Black Box Blues: Rethinking Reliance on a Magic Box

This article has been featured on The Daytona Beach News-Journal Online as an op-ed opinion piece. You can read the article on The News-Journal here.

For over a decade, Florida has been relying on Kentucky-based CMI, Inc.’s Intoxilyzer 8000™ as the primary means of detecting and prosecuting suspected “drunk drivers.” This machine, at the heart of Florida’s DUI prosecution methodology, is fundamentally flawed—whether it is the machine’s mysterious secret inner workings, flawed scientific assumptions, or operation and maintenance error. Justice should not be meted out by a machine, but sadly, too many humans are being judged guilty by an imperfect machine.

Florida’s Breath Test Machine

It is not illegal to drink and drive, rather, the law prohibits “impaired driving.” Impaired is defined as “worsened in a material regard,” which isn’t of much help. However, a blood alcohol content (BAC) above 0.08g/210l creates a presumption the driver is impaired, i.e. drunk.  The machine’s output, a number measuring grams of alcohol per 210 liters of air, is commonly referred to as the resulting BAC. Florida’s legal limit is an arbitrary 0.08g/210l air (it used to be higher)—at or above that number you are presumed to be impaired.  Below 0.05g/210l air you are presumed sober. The middle is without a presumption aside from the omnipresent presumption of innocence.

An individual has no way to know their own BAC at any given point in time, although a handy BAC calculator can be used to estimate one’s BAC using the Widmark formula. Even Widmark’s formula differentiates between men and women—something the machine does not. The old maxim of one drink per hour is hardly bullet proof.

A suspect who is arrested for DUI will usually be held overnight or until their BAC is below 0.05g/210l air. A first time DUI conviction carries with it hefty fines and the real risk of jail as well as a license suspension. Refusing to give a breath test will also usually result in a driver’s license suspension. Without a license or with a restricted permit, life for the working Floridian is a lot harder. Between legal fees, insurance raises, mandatory court costs, and mandatory classes, even a first time DUI can cost thousands of dollars—plus jail. That is why it is so important that DUI prosecution be far more involved than a computer print out.

The breath test machine theoretically detects alcohol in a driver’s breath then produces a reading used to determine if the suspect’s blood alcohol content is above or below the legal limit. The machine works on the assumption that there is a relationship between alcohol content in deep lung air and alcohol content in the circulating blood stream, i.e. in the driver’s brain. Even here, there is a fundamental flaw because this assumption is based on a scientific principle that holds true in an airtight system (closed system), which the human body is not. Further, the ratio is not universal. Issues like this are rampant and difficult to fully detect because the machine’s logic—its programming—is kept secret.

Breath Test Machine’s Mysterious Inner Workings

Using a process called infrared spectra photography, a beam of light is shot through an air chamber where the alcohol absorbs or blocks transmission of the light waves at certain “spots.” By checking two “spots” the machine “detects” the presence of alcohol, whose chemical makeups cause a particular pattern of absorption. The inner workings of the machine, the source code “soul” that processes the deep lung air captured in a tube to spit out a number, is jealously guarded—no one outside the company and perhaps certain law enforcement officials know the exact algorithm used.

One thing is for certain: The machine makes a lot of assumptions that are unsupported by science, and does not accept an input for age, weight, gender, race, or any other variable (the machine collects some token information at the top of the printed ticket but does not factor the input into the internal equation being run).

Garbage In, Garbage Out

Even if the machine is functioning as it was designed, there are still no assurances of an accurate determination of a subject’s sobriety. Again, the machine’s numerous (erroneous) assumptions—including that the human body is a closed system (necessary for the application of Boyle’s Law and Henry’s Law), that everyone’s bodies process and metabolize alcohol in the same way, that everyone’s physiology is the same, etc. (the list goes on and on)—are coded into the process using a “secret formula” that CMI has refused to disclose and which, in Florida, has been modified by the Florida Department of Law Enforcement (FDLE), again without disclosure. The “secret” source code is hidden, despite Florida’s Sunshine Law and the State’s purchase agreement terms, and despite years of litigation under the guise of protecting CMI’s trade secrets. Engineers have a saying, “garbage in, garbage out.” When you feed data in a “black box” with secret inner workings, this saying is particularly apt.

