• Subscribe

How close do computer models come to reality?

Speed read
Computer models:
  • See patterns beyond human perception.
  • Answer in minutes what would take humans years.
  • Match observations and demonstrate predictive power.

Darwin’s theory of evolution is simple. It requires only three necessary (and sufficient) components for the process to work: inheritance, variations, and differential survival (or 'selection').

<strong>Natural selection.</strong> One of the mechanisms responsible for making evolution happen. Courtesy Elembis. <a href='http://creativecommons.org/licenses/by-sa/4.0/'>CC BY-SA.</a>

Inheritance guarantees that anything new discovered by the process is not lost. Variation ensures that new things are being tried out constantly. And differential survival implies that differences matter — variations that help rather than hurt have consequences for the descendants of the first individual that carried that beneficial change.

But even though these principles are straightforward, how they play out is far from simple. We might be able to work out in our head how one beneficial change (say, a larger body size to withstand a predator’s assaults) can also have negative consequences (more time spent foraging to support the body weight exposes the individual to more predation). Such simple trade-offs can be captured by mathematical formulas, and their consequences can be worked out.

But in real biology, every single trait could conceivably affect every other. It’s not easy to work out the net benefit of a set of traits, either in your head or with mathematics. This is where computers come in.

Computers run through scenarios, fast

I frequently hear the phrase: 'With a computer, you can get any result you want.' But this is not true. What a computer does is keep track of things for you.

To a large extent, this is what mathematics does too. It allows us to use symbols to embody complex relationships that we can then manipulate according to strict rules.

The computer is no different, except it allows us to keep track of many more variables, and to work out the consequences of the relationships over long periods of time. Since we set strict rules, of course, we can’t get 'anything we want.' We get only what is allowed according to the rules.

But what are those rules?

In mathematics, you start with a set of assumptions, and you work out the consequences according to the rules of logic. This is still true inside a computer, but now we can also implement very specific rules — for example, the laws of chemistry, the effects of friction, or the cost of finding a mate.

Researchers in a variety of fields turn to computer simulations to help them test ideas that they can’t investigate any other way. Astrophysicists use these kinds of models to simulate how stars form. Material scientists simulate the aging of nuclear weapons to predict if they will still work in the future.

In evolutionary biology, we might ask which factor shaped a particular trait or behavior. For instance, my colleague Kay Holekamp has been observing hyenas in Kenya for over 25 years, and she’s collected an enormous data set about the hunting habits (among other traits) of these animals. But even all those observations can’t tell us why she sees what she sees in the field. The reasons may lie in pressures that the population was under in the past, or maybe the pressures manifest themselves only over thousands of generations.

<strong>Computer vision.</strong> Even decades of observation leave us with questions about why animals behave in certain ways — that's where computers come in. Courtesy Anne Engh. <a href='http://creativecommons.org/licenses/by-nc-nd/4.0/'>CC BY-NC-ND</a>.To answer questions such as “Why don’t the highest-ranking female hyenas participate in the hunt?,” we have to study the consequences of different assumptions on the long-term survival of the group.

Evolutionary theory says that only beneficial traits survive in the long run, but it can often be hard to understand how a certain trait might help. This is because of all those trade-offs, and sometimes the benefit of a trait only becomes clear after a long time. After all, evolution has had millions of years of trials, failures and successes. Even 50 years of observation might not reveal to us the long-term consequences of a set of traits and how they interact and play out in a complex world.

But a computer might work this out in minutes, as a population of 1,000 gazelles and a group of, say, 150 hyenas can be followed over thousands of simulated generations.

Matching theory to observation

In evolutionary science, computers thus are prediction machines: they answer questions like 'What would happen under these rules, given I started in this world with these starting conditions?'

In our study of the evolutionary origins of risk aversion, for example, we could ask what happens to risk aversion if the total population was large, but composed of small groups with migration between them. Running the scenario, we found that risk aversion still evolved unless the migration rate was exceedingly high.

Of course, if you start with the wrong rules, or inappropriate starting conditions, the results may not match what we observe in reality. But this is exactly what we require in the scientific process. If the predictions are wrong, then we must modify either the rules, or the initial conditions (or both).

Once we do obtain a match between the computer simulations and real-world observations, we must then test whether these rules also predict other things that we didn’t set out to test in the first place. For example, do the same set of rules also explain the observation that the spoils of a kill are not distributed equally among the hyenas?

This kind of thinking is no different from the way theory and experiment have worked to build the powerful framework of theoretical physics. In that quest, theories were laid down, for the most part, mathematically. In evolutionary biology, though, this is usually not possible simply because biology is too complicated.

Evolutionary simulations allow us to test hypotheses, but they’re not asking or even answering questions. We ask “What if,” and the computer dutifully responds: 'In this case, this is what you would get.' The computer helps us 'think forward in time' with blazing speed, and in evolutionary science this is precisely what is required to generate understanding.

This article was originally published on The Conversation, and has been edited for length and style. Read the original article here.

Join the conversation

Do you have story ideas or something to contribute?
Let us know!

Copyright © 2017 Science Node ™  |  Privacy Notice  |  Sitemap

Disclaimer: While Science Node ™ does its best to provide complete and up-to-date information, it does not warrant that the information is error-free and disclaims all liability with respect to results from the use of the information.


We encourage you to republish this article online and in print, it’s free under our creative commons attribution license, but please follow some simple guidelines:
  1. You have to credit our authors.
  2. You have to credit ScienceNode.org — where possible include our logo with a link back to the original article.
  3. You can simply run the first few lines of the article and then add: “Read the full article on ScienceNode.org” containing a link back to the original article.
  4. The easiest way to get the article on your site is to embed the code below.