Reliability and accuracy of computer-produced evidence

The issue

6.15 The uniform Evidence Acts contain a number of provisions facilitating proof of electronic evidence.[31] Sections 146–147 facilitate proof of ‘evidence produced by processes, machines and other devices’[32] and are intended, among other things, to facilitate the admission of computer-produced evidence.

6.16 Section 146 of the uniform Evidence Acts creates a rebuttable presumption that, where a party tenders a document or thing that has been produced by a process or device, if the device or process is one that, if properly used, ordinarily produces a particular outcome, then in producing the document or thing on this occasion, the device or process has produced that outcome. For example, it would not be necessary to call evidence to prove that a photocopier normally produced complete copies of documents and that it was working properly when it was used to photocopy the relevant document. Section 147 provides a similar rebuttable presumption in relation to documents produced by processes, machines and other devices in the course of business.

6.17 In DP 69, the Commissions asked whether the uniform Evidence Acts should be amended to impose a more rigorous requirement for the presumption of reliability and accuracy of computer-produced evidence.[33]

6.18 This question arose out of submissions to IP 28 by the Criminal Law Committee of the Law Society of South Australia and the Legal Services Commission of South Australia comparing the provisions of South Australia’s evidence legislation dealing with evidence produced by processes, machines and other devices with s 146 of the uniform Evidence Acts.[34] Both commented that s 45C of the Evidence Act 1929 (SA) seems to be more comprehensive than s 146 of the uniform Evidence Acts ‘in ensuring that a device producing a document is in itself not prone to error’.[35]

6.19 These submissions also pointed out that the uniform Evidence Acts have no direct equivalent of s 59B of the Evidence Act 1929 (SA), which requires a court to be satisfied that there have been no alterations made to the machine, such as tampering with the hard drive of the computer.[36]

The discussion in DP 69

6.20 In DP 69, the Commissions examined the South Australian approach to presumptions as to the accuracy of reproductions.[37] Briefly, s 45C of the Evidence Act 1929 (SA) allows the court to: rely on its own knowledge of the nature and reliability of the processes by which the reproduction was made; rely on the certification of someone with knowledge and experience of these processes or who has compared the contents of both documents and found them to be identical; or act on any other basis it considers appropriate in the circumstances.[38]

6.21 In addition, s 59B of the Evidence Act 1929 (SA) makes ‘computer output’[39] admissible subject to the court being satisfied as to a number of matters relating, broadly, to the proper programming, use and functioning of the computer; correct data entry; and the admissibility of the material from which the data was produced. The Commissions noted that the merit in this approach is that it ‘recognises in a direct way the need to address the issue of whether a computer has operated correctly in producing material that is to be admitted’.[40]

6.22 The Commissions observed that ss 45C and 59B provide alternative approaches to the admissibility of computer-produced evidence that have the outward appeal of being broad and investing the court with wide judicial discretion to admit into evidence photographic, electronic and other reproductions.

6.23 However, the Commissions commented that s 45C is flawed in that it relies entirely on the reliability of the ‘approved process’ without further, or actual, investigation into that process. The Commissions further observed that s 59B is based on the Civil Evidence Act 1968 (UK), which was criticised by the Law Commission of England and Wales in a 1993 review of that Act. The Law Commission observed:

[T]here is a heavy reliance on the need to prove that the document has been produced in the normal course of business and in an uninterrupted course of activity. It is at least questionable whether these requirements provide any real safeguards in relation to the reliability of the hardware or software concerned.[41]

6.24 In DP 69, the Commissions also examined research into the reliability of computers carried out by Dr Cameron Spenceley.[42] Dr Spencely developed an approach to the treatment of computer-produced evidence termed a ‘redundancy test’ approach. This relies on implementing a ‘redundant mechanism’[43] in the environment in which the computer is used to address the problem of reliability of computer output.[44] A ‘redundant mechanism’ does not increase the functional capacity of the computer system itself, but operates to prevent or mitigate unreliability in that system.[45] That is, it operates to provide some level of verification that a failure in the computer has not occurred.

