COURT FILE NO.: CR-20-145
DATE: 20220511
ONTARIO
SUPERIOR COURT OF JUSTICE
B E T W E E N:
Her Majesty the Queen
Janet Booy and Mark Dean, for the Applicant
Applicant
- and -
Richard Taylor
Jennifer Penman and Kristen Dulysh, for the Respondent
Respondent
HEARD: May 6, 9, 2022
RULING – HEALTH APP EXPERT EVIDENCE
The Honourable Mr. Justice Skarica
Overview
[1] It is alleged that at approximately 3:30am on July 9, 2018 Richard Taylor entered the residence of his mother and step father and set their bedroom ablaze. The fire killed the accused’s mother who was pronounced dead at 4:10am and severely burned the accused’s step father who died later that day. The step father made several dying declarations that “Rich”, “Rick” was responsible.
[2] The accused’s phone was seized and analyzed by the police. The analysis revealed a “Health App” on the phone recorded that on July 9, 2018 from 4:04am to 4:41am the user took 1806 steps for a total of 1075.67 meters. The Crown submission is that Mr. Taylor was pacing around his home after returning from the scene of the murder.
Issue
[3] Is expert opinion evidence regarding the steps and distance, obtained from the phone, admissible at Mr. Taylor’s trial?
Facts
[4] A fire was set at about 3:30am on Monday, July 9, 2018 in the bedroom of 8 Greening Court.
[5] As a result of the fire, Carla Rutherford (Carla) was pronounced dead at 4:10am on July 9, 2018. Her husband Alan Rutherford (Alan), briefly survived but was badly burnt. Alan told a variety of witnesses at the scene that “Rich”, “Rick”, his son-in-law and his wife’s son was responsible.
[6] The accused Richard Taylor is referred to as “Rich” and is the son of Carla Rutherford. Alan Rutherford died later that day at 3:01pm.
[7] The accused Richard Taylor’s phone was seized and analyzed by the police. The phone contained a “Health App”. These Apps are designed to record steps taken and distance travelled by the user.
[8] An analysis of the data revealed that the user of the phone on July 9, 2018 – the morning of the fire – took 1806 steps for a total of 1075.67 meters between 4:04am to 4:47am.
[9] The Crown contends that this evidence, which would have to given by an expert in phone data analysis, is admissible and supports an inference that the accused was pacing at his home shortly after the fire and accordingly, this is circumstantial evidence of guilt.
[10] The defence position is that the evidence is not admissible due to the underlying unreliable science and technology of the health data application.
[11] In the voir dire, Staff Sergeant David McKenzie was qualified to give expert evidence regarding Mobile Phone Forensics.
[12] Office McKenzie testified that the Health App takes data from sensors in the phone and converts it into tables.
[13] Office McKenzie testified that data obtained from an iPhone only provides data the phone allows you to obtain.
[14] Crime Analyst Jovan Krasulja prepared two reports – see Exhibits 12 and 13 and also see Tab 2 of the Crown Application materials – that indicate no health data appears on the phone between midnight and 4:00am on July 9, 2018. Between 4:04am to 4:47am the user of the iPhone 6 was documented taking 1806 steps for a total of 1075.67 meters. The 1806 steps that morning comprises 62% of the steps taken for the 4:00am hour between April 16, 2016 to July 9, 2018.
[15] The cross-examination of Mr. McKenzie revealed a number of concerns regarding data obtained from the Health App, including the following:
How the sensors on the phone actually work is unknown because that is proprietary information.
The Health App was created for personal interest and was not designed for court use.
The most up to date study regarding the Health App is from the Netherlands done by authors Jan Peter Van Zandwijk and Abdul Boztas in 2019 (Netherlands study). The Netherlands study was done for iPhones 6, 7, and 8 with five subjects. The study revealed that measured steps had an average error rate of 2 percent. However the study indicated, “the reliability of the registered distances, however, depends on a number of factors including walking speed and walking style of the subject and can deviate up to 30-40% from the true value.”
