How Do You Cross-Examine Siri If You Think She's Lying?

• Date: 01 January 2022
• Author: John M. McNichols

Published in Litigation News Volume 47, Number 2, Winter 2022. © 20 22 by the American Bar A ssociation. Repro duced with permissio n. All rights reser ved. This informati on or any portion the reof may not be copie d or disseminated in any

form or by any means or sto red in an electronic da tabase or retrieval sy stem without the ex press writt en consent of the Amer ican Bar Associatio n.

By John M. McN ichols, Litigatio n News Associate Editor

How Do You Cross-Examine Siri If

You Think She’s Lying?

closer to mimicking human thought, the need to challenge and

“cross-examine” the evidence they provide may have arrived.

What Is AI?

Although there is no singular denition of AI, most technol-

ogy writers use the term to describe the algorithmic coding

that enables computers to analyze data and engage in autono-

mous decision-making to achieve pre-programmed objectives.

AI is now a part of our daily lives, powering Apple’s voice-

operated personal assistant “Siri,” self-driving cars, and—less

interestingly, except perhaps to lawyers—document review

software that can distinguish relevant e-discovery from irrel-

evant chaff. A computer’s ability to undertake such tasks is

often the result of “machine learning,” the process of recog-

nizing patterns in data to make assessments and predictions

without human direction.

Unsurprisingly, AI has also found its way into the court-

room. There is no shortage of software designed to create

forensic evidence, from blood-alcohol concentrations to

the presence a particular person’s DNA at a crime scene. In

many instances, the software’s report or “conclusion” is not a

binary “yes”or“no”—or, as in Quick’s case, a nite number—

but rather is an expression of probability, more closely resem-

bling a human opinion than a mathematical calculation.

© Getty Image s

n March 1957, an ofcer stopped Preston Quick for

driving 40 in a 30-mile-per-hour zone. To blame? A lead

foot, but also a newly invented electric timer for clocking

driving speeds. At trial, Quick challenged the admissibil-

ity of the timer’s readout, arguing that it constituted

hearsay because it was an out-of-court assertion offered for

its truth. The St. Louis Court of Appeals rejected Quick’s

argument, explaining that under such logic, nearly every form

of instrument-based evidence would be inadmissible—even

the sounds of a stethoscope in a doctor’s ear. The idea that

Quick might have the right to probe how the timer arrived at

its ultimate conclusion—or to cross-examine someone other

than the arresting ofcer who testied about its operation—

did not factor into the court’s analysis at all.

In the decades since, most courts have followed suit, rou-

tinely admitting “machine testimony” that would undeniably

be hearsay if originating from a human declarant. In doing so,

courts have cited the perceived “objectivity” of such evidence

over fallible human testimony. Meanwhile, the Federal Rules of

Evidence dene “hearsay” to encompass only statements by “a

person.” But the advent of articial intelligence (AI) has called

into question whether the traditional approach should still

apply. The complexity of today’s devices has long surpassed that

of the timer Quick encountered, and as machines come ever

2 | LITIGATION SECTIO N

TECHNOLOGY