majority's decision.

MacLeod J's discussion of the general factors for the admission of computer generated evidence in Pettick resonate well with the test expressed in Swinton. As in Swinton, proof of reliability is the key to the admission of computer generated evidence. MacLeod J. noted that,

[T]here must be some evidence of the configuration of the program, how it produces its results, the indicia of care to be used in the data to the computer and the testing required to determine whether the program and computer effectively produces the desired and reliable result. It may be necessary to know who wrote the program, what it was written to do or accomplish, and what assumptions were made, if any, therein. We should know about the margin of error, if any, which may exist in the results generated. In short, it is necessary to demonstrate what the program was intended to do, and how it does it, and that it has the scope and capacity to do just that. This is quite apart from the question of whether raw data was properly assembled and properly fed into the computer.

In making these statements, MacLeod J.'s concerns track a number of factors set out in the test in Swinton. The statements illustrate the need for proving that the software is reliable; in good working order; the equipment operated and was programmed correctly; proper procedures were followed for the input and output of data; and a reliable result was generated.

The similarity between Pettick and Swinton becomes less clear when attempting to discern the value placed on the use of standardized computer equipment. As noted in Swinton, this is clearly a key element for admissibility. In Pettick, however, the importance of this element is unclear. In making reference to the issue, MacLeod J. noted that the computer program used by the expert to generate the mathematical calculations was "not a common program of which judicial notice could be taken." It is unclear what "common" means in this context. To borrow from the Swinton analysis, does "common" mean that the Court would potentially exclude the computer generated evidence because the program and/or computer was not considered to be standard equipment in the field? The answer to this question is uncertain. One wonders if this factor will even be a consideration for a future Canadian court or, if it is, the amount of weight it is given in determining the reliability of the computer generated evidence. If the factor is given weight, it will be important to balance reliability with flexibility to ensure that innovation does not become an impediment to the truth.

Despite the similarities, Pettick does differ from Swinton in at least two important respects. First, MacLeod J. immediately made a distinction between different types of computer evidence and the need for different evidentiary thresholds to establish reliability. On the one hand, there are computer records, which are taken from computer-stored information sources. On the other, there is information that is actively processed and generated as opposed to being simply retrieved by a computer, such as animations or simulations. In distinguishing between these computer generated evidentiary types, MacLeod J. was clearly opposed to a single standard and suggested that a higher standard of admissibility would be required for computer generated information than computer generated records. MacLeod J.'s distinction and analysis is important, particularly where the computer is generating what might be considered as opinion evidence.

Second, it does not appear that the individual testifying and the individual generating the computer evidence need be the same person. Although the Pettick Court did grant that it is better to produce a witness who performed the computer operations, the threshold for admission appears to be a greater or lesser degree of first hand knowledge, depending on the type of computer evidence in question. For example, when dealing with computer records, MacLeod J. suggested that a witness with a mix of first hand and hearsay knowledge of the facts proving the content of the records may be acceptable depending on the circumstances of the case. In the case of computer generated evidence such as computer simulations, the Court suggested a higher threshold, in that the testimony of an expert be required. However, nowhere did the Court indicate that the expert testifying must be the same person who performed the computer operations that generated the evidence. The Court failed to indicate the degree of first hand knowledge that the expert testifying should possess to prove the content of the evidence. Yet, some knowledge is clearly required based on the disposition of this particular issue. The Court refused to accept the evidence submitted by the expert who failed to participate in the generation of, or even understand, the computer generated report he presented. In the words of MacLeod J., without the testimony of the expert who prepared the computer generated report, the evidence could not be admitted as it "was that of the computer, without any way to test its reliability." In seemingly choosing to forgo the need for the expert to both deliver and prepare the evidence, the Court's assessment for the admissibility of the computer generated evidence was given a measure of flexibility that is missing in Swinton.

Conclusion

The divergent criteria for the admissibility of computer generated evidence in Swinton and Jamieson raise important evidentiary issues. In Swinton, the Court attempted to set down a single standard for the admission of computer generated evidence. Although the specificity of the test does potentially provide courts with a reasonable degree of guidance in approaching the issue, it could be argued that Swinton's interpretation of the test may be too restrictive and actually hinder the search for the truth. In Jamieson, the broad criteria considered by the Court allows for a great deal of flexibility, but it is possible that the test may need to fleshed out further to develop the intermediary steps to trace the evolution from raw data to computer generated evidence. Perhaps, Canadian courts will look towards developing a middle ground that echoes the approach in Pettick, which discussed more general criteria for computer generated evidence, balanced with a sufficient degree of flexibility to allow for the acceptance of innovative evidentiary tools and approaches. In the interim, it will be interesting to see if the Canadian approach to the admission of computer generated evidence will change as a result of Swinton or Jamieson or if Canadian courts look to Pettick for guidance. Regardless, as computers become more and more ensconced in the evidentiary process, developing standards for assessing the admissibility of computer generated evidence will become increasingly important.

Endnotes

Contact James Kosa for additional information on Admissibility of Computer Generated Evidence.