Breath Test Machine Maintenance Schedule

The breath test machines, which theoretically run self-inspections before each test, are maintained locally by an Agency Inspector with two days of training and continuing education, who checks his or her agency’s machine on a monthly basis (not every 30 days, but once each month, meaning in theory almost sixty days can pass between “valid” monthly inspections). The inspection involves swishing homemade mixtures of alcohol and water around in your mouth and then blowing in the machine. The results need to be within a generous margin of error. Then, yearly, a department inspector examines each machine after it is shipped to Tallahassee—no such inspection is performed when it is returned to service, rather the local Agency Inspector checks it out again.

Considering Florida’s climate and the reality of life as a postal parcel (short and probably brutal, if not very shaky), I doubt laboratories would sanction that kind of treatment of their sensitive diagnostic tools. But the risk of transit damage is the least of your worries. The machine is constantly assailed by radio waves and other environmental factors, and since the machine is designed to be portable and used roadside, some are even operating out of on-the-road patrol vehicles. It is a hard knock life for a machine—a machine with a frighteningly limited (and usually long since expired) warranty. There are countless potential attacks on the inspection and maintenance of the machine.

Margin of Error

Even if the source code issue is resolved to one’s satisfaction, there remain many other valid concerns about the accuracy of the reported results. Florida law requires two (2) “valid” samples within +/- 0.02g/210l of air. This is a 25% margin of error (0.02 represents a quarter of 0.08), and a subject reporting results of 0.07g/210l and 0.09g/210l would be perfectly acceptable despite the large variation.

As I like to argue to the jury, the baseline of the machine is a 25% margin of error—and of course, the other side disagrees. However you look at it, the machine’s results are within a margin of error I find unacceptable. How many people would trust an airplane autopilot with this degree of exactitude? In the event the two samples are too far apart, the machine prompts for a third sample. This third sample can be within +/- of 0.02g/210l of either of the previous two numbers. So this means any result between 0.05 and 0.11 would be a valid test and would be considered a legally acceptable and admissible result.

The Human Equation

This is just a brief overview of the questionable “science” and results that arise from and in the use of the Intoxilyzer 8000™ in Florida Court. Is it any wonder I consider a video tape of a suspect’s behavior roadside to be a far greater tool for determining if someone is impaired or not? How a person walks and talks, interacts with others, and generally comports themselves on tape is an excellent tool in determining impairment—not “evaluating” how a subject (suspect) says the alphabet or balances on one foot doing field sobriety exercises (FSEs). Whether due to age, infirmity or inherent clumsiness, many people simply cannot perform these exercises under ideal conditions, much less roadside while facing imminent arrest.

Although FSEs are certainly a poor judge of sobriety, they are the metric by which most police officers decide whether or not to arrest a driver. FSEs were designed to be performed in a standardized way despite, again, offering no variance for individual traits. For some reason, DUI “scientists” are the most equalitarian people on the planet; they cannot treat anyone differently, not the Olympic athlete or the morbidly obese, not the young or the old. All are alike in the eyes of the FSEs.

Man vs. Machine

Drunk drivers are tremendously dangerous; driving is already dangerous enough without adding drugs or alcohol to the mix. I applaud the efforts of law enforcement to protect the motoring public. But I cannot excuse the poor science and flawed logic that underpin the system.

A recent article by The New York Times—“These Machines Can Put You in Jail. Don’t Trust Them.”—called into question the reliability of the breath test machine, citing many of the reasons I listed above and going into far greater depth in a much longer feature article. I believe the Intoxilyzer 8000™ is too flawed to form the basis of a criminal conviction. I trust a jury of my peers and their own inherent senses far more than I trust an archaic machine with less processing power than any piece of consumer electronics I have owned in a decade. A blood test is the only potentially reliable way of determining blood alcohol level, provided that the sample is properly extracted and tested.

As Florida decriminalizes medical marijuana (and in States with or moving towards recreational marijuana) we will likely need a means to quickly test for impairment, begging the question: How do you create an impairment gauge in the first place?

Whatever machine comes next, we must remain critical and we must demand transparency and access to the inner workings. While “garbage in, garbage out” may be the computer engineer mantra, as a trial attorney I know if you put truth in, you will get justice out. I trust in the jury system, not a black box. So should you.

Aaron Delgado is an experienced Criminal Trial Lawyer and a founding partner of Aaron Delgado & Associates. He has tried over fifty DUI trials to verdict, lectures statewide on DUI and trial law, and has been honored by the Daytona Beach News-Journal as the area’s top DUI lawyer.