6.25 The test of admissibility for computer-produced output that Spenceley proposes is that the party adducing the evidence should be able to demonstrate that:

(a) some mechanism(s) of redundancy (however formulated and implemented) was or were utilised in connection with the production of particular material in the setting in which it was produced; and that

(b) it is reasonably likely that any error(s) in the operation of that computer that affected the accuracy of information contained in that material would have been detected by such mechanism(s).[46]

6.26 The Commissions considered that a ‘redundancy test’ offers a more rigorous requirement for admissibility of computer-produced material that arguably balances the need to ensure reliability of evidence with the need for an efficient practice for use in litigation.[47] It was noted in DP 69 that relatively simple and cheap verifying measures could bebuilt into the computer environment that could at least mitigate the risks of computer unreliability.[48] Then all that would be required in the courtroom would be to describe the measures that had been put in place in a particular setting.[49]

6.27 The Commissions also canvassed an alternative viewpoint to that which advocates a more rigorous test for admitting computer-produced evidence.[50] Adherents to this view argue that there are significant benefits to be derived from the presumption of reliability and accuracy of computer output, because this facilitates the admissibilty of the numerous documents and business records generated from computer stored information.[51]The argument is that s 59B has not made it easy to have computer-produced documents admitted into evidence.[52]

6.28 In DP 69, the Commissions note that in the few cases in which the section has been considered, the South Australian courts have held that the conditions of s 59B were not complied with in at least three cases.[53] Reference was made to Emmanuel Laryea’s argument that ss 146 and 147 eliminate the problem arising under s 59B of computer evidence being rejected where there is no apparent system malfunction.[54] Even so, Laryea concludes:

It must be ensured … that adequate safeguards for testing computer evidence are put in place. Courts should be given, and use, wide powers to ensure that computer systems and electronic data are sufficiently tested for integrity and reliability when necessary.[55]

6.29 In DP 69, the Commissions observed that the case law dealing with ss 146 and 147 of the uniform Evidence Acts has not indicated that there are any problems with the operation of these provisions.

Submissions to DP 69

6.30 Of the seven submissions addressing this question, three supported a more rigorous test[56] and four opposed it.[57]

Submissions opposed to a more rigorous test

6.31 The Office of the Director of Public Prosecutions (NSW) (NSW DPP) does not support a higher threshold for admissibility of computer-produced documents for a number of reasons.[58] These can be summarised as follows:

  • A more rigorous test is not justified; there is no solid evidence that such a provision is needed and no cases of wrongful conviction from computer-generated error.

  • Litigation in Australia depends on an adversarial system and the burden of proof that rests on the prosecuting party, or plaintiff, ensures proper testing of evidence of this sort.

  • It would impose a higher threshold than for other ‘machine produced evidence’.

  • Data manipulation can occur with any machine-generated information, such as photos, tapes and videos.

  • The party challenging the accuracy of the evidence would have to be given the opportunity to inspect the relevant computer and perform their own tests—a costly, and time-consuming exercise.

6.32 The NSW DPP is particularly opposed to the ‘redundancy test’ approach for reasons that include the following:

  • The meaning of the term ‘redundancy mechanism’ is not readily understood.

  • The ‘redundancy test’ is set at the civil standard of proof (it gives a basis for inferring that computer-produced material will be more likely than not to aid the identification of truth); as such, it is not relevant to criminal trials, or at best, favours the prosecution.

  • Similar problems arise if the verifying mechanism built into the computer system is itself either another computer or part of a computer; should the verifying mechanism also require a ‘redundancy mechanism’?

  • The example given in DP 69[59] of a customer checking a bank statement with the bank gives rise to several problems: the evidence would be hearsay and would be required to fall within one of the exceptions to the rule; the inquiry to the bank would result in the unsatisfactory solution of the bank checking its own computerised record; and, unless the ‘verifying measure’ is a guarantee of accuracy (which it is not), it may merely repeat or corroborate whatever in-built problem exists in the data generation process.

  • There would be significant compliance costs in the extra statements and witnesses required simply to overcome an unidentified, unquantified, assumed risk.

  • The cost of acquiring a ‘redundancy mechanism’ may put this beyond the reach of smaller litigants and may unfairly disadvantage them in litigation.

  • The impact of such a test is potentially far-reaching as there are so many documents and other material, such as records, tests and photos, produced on computer or using computer technology. Any requirement that computers be subject to a ‘redundancy mechanism’ could result in these items of evidence being routinely challenged as to an assumed inaccuracy.

6.33 The Australian GovernmentAttorney-General’s Department pointed out that, in a criminal case, the prosecution may have little choice about the type of documentary material available to it. It submits that it is unlikely to be in the interests of justice to require a court to reject evidence that appears cogent and reliable (and which may be corroborated by other material) simply because it does not satisfy formal preconditions for admissibility.[60]

6.34 The CDPP submits that there are significant benefits to be derived from the presumption of accuracy of computer output.[61] The presumption facilitates the admissibility of the numerous documents and business records generated from computer-stored information. When such evidence is becoming more pervasive, it is questionable whether more rigorous tests for its admission should be put in place. Section 59B of the Evidence Act 1929 (SA), for example, has made it harder to get computer produced documents into evidence. Furthermore, it points out, as did the NSW DPP, that other forms of evidence are also prone to manipulation and falsification and yet are routinely accepted.

