One of the few virtual, screen-based interfaces in the main control room. Doesn’t it feel abstruse and anxiety-inducing, compared to the solid physicality of all the previous photos?

What happens if this screen has a glitch during an emergency? Or the mouse cord gets frayed? Kind of puts "blue screen of death" in an entirely new context.

Co.Design

Look at This Crazy Russian Nuke Plant: Are 10,000 Buttons Safer Than One Screen?

Buttons, knobs, and dials everywhere — and barely any screens. A nuclear plant in Smolensk makes us wonder: are physical interfaces safer than virtual ones?

Touchscreens (and screens in general) are awesome. So sleek! So dynamic! So futuristic! Yeah, well what if they're fundamentally confusing and dangerous, too?

It sounds like heresy, but I couldn't help wondering that after seeing the rows and rows of physical buttons, gauges, and readouts that Russian technicians use to run this nuclear power plant. Built post-Chernobyl, the plant boasts a "30km-wide security zone around the plant itself, filled with all sorts of sensors and monitoring devices that measure the condition of the environment to report any smallest deviation from normal radiation doses." But what about all those clunky, straight-outta-Star-Trek knobs and lights — what if they're a safety feature, too? The plant was completed in 1990, so color computer screens, mice, and "normal" high-tech user interfaces were certainly available. How could a dizzyingly dense grid of plastic buttons be better than a single screen that can change to display only what the technician needs at any given moment?

Our brains and hands evolved the way they did for a reason.

Well, here's the thing, as Christopher Mims at Technology Review brilliantly points out: touch is a powerful, powerful thing. And not the sterile, featureless version that passes for "touch" on your iPad. I'm talking about the physical, primal, ultra-high-res sensorium that you experience from interacting with everyday objects in the real world. Our brains and hands evolved they way they did for a reason, and virtual displays and interfaces simply don't "click" with the kind of infomation-processing we've evolved to do so well. Deep, spatial sense-memory — "colored THING in THAT location that feels like THIS and STAYS there" — is how our savannah-dwelling ancestors navigated their environment and avoided getting killed, and it's still true today.

So while there were surely other factors influencing the Russian nuke-designers' UI philosophy (economics being a huge one), it's hard not to wonder if they shied away from screens on purpose. When you're always one wrong command away from another Chernobyl, it pays to play by the brain's built-in rules — not Steve Jobs's.

[For our previous cover of the nuke plant click here | Hat tip to Technology Review]

Add New Comment

11 Comments

  • Larry Marine

    As a veteran UX designer, I find it absurd, yet all too common, that someone could focus on such a minor aspect of good design as the difference between touch and virtual and try to pass it off as some kind of profound insight. Of course, tactile feedback is somewhat preferred over a virtual interface, but we learned that decades ago. Tactile interfaces limit what you can do and provide little more than raw data and control, making the system completely dependent on the highly variable skills and expertise of each user.

    In the early decades of automobiles, the driver had a spark advance level on the steering column to advance the spark as he/she increased the RPMs. Failure to do so properly resulted in backfiring. Engineers finally realized that drivers were overwhelmed with this due to the many other tasks requiring attention during the course of driving and rarely operated the spark advance correctly. They developed an integrated spark advance technology that removed this task from the many other tasks and effectively automated the function. The same can be said about the automatic transmission, as well. Yet, contrary to the author's premise, removing these "tactile" controls dramatically improved the operation of the "system."

    I contend that if the buttons are an improvement, then it's not because they are tactile, its because the previous design was significantly much worse. I have investigated and designed complex systems in my decades of UX design and can attest that every complex system interface I've seen were so terribly designed that they promoted critical human errors. They all suffered from the same design failure, as well. Just like the spark advance on the early automobiles, they all placed too high of a reliance on the individual users' skills.

    Those buttons on the Russian Nuke plant, though providing nice tactile feedback when touched, rely too much on the highly variable skills of each and every user, regardless of their training. As a general rule of thumb, the more training a system requires in order to operate it effectively, the more poorly designed the "system" is. If a real emergency occurred, it's quite likely that the visual barrage of blinking lights would actually cognitively cripple even these highly skilled users.

    This article is little more than the design equivalent of folk psychology.

  • Michael K

    To suggest that control systems like the one depicted above are in any way safer than modern control systems is ridiculous! I wont dive too deep into the reasons, but with modern systems, the ability to trend long term operating conditions, easily update logic and add sensors and other controls make modern systems significantly more capable and preferable by users then old knob and dial systems. In my experience, every industrial operator I've met would never return to the old way of doing controls!

