decided: October 16, 1989.
COMMONWEALTH OF PENNSYLVANIA, APPELLEE,
ANDREW KLINGHOFFER, APPELLANT
Discretionary Appeal No. 40 E.D. Appeal Dkt. 1988 from the Judgment of the Superior Court, 370 Pa. Super. 648, 533 A.2d 1075 (1987), of Pennsylvania at No. 90, Philadelphia 1986, Affirming the Judgment of Sentence of the Court of Common Pleas of Philadelphia County, Information Nos. 222-228, December Term, 1983.
Stanford Shmukler, Philadelphia, for appellant.
Gaele McLaughlin Barthold, Deputy Dist. Atty., Ronald Eisenberg, Chief, Appeals Div., Jo-Ann M. Verrier, Philadelphia, for appellee.
Nix, C.j., and Larsen, Flaherty, McDermott, Zappala, Papadakos and Stout, JJ. Stout, former Justice, did not participate in the decision of this case. Larsen, J., files a dissenting opinion in which Papadakos, J., joins.
Author: Per Curiam
[ 522 Pa. Page 563]
This appeal is dismissed as having been improvidently granted.
LARSEN, Justice, dissenting.
In my opinion, this appeal was not improvidently granted. To the contrary, this appeal squarely presents an issue of substantial importance in our increasingly computerized society, namely the scope of the evidentiary foundation which should be required prior to the admission of a computer-generated expert opinion on a critical issue.
Appellant was convicted by a jury in the Court of Common Pleas of Philadelphia County of two counts of vehicular homicide and related charges stemming from an accident on the Benjamin Franklin Bridge wherein appellant's vehicle crossed over two lanes and collided with two cars in the oncoming lanes of traffic, killing two people. The critical issue in the case was the speed at which appellant was operating the vehicle at the time of the accident.
Appellant had stated to police officers (while receiving treatment in a hospital) that he had been traveling at the same rate of speed as the other vehicles, about fifty-five to sixty miles per hour (mph). The posted speed limit was forty-five mph. Appellant's speedometer was frozen at sixty-eight mph. The Commonwealth introduced Dr. Thomas, an expert in the field of "vehicular dynamics" who testified that, in his expert opinion, appellant's vehicle had been traveling at approximately 75 mph at the point of the first collision. This expert opinion was generated by a computer using a program known as "Applecrash." Obviously, whether the jury believed appellant's version of the accident (that he had been traveling at the same speed as
[ 522 Pa. Page 564]
other vehicles which was about fifty-five to sixty mph) or the Commonwealth's version (seventy five mph) was a critical factual issue, the resolution of which might well have been dispositive of appellant's guilt on the charges for which he was on trial.
When an expert opinion on a crucial issue in a trial is generated by a computer program, how much information about that program and the data that went into it should the trial court require of its proponent in order to prima facie establish the reliability of that computer opinion, i.e., what is an adequate evidentiary foundation to permit the introduction of such an expert computer opinion? Also, how much pretrial discovery of such information and data is the opponent entitled to, and when is he entitled to it, in order to ensure a fair and adequate predicate for effective cross-examination of the expert witness and a fair and adequate basis on which to challenge the accuracy of the computer generated opinion? We granted allocatur in part to address those issues, and I believe we should take the opportunity presented here to answer those questions and establish at least some preliminary guidelines for the introduction and use of computer generated evidence.
Appellant claims that the Commonwealth failed to disclose in timely fashion all of the relevant information concerning the computer program and the data entered into said program by the expert witness, which information was necessary in order to accurately assess the reliability of the computer generated opinion as to speed. Thus, appellant claims there was an inadequate foundation laid to allow the admission of the expert opinion, and that he had insufficient information that was necessary to provide him with a meaningful opportunity to effectively cross-examine the witness and challenge the accuracy of the computer's opinion.
The Commonwealth counters that appellant timely received all of the discovery to which he was entitled under our rules of criminal procedure, Pa.R.Crim.P. Rule 305, and that the record adequately discloses (and timely disclosed to appellant) the program used to generate the expert opinion
[ 522 Pa. Page 565]
as to speed and all of the data fed into the computer by Dr. Thomas to arrive at the computer conclusion of 75 mph.
The record is hazy as to whether all of the data which Dr. Thomas used in running the Applecrash program was, in fact, on the record. However, it does appear that Dr. Thomas failed to preserve or maintain a permanent record of the input he utilized, although he did prepare for trial from memory (at the Commonwealth's request) a one page sheet which contained some of the data/input.