6.35 The CDPP emphasises that the presumptions in ss 146 and 147 are rebuttable. It submits that existing mechanisms to enable the testing of computer evidence are sufficient.

6.36 The CDPP strongly believes, both from its own experience of evidentiary law and the fact that a review of the case law does not reveal any problems with the operation of ss 146 and 147, that the creation of additional requirements for admission of computer records is not warranted.

Submissions in favour of a more rigorous test

6.37 The Office of the Victorian Privacy Commissioner observes that, in numerous instances in Victoria, technology-generated evidence (particularly speed camera evidence) has been shown to be less than reliable.[62] It submits that it is critical to maintain public confidence in the judicial process and that this can be eroded by even isolated instances of the admission of inaccurate computer evidence. It submits that the public’s confidence in the accuracy and reliability of some technologies has already been shaken. It is therefore important to subject these technologies to scrutiny and maintain the highest standards of testing computer evidence, particularly as computer systems become more sophisticated and complex.

6.38 The other two submissions supporting a more rigorous test for the admissibility do not extensively detail the reasons for their support. The Law Society of South Australia favours the adoption of the approach taken in s 59B of the Evidence Act 1929 (SA).[63]

6.39 The Law Society of New South Wales states that ‘in an age of computer hacking and viruses the rebuttable presumption in s 146 of the uniform Evidence Acts is of concern’.[64] It points out that s 146 envisages application to machine-produced evidence such as photocopies (this is the example given in the legislation), but simple data copying is considerably different from computer-produced data, which can be stored and manipulated. It submits that the existence of quality control or internal control systems should be sufficient for computer-produced evidence to be considered prima facie accurate and reliable. However, it questions what the standard of quality control should be and suggests that that there may have to be different standards for different litigants. (Compare, for example, the computer records and systems of a sole trader with those of a multinational corporation.) It also submits that the concerns raised about the accuracy and reliability of computer-produced evidence apply to other electronic communications such as SMSs.

The Commissions’ view

6.40 The Commissions have made it clear in this Inquiry that a major overhaul of the legislation is neither warranted nor desirable. It follows that a persuasive case for change should exist before the Commissions recommend a legislative amendment. As was stated in DP 69, the Commissions are interested in identifying those parts of the uniform Evidence Acts that may benefit from some fine-tuning in the light of experience.[65]

6.41 The submissions opposing a change in the threshold of proof for computer-produced evidence highlight the lack of evidence, both from their own experiences and from knowledge of the case law, of problems arising from the operation of ss 146 and 147. These submissions have argued strongly that a more rigorous test is not justified.

6.42 Given the division of opinion on this issue, the strongly-held views of those opposed to amendment and the lack of empirical evidence justifying a more rigorous test, the Commissions are persuaded that a case for change has not been made out.

[31] For example, s 48 permits the tendering of a copy of a document produced ‘by a device that reproduces the contents of documents’: Uniform Evidence Acts s 48(1)(b)(ii). This provision allows photocopies and computer-produced copies of documents to be admitted as evidence: S Odgers, Uniform Evidence Law (6th ed, 2004), [1.2.4920].

[32] Uniform Evidence Acts ss 146–147.

[33] Australian Law Reform Commission, New South Wales Law Reform Commission and Victorian Law Reform Commission, Review of the Uniform Evidence Acts, ALRC DP 69, NSWLRC DP 47, VLRC DP (2005), Q 6–1.

[34] Criminal Law Committee of the Law Society of South Australia, Submission E 35, 7 March 2005; Legal Services Commission of South Australia, Submission E 29, 22 February 2005.

[35] Legal Services Commission of South Australia, Submission E 29, 22 February 2005.

[36] Ibid; Criminal Law Committee of the Law Society of South Australia, Submission E 35, 7 March 2005.

[37] These are reproductions made by ‘an instantaneous process’ or produced from a record made by a process in which the contents of a document are recorded by photographic, electronic or other means: Evidence Act 1929 (SA) s 45C(3). Section 45C(4) creates a rebuttable presumption that a reproduction made by ‘an approved process’ (as defined by regulations) accurately reproduces the contents of the document purportedly reproduced.

