You’re a modeler, but you might not know it

Our CTO gave a presentation about modeling last week: what it is and why we do it. We use the word “model” a lot at my work. We use it to refer to ontologies, code that applies logic or mathematical analysis to data, and the concepts structuring an application. Little wonder that “model” starts to feel vaguely defined, even for people who build modeling software.

Roy’s talk described his personal history of modeling, ranging from building model airplanes to modeling shear and moment in architectural structures, to computer graphics, to the semantic modeling we do at Thetus. Roy gets around. Or, as he puts it, he’s been lucky enough in his life to show up at places where interesting things are happening.

His argument, and mine, is that modeling is a basic conceptual process we use to structure and make sense of the world around us. Models let us abstract, predict, and filter out the mass of information that surrounds us. And not just digital information: our analog, real-world environment sends us a continuous stream of information to our senses to process. This is the original too much information problem, and we couldn’t handle it without models. In my work, I build models to describe things in the world. This its own art form, but not one that comes naturally to most of us. Which brings up the question: if models are so fundamental to the way we think, why is it so hard to create a good one? And what exactly is a good model? Are there distinctive characteristics that let you know a model is good, or do you just know a good one when you see it?

Probably part of the reason it can be hard to model a domain is exactly because modeling is ingrained in our cognitive processes. Not only can we not articulate how we do it, we’re often not aware we’re doing it at all.

Take the sound English speakers make for the letter p. Most of us would say that people make this sound, and and we hear what they say.  In actuality, speakers make a range of sounds that hearers map onto the phoneme for the English p.  A Thai hearer maps several of those sounds into separate phonemes instead of just one. We don’t think of hearing speech in terms of models; we just think we’re listening to what objectively is. But in practice, we carry models of our phonemes with us that we use to interpret what we hear. We use the model to structure and filter physical reality. Fundamental.

The question of what makes a good model is a bigger one, and one I spend a lot of time thinking about. More on that to come.