For purposes of this opinion, I have simplified the arguments of the parties and the state of the record, and I omit discussion of the conflicting inferences which the parties would have us read from the record. Suffice it to say that it is not clear precisely what data Dr. Thomas fed into and utilized in running the Applecrash program to generate the computer's opinion on the speed of appellant's vehicle, and the parties are in dispute as to when appellant discovered that the Applecrash program had been used, when he received that programming information, and as to the completeness of the information which Dr. Thomas recalled utilizing. The point is, should a reviewing court have to guess, or should we not establish specific guidelines for the admissibility of computer generated expert opinions and pretrial discovery of the programming information and data that went into the computer, and require that a proper foundation meeting those guidelines be demonstrated on the record.
As the trial judge candidly stated as he struggled with the issue of the admissibility of and foundation for the computer generated expert opinion: "It's a very esoteric field . . . . The computer world is still a great mystery to me as it is to ninety-nine point nine nine percent of the population." Notes of Testimony, Trial at 5.39. The judge's point is well taken, although his percentage estimate of the computer ignorant might be subject to debate.
Despite the court's frank acknowledgment of the esoteric nature and "mystery" of computer programs such as Applecrash, it decided that cross-examination would be sufficient
[ 522 Pa. Page 566]
to point out and uncover the potential gaps and flaws in an expert opinion generated by such a program. However, without adequate information as to the precise data that went into that program and sufficient time to analyze the data and the program, the effectiveness of cross-examination is, it seems to me, seriously undermined.
Appellant has provided us with the following excerpts from Judge Van Graafeiland's dissenting opinion in Perma Research and Development v. Singer Co., 542 F.2d 111, 124-26 (2nd Cir. 1976):
As courts are drawn willy-nilly into the magic world of computerization, it is of utmost importance that appropriate standards be set for the introduction of computerized evidence . . . .
Although the computer has tremendous potential for improving our system of justice by generating more meaningful evidence than was previously available, it presents a real danger of being the vehicle of introducing erroneous, misleading, or unreliable evidence. The possibility of an undetected error in computer-generated evidence is a function of many factors: the underlying data may be hearsay; errors may be introduced in any one of several stages of processing; the computer might be erroneously programmed, programmed to permit an error to go undetected, or programmed to introduce error into the data; and the computer may inaccurately display the data or display it in a biased manner. Because of the complexities of examining the creation of computer-generated evidence and the deceptively neat package in which the computer can display its work product, courts and practitioners must exercise more care with computer generated evidence than with evidence generated by more traditional means.
Roberts, A Practitioner's Primer on Computer-Generated Evidence, 41 U.Chi.L.Rev. 254, 255-56 (1974).
There are those knowledgeable in the field of computerization who believe that new evidentiary rules will be required to channel and control the use of this new
[ 522 Pa. Page 567]
medium. Freed, Computer Records and the Law-Retrospect and Prospect, 15 Jurimetrics J. 207, 208 (1975). Indeed, much non-judicial effort has already been expended in this field, and specific recommendations have been made. See e.g., Federal Judicial Center, Manual for Complex Litigation § 2.717 (1973); ABA Sub-Committee on Data Processing, Principles of Introduction of Machine Prepared Studies (1964). See also Bibliography of Representative Articles and Comments in Freed, supra, 15 Jurimetrics J. at 218.
Judicial decisions to date have largely skirted the edge of the problem because they have been concerned mainly with computerized records made in the regular course of business. [Citations omitted]. Routinely prepared records, admitted pursuant to business records acts such as 28 U.S.C. § 1732 are well recognized exceptions to the hearsay rule, because their regular use in the business of the company insures a high degree of accuracy. Proof of day-to-day business reliance upon computerized records should therefore make less onerous the burden of laying a proper foundation for their admission. United States v. Russo, supra, 480 F.2d  at 1239-40.
Where, however, a computer is programmed to produce information specifically for purposes of litigation, an entirely different picture is presented. Its product, which is hearsay and conclusory, is not admissible under 28 U.S.C. § 1732 or similar state statutes. [Citations omitted]. Under such circumstances, a court should not permit a witness to state the results of a computer's operations without having the program available for the scrutiny of opposing counsel and his use on cross-examination. United States v. Dioguardi, 428 F.2d 1033 (2nd Cir.), cert. denied, 400 U.S. 825, 91 S.Ct. 50, 27 L.Ed.2d 54 (1970). Moreover, such availability should be made known sufficiently in advance of trial so that the adverse party will have an opportunity to examine and test the inputs, program and outputs prior to trial. United States v. Russo, supra, 480 F.2d at 1241.
