BY: NICOLA DENNIS

When we were building our house, we were confused by the word “profile”. That piece of paper - oh it's a profile. That 3D computer graphic, that grid of string that we mustn’t touch, the walls of our house - all referred to as "profile". Kindly itemise your expenses for the bank’s profile. Can we use your photos to build our social media profile? Profile the profile for the profile, please.

In the world of science, it is a similar story for the word “model”. In the broadest sense of the word it means something along the lines of “organising ideas”.

In more practical terms it could mean anything. It might be an excel spreadsheet with a few calculations in it. Or, data could be examined with a statistical model to see if, say, one species of pasture grows better than another. A model could be a whiteboard full of scribbles. Or, it might be an animal. At the university, we had “model animals” which were genetically modified with human genes.

If you hang around with people who really enjoy the sound of their own voice, the model isn’t anything at all. It's a metaphor for the work that they want you to believe has taken place - “we modelled this project on a tree. Here are the branches filled with empty promises. The roots represent the money you have parted with to climb up the trunk of enlightenment.”

In my previous job, I was in the business of programming simulation models. I don’t even know if that is the correct term for them because we just called them “models”, but it is what I will call them for the sake of this article.

Simulation models are the types of models where you put in some information and they “tell” you what is likely to happen. Think Farmax for farm financials and pasture growth or Overseer for nutrient leaching and greenhouse gas emissions.

To be clear, I did not have a hand in either Farmax or Overseer. The models I worked on were obscure ones that simulated genetic scenarios for clients in charge of national breeding programmes. But, I expect the process of knocking these models together is much the same regardless of what they are simulating.

WHAT IS THE BENEFIT OF A MODEL?

Aside from being a fun way to jaunt around on someone else's dime, simulation models are useful to test a bunch of different scenarios.

A physical experiment on the genetics of cow fertility might take decades to seek out all the cows with the correct genetics for the experiment and then follow them through their lifetime. And, once you were done, you might wonder what the results would have been like if you had made different management decisions.

So, using a simulation model speeds things up. It is also one of the cheaper things you can achieve in the science world. "Cheap" is definitely in the eye of the beholder, but, in general, software is inexpensive compared to field trials.

Software is also easily copied and distributed. So, it is easy enough to hand the model to others to use. If the model is robust enough to be handled by the general public then it can be another tool in the farmer toolbox.

It's not all rainbows, however. Some of the weirdest accusations and some of the most bewildering government policies begin in a model. Because the model holds its user (and anyone who might disagree with them) a little further back from the science in use, some very strange statements can become fixtures in the public arena. For example, "Scientists say a single hamburger patty uses 2500 litres of water…"

WHEN IT ALL MAKES NO SENSE

For every complex problem, there is a simple and elegant solution that will never work. The kind of solution that makes a ton of sense on paper, but breezes right over the important inner-workings of the problem, or incentivises the wrong behaviour and makes the situation worse.

Our wide-eyed model can get us into a bit of trouble here because it really is just an echo chamber for scientific theory if used in the wrong hands. “Yes Master” the model says, helpfully, “I took all the information you told me you liked and proved, once again, that your thinking is correct”.

If the model has enough levers (looking at you, Overseer), then a shameless operator can push buttons until they are rewarded with their desired answer, stamped with a vague scientific seal of approval.

Models should not be the ultimate dictator of reality, or worse, used to police government regulations. But, because they are at the cheaper end of science delivery (i.e. often much more affordable to use than monitoring real world effects), there does seem to be an over reliance on models.

When used correctly, models are a great way to foster communication and collaboration amongst scientists from different disciplines. That in itself is a huge achievement because it rarely happens by default.

However, when it comes to the sharp and pointy litigation end of fixing real world problems, simulation models should not be the only voice in the room. The room needs to include real life humans providing real life observations and robust discourse. Even if it means a whole bunch of these boffins are abusing the word “model”.

HOW TO BUILD A SIMULATION MODEL

First, one scours the literature. One of the models I built was for simulating cow fertility. So, I dug deep into science papers to answer questions such as “how heavy does a heifer need to be to hit puberty?”, “What is the heritability of cow body condition scores?”, “How does milk production affect fertility?”, etc.

When this was complete, there was a kind of skeleton of a model that would simulate a herd of cows going through their lifetime. Except there were lots of things we didn’t know because the science wasn’t there yet, such as “do cows with lower fertility breeding values have longer cycle lengths?”, or “How long does an unsuccessful cow pregnancy usually last for?” So, these things had to be guessed in the first instance.

That was the fun bit. I got to travel to Europe to pressure dairy cow fertility experts into giving their best guesses. And then, it all had to be checked and calibrated against data from real cows.

As the science caught up, we were able to see how close we had gotten to the real answer with educated guesses. It was pretty damn close even if I do say so myself.