Invisible. Uncontrollable. Incomprehensible.


You might hope that making more things digital would make more things countable and controllable – and, perhaps, even more understandable. Not so simple, says Peter Coffee.

Invisible Man 1933
Invisible, uncontrollable and incomprehensible – HG Wells-style

In just the past week, I’ve seen a salvo of news stories of a genre that one might label as “I can’t tell what happened, I can’t control what happens, I don’t understand what’s happening.” It’s perverse that this is occurring in a time when “digital transformation” is the buzz phrase of choice – considering that many people think of “digital” (versus analog) as having inherent virtues of precision, accuracy (not the same thing), and clarity.

In many ways, it seems, the modern digital age is still (or perhaps even increasingly) fuzzy.

Naïvely, we might hope that making more things digital would make more things countable and controllable – and, perhaps, even more understandable. When we used to record phenomena by analog methods—pen traces on calibrated pinwheels of paper, for example—we didn’t get the ability to perform computation or other processing on that data as readily as we take for granted today. “Digital” can (and should) mean transformable and analyzable, not just encodable (as I observed here three years ago).

Pre-digitally, we managed after a fashion. Mechanical instruments to perform analog processing, working from our pen-and-ink chart-recorder records, are among the charming anachronisms of generations past. Oh, wait, of my generation: we had these things at the chemical plant where I worked in the late 1970s, when we put in the first all-digital control room. It felt almost Promethean: bringing the fires of digital virtue down to analog Man. Weren’t we something.

On the other hand, those pen-and-ink charts had certain virtues of certainty: what you saw was what there was. No mystery, no ambiguity.

Post-digitization problems begin immediately, though, with the issue that I mentioned toward the end of my comments here a month ago, when I observed that all digital representations have to make a decision about how many bits to employ in representing a quantity – and that they therefore have to embed a decision of how good is “good enough.” Or “big enough,” or “[metric] enough” (for example, the range of colors that can be represented on a screen or other output). You can get a pretty good argument going, for another example, on the subject of what’s wrong with the most common standard for representing enormously large or small numbers using floating-point formats.

These are intrinsically digital problems with intrinsically digital solutions. If one is not happy with floating-point representations, there’s the option of representing numbers exactly as strings of decimal digits, their length limited only by available memory. This and other similar alternatives have daunting consequences of large memory usage and poor computational speed, but trade-offs can be considered and choices can be made.

So far, then, we’re talking about things that may not be conspicuous, but also are not malicious: we can see them if we know where to look, and we can change them with perfectly predictable effect. All good – but we’re just getting started.

The invisible bits

Among the news items, for example, that triggered my initial line of thought this month was a complaint about a retail web site that would not let a customer place more than thirty-two items in an on-line shopping cart. I saw that “32” value, with its suspiciously power-of-two aura, and had a sinking feeling: “Please, don’t let this be a deeply buried five-bit field that can only represent order item ID values from Item 0 through Item 31.” To the customer, the root of this kind of problem is completely invisible.

Further, when one customer challenged the retailer in question with a posting on the retailer’s Facebook page, the dialog exemplified much of what’s wrong with superficial attempts at making enterprises seem more social:

Q: “Why in the world is there now a limit on the number of items I can have in my cart? I get an error telling me I can’t have more than 32 items.”

A: “We are sorry if you’re having problems with your order. Can you please provide detailed information about your order issue so we can see what we can do to assist?”

Detailed information? Would that be “Well, one of the items is red”? This level of dialog feels like a throwback to ELIZA. “Tell me more about your mother. I mean, your order.”

Again, though, there is a digital solution to this digital problem. If there’s an arbitrary limit on the size of a single order, there can always be multiple orders as a work-around. Annoying, but not out of control.

No button to push, no lever to pull

My triggering example for “uncontrollable” was another news item, concerning a dispute in the delivery of electrical power in twenty-five European countries. Network glitches disturbed the sine-wave frequency of the alternating-current mains supply, causing conventional “synchronous motor” clocks to drift by many minutes during the period of conflict.

To customers, this was an unheralded and completely mysterious behavior in a system that offers no readout to explain it, no knobs to turn or switches to flip to fix it. The assumption of accurate, constant AC frequency is literally hard-wired into the system and the devices that rely upon it. Utterly uncontrollable – and in the same vein as many of our most modern systems, whose ancestors offered many more ways to mitigate their misbehavior.

To mention only one such example of reduced controllability, the mechanical hand brake in older automobiles provided a useful and controllable alternative for slowing or even maneuvering the car, while the electrically operated parking brake on even some of today’s sports cars is digital – as in it’s on, or it’s off. I’m not a fan.

A mystery wrapped in a perplexity

The final term of pain that’s on my mind this month is “incomprehensible”: it was triggered by a remarkably public academic argument. Unlike, perhaps, most spats among economists, this one involves the manner of estimating a value of considerable general interest: how much money are ride-share drivers actually making, after allowing for all the expenses associated with owning and operating their cars?

Several months ago, my wife and I discussed this question while waiting for a red light to turn green as she drove me to the airport: in just that short time, we came to the conclusion that on an all-in basis, an Uber driver making that airport run (for about a third the cost of a taxi service) was probably just breaking even, if (a crucial qualification) all car ownership costs were fully amortized over a scenario of full-time Uber driving. Researchers are now arguing this very point with Uber, based on wide-ranging interpretations of how drivers were asked and how they answered various questions for a newly published study.

As I said, incomprehensible. The numbers are all there to see. The drivers are in control, being able to turn on the app and accept trips only when surge pricing makes it worth their while (if they are so inclined). And yet, it seems to be genuinely difficult to figure out what’s actually happening at their bottom line.

Now that it’s working, do we like what it does?

Mere technical improvement in systems, in short, even if including magical spells of “now with digits,” can sometimes merely open new doors into behaviors that are much more challenging to understand and address. Anyone involved in planning or managing these transformations needs to be giving early attention to the next-day story: “And then what happened?”

Unless, of course, you’d like to be the cautionary-tale topic of my next Diginomica contribution.

Image credit - Invisible Man (1933) via Vimeo

    Leave a Reply

    Your email address will not be published. Required fields are marked *