You can get a false positive as a step i.e. going over a speed bump can be registered as a step.
The Netherlands study indicates, that at page S132, that given this false positive situation, “we conclude that it is possible to use data from the Health App for forensic purposes in cases where it is known (or assumed) that walking or running has taken place.” The report indicated that converse need not be true, concluding that under specific conditions there might be false positive registrations in the Health App. The study indicates, as an example, that sometimes steps can be registered when driving a car over speed bumps.
The Netherlands study, in the introduction, points out the Health App relies on sensors such as accelerometers and gyroscopes. The manufacturers are not explicit about the way these signals are processed which “makes it hard to independently judge the accuracy of health information derived from this sensor information.”
Law
[16] The Supreme Court of Canada outlines the current legal framework for expert opinion evidence in the leading case of White Burgess Langille Inman v. Abbott and Haliburton Co., 2015 SCC 23, [2015] S.C.J. No. 23.
[17] Justice Cromwell, for the court, indicated at paras. 22-24:
22 Abbey (ONCA) introduced helpful analytical clarity by dividing the inquiry into two steps. With minor adjustments, I would adopt that approach.
23 At the first step, the proponent of the evidence must establish the threshold requirements of admissibility. These are the four Mohan factors (relevance, necessity, absence of an exclusionary rule and a properly qualified expert) and in addition, in the case of an opinion based on novel or contested science or science used for a novel purpose, the reliability of the underlying science for that purpose: J.-L.J., at paras. 33, 35-36 and 47; Trochym, at para. 27; Lederman, Bryant and Fuerst, at pp. 788-89 and 800-801. Relevance at this threshold stage refers to logical relevance: Abbey (ONCA), at para. 82; J.-L.J., at para. 47. Evidence that does not meet these threshold requirements should be excluded. Note that I would retain necessity as a threshold requirement: D.D., at para. 57; see D. M. Paciocco and L. Stuesser, The Law of Evidence (7th ed. 2015), at pp. 209-10; R. v. Boswell, 2011 ONCA 283, 85 C.R. (6th) 290, at para. 13; R. v. C. (M.), 2014 ONCA 611, 13 C.R. (7th) 396, at para. 72.
24 At the second discretionary gatekeeping step, the judge balances the potential risks and benefits of admitting the evidence in order to decide whether the potential benefits justify the risks. The required balancing exercise has been described in various ways. In Mohan, Sopinka J. spoke of the “reliability versus effect factor” (p. 21), while in J.-L.J., Binnie J. spoke about “relevance, reliability and necessity” being “measured against the counterweights of consumption of time, prejudice and confusion”: para. 47. Doherty J.A. summed it up well in Abbey, stating that the “trial judge must decide whether expert evidence that meets the preconditions to admissibility is sufficiently beneficial to the trial process to warrant its admission despite the potential harm to the trial process that may flow from the admission of the expert evidence”: para. 76.
Analysis
[18] It is clear that the analysis of Health App data is relatively new science. There appears to be very few scientific articles dealing with the reliabilities of Health App data.
[19] Both Crown witnesses – Staff Sergeant McKenzie and Crime Analyst Jovan Krasulja testified that the Netherlands study in 2019 is the latest up to date study that analyzes the accuracy of step and distance data. It is noted that that the Netherlands study is not exhaustive as it deals with just 5 subjects – a very small sample.
[20] The Netherlands study raises the following concerns relevant to the assessment of whether the Health data is reliable. These concerns include:
How the sensors on the phone actually work is not known. According to the study, “manufacturers are not very explicit about the way these sensor signals are processed which makes it hard to independently judge the accuracy of health information derived from the sensor information.”
The Health App was created to collect data on daily activities for health purposes. While the Health App obviously has forensic applications if the data is accurate and reliable, the Health App was not designed for court use.
The results of the Netherlands study found that the steps registered on the iPhone agree very closely with those measured manually with an average error of 2 percent. The reliability of registered distances however depends on a number of factors including walking speed and walking style and can deviate up to 30-40% of the true value.
