Our recent issues with publishing Diaspora at Lulu have been illuminating, and I’ve already written at some length about how Lulu leverages automation in order to succeed. So let’s talk a little about how humans fail automation — that is, the way the remaining (and necessary) human links in an automated chain fail when we choose to behave like machines.

As we automate more and more we tend to come to see ourselves as part of the machinery, and we certainly are. Unfortunately, we also tend to see this as liberating us from the more complicated aspects of our work — the communication and the negotiation and the decision-making — and this is not the case. These things are exactly why we remain in the loop.

There are a couple of places where humans have reveled in becoming machine-like to grave effect. The first is the whole idea of zero-tolerance. Giving humans the option to point at a rule and declare “there’s nothing I can do” is effectively replacing that human with a machine (that machine being, even more sadly, themselves). This is deeply broken. One of the chief values of having humans in authoritative roles is having an intelligent decision-maker think about and resolve issues that don’t need to be followed through the apparatus to punishment or whatever. If we want a human to make a non-choice like that, we should imagine them replaced by a machine, because that is what we have done. This is unacceptable. It de-socialises human behaviour. Let’s not do that.

So how does this relate to our issue at Lulu? Well the crux of the problem was purely technical. There is an internal representation of page numbers in any PDF document that is not necessarily the same as the printed (folio) representation of page numbers. For some reason (and this is certainly a defect, but not central to the problem), a printer that Lulu uses throws an error when this internal representation is more than some fixed number of characters. Up until our problem started, the humans in this print shop had been manually changing the internal representation to something suitable and then printing.

Okay pause. Red flag. When you, a human, do the same thing over and over again and you hate it, STOP. Raise the issue so that it can be resolved at a higher level or, (much much better) automate the problem away (this, incidentally, is why everyone in an automated environment should be facile with a comfortable high-level programming language). If someone had said “oh man I can bang out twenty lines of Perl that detect and resolve this”, both sides of the issue would go away — I wouldn’t care about the internal page representations and the printer wouldn’t care either. Lubrication! Humans to the rescue!

Anyway, they didn’t automate this, which would have followed the fundamental rule of implementing interfaces (interpret the spec broadly for receipt and tightly for send). The fallback option, speak up, was also not followed. Instead they tolerated the manual addition to the process until one day they stopped tolerating it.

And here’s the next big failure of humans as part of a chain of automation: they chose not to communicate it. They threw the error (the wrong one incidentally, which is even less help) and shut up. Now at this point they not only know what’s wrong, but they also know exactly how to fix it. Seconds after the issue arises. So how do we get to fourteen days later before the problem is resolved?

Well, the print shop obviously either felt no obligation to suggest the fix (not our job, man) or did so but was not heard by the right person (not our job, man). Both of these are humans acting like machines — doing only what is required or what is on the procedural checklist — instead of humans. Instead of wanting to correct the problem, they perceived their role as strictly defined and immutable: a cog.

And so eventually it took someone with a personal interest (thankfully a human success — a friend contacts a friend and I make contact with someone who really cares about solving the problem because there is social capital at stake) chasing down the process and talking with humans and understanding the problem. That this should be such an intractable problem for so long speaks to the ultimate failure of people acting as though they are a cog in the automation — we actually prefer it that way. It is less stressful to point to the written procedure and claim perfect compliance. It is also entirely defective as a human function. Sadly, a good many industries encourage exactly this failure and people are certainly going to adapt to a situation that rewards ass-saving behaviour by providing rules that demand the path of least resistance.

So Nick, thank you for embracing the problem as a thinking human being and driving to the heart of the matter (perhaps literally). Fred, thank you for making the problem a social one rather than a technical one — even people outside the strict boundary of the problem space can help, and choosing to feel obligated to do so is a virtue we should all strive for.

We remain the smartest and most flexible part of any machine, most especially when we adapt the machine to reduce our busy-work. We should hate busy-work but we should not refuse it on a technicality — we need to exercise our novel (for now) capabilities and change the damned machine when it needs changing.

Is Diaspora up? As of this writing, not officially. It is, but there’s a coupon pending as a thank you to our patient customers, so unless it’s urgent, I’d wait for the official announcement. But yes, the problem is fixed, and there are two very human stars in the story. Thank you both.

–BMurray