[38] Evidence Act 1929 (SA) s 45C(2).

[39] Ibid s 59A includes definitions for ‘computer’, ‘computer output’ and ‘data’.

[40] C Spenceley, ‘Evidentiary Treatment of Computer-Produced Material: A Reliability Based Evaluation’, Thesis, University of Sydney, 2003, 233.

[41] Law Commission, The Hearsay Rule in Civil Proceedings, Report 216 (1993), [3.15].

[42] C Spenceley, ‘Evidentiary Treatment of Computer-Produced Material: A Reliability Based Evaluation’, Thesis, University of Sydney, 2003.

[43] ‘Redundant mechanisms’ can involve hardware solutions, software solutions, human solutions, or any combination of the three. Examples of ‘redundant mechanisms’ are: manual verification of output by a person with knowledge of, or at least familiarity with, the expected output; or comparison of the output of interest with the output from a parallel computer system: Ibid, 255.

[44] Ibid, 254–263.

[45] Ibid, 255. ‘System A’ is the system that arrives at the result for which the computer is being used. ‘System B’ is a parallel system that verifies the result, but is redundant in so far as it is not actually needed to arrive at the result.

[46] Spenceley argues that computer input should be treated exactly the same way as it would be treated if offered directly as evidence: Ibid, 263–265.

[47] Australian Law Reform Commission, New South Wales Law Reform Commission and Victorian Law Reform Commission, Review of the Uniform Evidence Acts, ALRC DP 69, NSWLRC DP 47, VLRC DP (2005), [6.37].

[48] Ibid, [6.37].

[49] C Spenceley, ‘Evidentiary Treatment of Computer-Produced Material: A Reliability Based Evaluation’, Thesis, University of Sydney, 2003, 263.

[50] Australian Law Reform Commission, New South Wales Law Reform Commission and Victorian Law Reform Commission, Review of the Uniform Evidence Acts, ALRC DP 69, NSWLRC DP 47, VLRC DP (2005), [6.25]-[6.28].

[51] L Crowley-Smith, ‘The Evidence Act 1995 (Cth): Should Computer Data be Presumed Accurate?’ (1996) 22(1) Monash University Law Review 166, 173.

[52] E Laryea, ‘The Evidential Status of Electronic Data’ (1999) 3 National Law Review 1, [27].

[53] Ibid, [27]. The evidence was admitted on other grounds in Mehesz v Redman (1979) 21 SASR 569 and R v Weatherall (1981) 27 SASR 238. The evidence was found inadmissible under s 59B and at common law in Steiner v Modbury Towing Pty Ltd (Unreported, Supreme Court of South Australia, Matheson J, 5 August 1998).

[54] E Laryea, ‘The Evidential Status of Electronic Data’ (1999) 3 National Law Review 1, [37].

[55] Ibid, [92].

[56] The Law Society of South Australia, Submission E 69, 15 September 2005;Office of the Victorian Privacy Commissioner, Submission E 115, 30 September 2005;The Criminal Law Committee and the Litigation Law and Practice Committee of the Law Society of New South Wales, Submission E 103, 22 September 2005.

[57] Director of Public Prosecutions (NSW), Submission E 17, 15 February 2005; Commonwealth Director of Public Prosecutions, Submission E 108, 16 September 2005; Attorney-General’s Department, Submission E 117, 5 October 2005; New South Wales Public Defenders, Submission E 89, 19 September 2005.

[58] Director of Public Prosecutions (NSW), Submission E 17, 15 February 2005.

[59] Australian Law Reform Commission, New South Wales Law Reform Commission and Victorian Law Reform Commission, Review of the Uniform Evidence Acts, ALRC DP 69, NSWLRC DP 47, VLRC DP (2005), [6.31] fn 58.

[60] Attorney-General’s Department, Submission E 117, 5 October 2005.

[61] Commonwealth Director of Public Prosecutions, Submission E 108, 16 September 2005.

[62] Office of the Victorian Privacy Commissioner, Submission E 115, 30 September 2005.

[63] The Law Society of South Australia, Submission E 69, 15 September 2005.

[64] The Criminal Law Committee and the Litigation Law and Practice Committee of the Law Society of New South Wales, Submission E 103, 22 September 2005.

[65] Australian Law Reform Commission, New South Wales Law Reform Commission and Victorian Law Reform Commission, Review of the Uniform Evidence Acts, ALRC DP 69, NSWLRC DP 47, VLRC DP (2005), [1.31].