Vagueness, Verification and Modern Systems
05 Oct 2019Vagueness and Philosophy vs. Science
Recently read this interesting article that argues against video verification in sports, that is using instant replay to assist referees in making close calls. The argument was basically as follows: inherently rules in sports (and other domains of human activity) are very difficult to make specific enough to eliminate all vagueness, and so because of this vagueness gathering more data (e.g. video evidence) will ultimately increase rather than decrease the difficulty of these decisions.
I think this may be right. I see this phenomenon regularly in my work in building data processing and analysis pipelines. When questions or interests are ill-defined (i.e. vague) then more data ultimately increases confusion rather than clarity. The closer you look at things you don’t understand—especially if you have no specific strategy for how you want to understand that phenomenon—you the more you don’t understand.
However, of course, I don’t take this to be an argument against gathering data. In my opinion, that’s a naive conclusion. The problem is with vagueness and non-specificity. As the precision and diversity of data/ information from our modern systems increases, we need to proportionally increase the specificness of the questions we ask. And most importantly, the criteria by which we determine a question has been adequately answered or a decision can be made, that is a decision rule. Essentially, we need to more clearly specify outcomes and how we come to them.
It seems like we’ve come a long way as a society in terms of evidence gathering (measurement) as well as how to aggregate evidence to describe phenomenon (statistics)
Working at a startup is kinda like you’re in some kind of rally car trying to win some kind of race, where the destination/ finish line is unknown, the specific directions unclear and your potential speed virtually limitless.
But the faster you go the more fuel you use and you only have a limited amount of fuel (your fuel tank’s only so big). So maybe you keep some fuel cans in reserve, or maybe some companies give you free fuel along the way, in exchange for putting some ads on your car. Or maybe you use that otherwise reserve fuel to boost your speed for a while, leaving none in reserve and you having to stop for more.
But, since the destination’s unknown, even if you decide to move faster it’s not clear you’re going in the right or wrong direction more quickly. And, even more, since you’re human your tolerance for speed is also limited: you might get sick from the turbulence and/or afraid of a crash.1 So basically you go as fast as you and your vehicle can tolerate.
For some reason, I envision this as some kind of scene from an old Charlie Chaplin film where’s there’s no sound and it’s slapstick all the way: he’s running around in this Model T looking thing driving as fast as he can on this muddy, bumpy road stopping every once and a while to get out and fill up using a fuel can, spilling and tripping as he does. It’s alot like that. Except, of course: startups aren’t Model T’s, they can go as fast you can tolerate. And it’s also not as funny.
Implications of Rolling Start + Unknown Route
Oh and by the way, there’s no official start to these races. They only start for you when you start them, but they’ve typically started for everyone else already and you don’t know how far ahead the others are because you can’t see them as they’re scattered along “the route”. And, even more, even when you can see them, you don’t know precisely where the finish is anyway so it’s difficult to tell who’s ahead. It could be that even though you’ve started later, you’re closer to the finish because your product at its inception is more like the final, winning solution in some possibly very fundamental way.
A nice, concrete example here might be Facebook: there were many social networking websites existing in 2004 when it came onto the scene, but somehow very clearly it won out. The main reasons they won are (a) they solved the identity problem in that they tied people’s accounts to their college email addresses (as a way of verifying their identity) and (b) they moved quickly. So they started out closer to the final, winning solution at their starting point, and they moved more quickly than the others in realizing that solution
And I think all of this implies for other startups that it’s definitely worth giving thought to your starting point, how you can somehow position yourself fundamental closer to the final, winning solution. Which isn’t easy, you have to envision the end point, the finish line (the final solution) without a proper map and then plot the route from there.2
Times N
And this is all also a simplification, because you’re usually not focused on a single big race,3 you’re really thinking about some handful of small races N. And you have some fleet of vehicles, each with limited fuel, drivers and all the rest. And, even more, there are an almost infinite number of potential races you can join, but you have to find them, in some cases pay some fee to enter, and then go through all the operational/ logistical hurdles to get your vehicles to some starting point (which again is vague and unspecific) with proper fuel, drivers and all the rest.
In short, it’s a challenging and tricky business trying to make something new into something valuable, to beat others who happen to be trying for the same thing.
1. G-forces aren’t pleasant.
2. Another tricky bit here is the challenge of not setting your sights too low (like possibly early social networking sites did), to envision the best end point possible. And useful here is a framework from AirBnB founder, Brian Chesky, called the 11-star experience. Think of a 5-star hotel, and how it compares to a 1-star hotel. What is the equivalent of a 5-star hotel of your product type? What about an 11-star equivalent?
3. Though likely this is the best way to do it.