[ 522 Pa. Page 568]
. . . Because of the computer's "ability to package hearsay and erroneous or misleading data in an extremely persuasive format", Roberts, supra, 41 U.Chi.L.Rev. at 279, this rule should be strictly adhered to whenever expert testimony is predicated upon specially prepared computerized calculations or simulations.
It is a mistake to liken the program of a computer to human calculation, because the program directs the performance of tasks that humans would not attempt, in a manner that they would not elect. Jacobs, Commission's Report on Computer Programs, 49 J.Pat.Off.Soc'y 372, 374 (1964). An error in programming can be repeated time after time, and simulation with an incorrect program is "worse than worthless". Favret, Introduction to Digital Computer Applications, supra, at 122. For this reason, programming requires great accuracy, more than that needed in other types of engineering. Ershov, Aesthetics and the Human Factor in Programming, 13 Jurimetrics J. 142 (1973).
Because the record in this case contains no information about the programming relied upon by plaintiff's experts, neither we nor defense counsel have any way of knowing whether it was complete or accurate.
Id., at 124-126.
See also Nelson, Garbage In, Garbage Out: The Need For New Approaches to Computer Evidence, 9 American Journal of Trial Advocacy 411 (1986) and Comment, Guidelines for the Admissibility of Evidence Generated by Computer for Purposes of Litigation, 15 University of California, Davis, Law Review 951 (1982).
Judge Van Graafeiland's opinion discusses some of the problems and dangers posed by the prospect of trial by computers. There are some serious problems and significant dangers. I will not attempt to craft comprehensive guidelines for the introduction of computer generated evidence in this opinion, for it would be merely an academic exercise at this time. I write to point out the need for this
[ 522 Pa. Page 569]
Court to study the problem and to establish clear, specific procedures for the admission of such evidence.*fn1
As the saying goes, "garbage in -- garbage out." If computer generated evidence and expert opinions are to be used to deprive a person of his liberty as here (and such use of computer evidence can be expected to be made with increasing frequency), surely the judicial system must demand a high degree of reliability and disclosure of adequate information as to programming and input to allow meaningful cross-examination, and to permit the jury to determine whether any garbage went into and contaminated the computer program. Moreover, we must require disclosure of the necessary information to the opponent prior to trial to allow sufficient time to locate a qualified expert in the field who speaks the unique foreign language of the particular computer program who can assist the opponent and his advocate in meaningfully interpreting the data and program and in challenging the accuracy of the results. While there are some exceptions, as a rule I would not expect our judges, juries or lawyers to be well-versed in the foreign computer language used, nor to have the requisite expertise and knowledge to understand how a computer generated a particular opinion. On the other hand, it is easy for a layman to grasp the computer's final result, for example "defendant was traveling at 75 mph," and to be seduced by the magic and technology which enables a computer to analyze staggering amounts of information and variables and transform that data into a conclusion.
[ 522 Pa. Page 570]
Such a conclusion -- a computer generated expert opinion -- might be accurate and useful in litigation, but only if the information fed into the program, the program itself, and the operation of the program have been accurate and complete. Neither computer nor its programmer are infallible, and unless there is, on the record, sufficient information and data about the program and the programming to establish a proper foundation and to permit meaningful cross-examination, a jury's verdict that may have been influenced by a computer generated opinion on a critical issue will be suspect. It has been said that only a small (two digit) number of men and women truly understand the formulae, processes, equations and reasoning that led to Einstein's theory of relativity and his landmark conclusion E = MC. Obviously more people understand the workings and language of computers, but not so many that I would allow the use of computer generated opinions and evidence at trial without adequate assurances of reliability and meaningful opportunity to permit the parties and advocates to effectively challenge the accuracy of such computer conclusions before the finder of fact.
*fn1 If this Court were to address the issues in this case, it would be appropriate to craft at least some minimal requirements to establish the prima facie reliability, vis a vis foundation and discovery, of computer generated expert opinions, and to measure the use of such computer evidence in this case against those minimum requirements. However, the field of computer analysis and programming is so unique and complex, and the language of the field is so foreign, I would not be comfortable prescribing permanent guidelines at this time. To the contrary, I believe this Court should refer the matter to the appropriate rules committee with instructions to enlist the assistance of competent, qualified computer experts to prepare a report to this Court and to draft proposed guidelines/rules.