There are specific conditions where the Health App can create false positives. An example given is that initial research indicates that sometimes steps can be registered when driving in a car over speed bumps.
Most relevant to this case is the following comment made in the Netherlands study at page S132: “we conclude that it is possible to use this data from the Health App for forensic purposes where it is known (or assumed) that walking or running has taken place. It is important to note that the converse need not be always be necessarily true; if steps or distances are registered by the Health App, this need not necessarily indicate that walking or running has taken place while carrying the telephone, since under specific conditions there might be false positive registrations in the Health App.”
[21] The accused was confronted with the Health App data, by Detective Ashbaugh at pages 180-182 of the accused’s police statement #3. The accused confirms that he went to bed after 3am. The accused indicates he didn’t leave the house and repeatedly indicates he doesn’t know how that data came to be on his phone. The accused speculates his son could have picked up the phone to play games but then indicates that his son probably didn’t do that.
[22] In the result, there is no evidence before me that the accused was walking or running with the phone in his physical possession between 4am and 5am on July 9, 2018.
[23] Given (1) the conclusion of the Netherlands study that the Health App data can be used for forensic purposes in cases where it is known or assumed that walking or running has taken place and (2) there is no evidence before me that the accused was walking or running with his phone at the material time, (3) combined with the other concerns outlined above, I am not satisfied of the reliability of the data, given the current state of the underlying science supporting the use of Health App, in the circumstances of this particular case – see White Burgess at para. 23.
[24] I wish to note as well that two of the most infamous cases regarding wrongful conviction in Canadian history also resulted from expert scientific opinion that was later discredited and found to be unreliable. In R. v. Truscott, 2007 ONCA 575, the Court of Appeal indicated that Mr. Truscott was convicted in the 1959 murder of a classmate. The Crown’s case at trial depended largely on the time of death evidence given by a forensic pathologist.
[25] Fresh evidence introduced decades later by experts in the field of pathology and gastroenterology convinced the Ontario Court of Appeal that the time of death testimony given at the accused’s original trial was not scientifically justified. Mr. Truscott’s murder conviction was set aside and an acquittal was entered.
[26] Similarly, the Morin Commission Report enquiring into the wrongful murder conviction of Guy Paul Morin, indicates in Chapter II of its executive summary, that the hair and fiber evidence of two forensic scientists, given at the instance of the prosecution, ultimately had little or no probative value in proving Mr. Morin’s guilt. However, the Commission found that the contribution of the CFS (Centre of Forensic Sciences) to Mr. Morin’s wrongful arrest, prosecution and conviction was substantial. The assistant Deputy Solicitor General Dr. James Young apologized on behalf of the CFS and corrective measures were undertaken.
[27] The history of the Truscott and Morin cases underline the wisdom of the Supreme Court of Canada in White Burgess in requiring a trial judge to scrutinize carefully the reliability of the science underlying expert opinion evidence.
Conclusion
[28] For the reasons outlined above, I find that, in the particular factual circumstances of this case, and the current state of scientific knowledge, the data obtained from the Health App is not reliable.
[29] I find that the prejudicial effect of the Health Care App evidence far exceeds any probative value, see R. v. Abbey, 2009 ONCA 624 at paras. 87-94.
[30] Accordingly, the Health Care data is not admissible at this trial. Further, the references to the Health Care App data outlined at pages 180-182 of the accused’s Statement #3 will need to be edited out of the statement.
Order
[31] The Crown’s application to introduce expert evidence regarding the data obtained from the “Health App” of Mr. Taylor’s cellphone is dismissed. This evidence is not admissible at Mr. Taylor’s trial.
Skarica J.
Released: May 11, 2022
COURT FILE NO.: CR-20-145
DATE: 20220511
ONTARIO
SUPERIOR COURT OF JUSTICE
B E T W E E N:
Her Majesty the Queen
Applicant
- and –
Richard Taylor
Respondent
RULING
Released: May 11, 2022