Apps and the principal-agent problem

I've been thinking a bit more about the paradoxes of online advertising, which I previously wrote about as an example of the importance of voluntary unfreedom. Key to that argument is the "your computer, your rules" doctrine, which is at the core of debates around DRM, the war on general-purpose computing, and other ideas around computational autonomy. Your home is your castle, and by analogy your computer is your internet castle, safe and under your control. Only, well, it's not.

It's no question to anyone who's been paying attention that the locked-down world of mobile operating systems is a vastly superior experience for the average user. The time between a normal person getting their hands on a Windows machine and it being unusably riddled with crapware is measured in minutes. Sure, you can have antivirus programs that try to find the crap and remove it, but they're fighting a doomed battle, like a digital King Canute kicking against an endless tide of cyber-excrement. Meanwhile, in mobile land, spyware is comparatively rare, viruses don't exist, and most likely the only time you'd see an antivirus would be while reading an only-90s-kids-will-remember post.

The reason the mobile model is superior is simple: it doesn't trust the user. Or, more precisely, it recognises there's been a fundamental shift in the trust model since the early days of personal computing. Once upon a time, if a program was running on your computer, it was your program, doing something you had told it to do. But the combination of "just download and run it" software, everything running in the background, and the transition from computers being for experts to computers being for everyone means those days are dead and gone. The things running on your computer aren't your programs; they're someone else's programs, and you have no idea if they're doing what you want.

This is the classic principal-agent problem. Although you have nominal autonomy over your computer, you can't directly control the programs running on it, you don't have direct visibility over their actions, and most of the time if you did you wouldn't understand what they're doing anyway. These programs are your agents and, just like a financial manager or a political representative, they may decide to act against your interests if that's more lucrative. Worse still, you have no way of knowing if this is the case until long after they've done terrible damage.

So the great appification, far from reflecting a disregard for users, is a recognition of this new world where most people aren't, and can't be, masters of their own computers. And, in a sense, I think this is actually kind of liberating. Any time I'm at a command prompt I'm a few fat-fingered keystrokes from accidentally ruining whatever I'm working on. Every command I run could potentially contain some code that deletes every file in my home directory. But on a mobile OS? They just don't assume everything that happens is deliberate.

Now I should say that this isn't a completely rosy picture. The principal-agent problem hasn't gone away on mobile (or on the web), it's just been shunted to somewhere else. Just like you can hire a sales manager if you don't understand whether your salespeople are doing the right thing, the people who make your mobile device can make sure your apps are doing the right thing. But, y'know, quis applicat ipsos applicationes? What if your sales manager doesn't know what they're doing? Do you hire a sales manager manager? You eventually have to trust someone.

Unfortunately, "in the end you can only trust yourself" makes a good tagline for an apocalyptic western, but is lousy as a philosophy for user experience. Locked-down devices are safer and better for most people, even if they drive me nuts when I want to mess with their innards. What we need isn't to push back against appification, but to accept it and make sure we have viable and trustworthy options for gatekeepers. One that isn't run by a tech company 5 minutes off the West Valley Freeway would be a good start.