  • Dominic Muren

    I wouldn't even assume it is an economic decision. There are plenty of known bad designs that keep being made because of bureaucratic entanglement or regulatory standards. For example, in the US, curb cuts offer a ridiculously cheap, accessible way for ordinary city dwellers to mitigate stormwater runoff while watering their parking strips ( http://www.ia.nrcs.usda.gov/fe... ) However, because of 1940s-era legislation to protect the function of combined sewer systems, these sorts of cheap, easy-to-install innovations are actually illegal in many cities in the States.

    More importantly though, is the question of whether a few screens really is the best option in this case. The control room you show (and the overwhelming majority of control rooms in a Google image search for "nuclear control room") adopt a distributed approach to information visualization -- that is, variables are displayed on individual readouts, grouped into relevant hierarchies, and presumably, with some degree of redundancy, and fault checking built in.

    In the centralized approach you describe -- where a few monitors display lots of different pieces of information -- you might have faster apprehension of information from the screen. But if anything should go wrong with the screen -- system crash, faulty pixels, lost power, etc, you would lose all of these readouts at once.

    My gut feeling is that this redundancy and fault-tolerance is what keeps these interfaces complicated despite the consequences. And sure enough, even the few LCD-based systems on an image search have an overwhelming overabundance of screens, probably so they can be swapped in in case of failure.

    Bottom line - design can make things less complicated, but only by removing complexity. Some problems are just complex, and simplifying them only makes them less controllable or predictable.

  • Nick Preston

    From reading these comments, I get the impression that none of the commenters have any clue what is involved in operating (or the 8-10 years required to be trained to understand and operate one) a Nuclear power plant. The systems are many and complex, a good operator needs to understand a little more than "just push the button when the light flashes". I have seen PHDs run screaming from the "high school educated" training program.

    What is missing from the virtual interfaces is the "feel for the machine" that you can only get from something tactile. Computer displays are a great tool for status displays and information summary, but fall short when used as a "Human Machine Interface". I have personally used both and know from experience exactly what the author of the article is saying.

    Get out there and lay your fingers on a REAL control panel It cant be beat for ease of operability and understanding of the entire plant.

  • Seth Close

    "When you're always one wrong command away from another Chernobyl..." is an ignorant statement. I build software for nuclear plants and, despite the misconceptions most people have garnered from The Simpsons, nuclear power plants today are incredibly safe. Most of the software is safety and oversight software. There is absolutely no way one software command could do much damage. Before you reference historical events, you may want to research them. Chernobyl was due to political pressure to run an unsafe experiment in a reactor without a containment dome. What's interesting is that the (completely contained) incident at Three Mile Island was due to too many instruments and alarms going off simultaneously. This article had some good ideas. It's too bad the author didn't do their due diligence on the background.

  • tom_ter

    In graduate school we had a guest lecturer from Russia who explained what he was working on -- GUIs that modified what was displayed/what options were available based on 'current situation'. This was pre-web stuff (early 1990s), aimed particularly at nuke plants or space ships. He explained that in commie USSR/post-Gorby Russia, only PHDs ran the nuke plants. In the US, the average education level of the workers = high school.

    WHat gives? He explained all those PHDs worked against Chernobyl -- when things went wrong, the smart guys spent way too long over-diagnosing what was happening. By the time options were ruled out, too late. He compared that to the US: the controls were designed for high schoolers to use. A problem light flashes, you hit the proverbial "off" button.

    Anyway, he was out in front of the meta interface concept.

  • Eric Rice

    The problem with that reactor control panel (and the thesis of this article) is twofold:
    1. It's huge. A reactor is a large, complex system, and it takes a lot of knobs, dials, and readouts to represent complete information on the system. Imagine trying to correlate and synthesize information happening on opposite sides of that room to understand what's going on.
    2. It does not intuitively represent what's going on in the system. Take, for instance, the next-to-last display in the set of pictures. To me, as an engineer who interacts with large, complex systems, that's not "abstruse or anxiety-inducing", it's representing state information of the system on a schematic. The field of buttons means nothing without extensive and immediately accessible knowledge of the system, but that computer display provides everything a trained operator needs to know about both the design and state of the system in a single place, allowing for much faster and more effective assessment of information.

    The computer display may feel "sterile", but it's representing information much more effectively and enabling much better response times ... it can also interpret data for the operators and bring potential issues to the operators' attention. I'll take the intellect of the computer over the visceral appeal of a bunch of buttons, thankyouverymuch.

  • Joel Blair

    Everyone loves to press a button. Just look at babies. There's something so satisfying about a good springing, clicky button. All this is lost on touch screens and touch sensitive interfaces. It's like playing a piano verses. . . well, playing a piano on the iPad.

  • Quentin Kramer

    My question is whether they would really need all those fancy detectors and arrays to figure out the plant melted down...I don't think at Chernobyl it was up for debate whether something had gone wrong :)

  • Jason Braun

    Besides would you really want a "dead Strip" problem on a nuclear reactor?