Type I and Type II errors — a practical perspective
Nearly 3 million years ago, a group of primordial humans were resting at the edge of a Savannah (grasslands) and a forest. The trees were lush on one side and on the other side, grass, as tall as 5 feet filled the ground. Naturally, visibility was low.
The head of the group was lolling around, while the others rested. He was keeping watch. He suddenly heard a rustle of the leaves from the forest end. The head of the group noticed it. He thought about it, but his lazy brain decided it that it was probably just the wind. He turned his back and continued lolling around. A tiger (whose species is now extinct) leapt out of the woods and promptly killed the head of the group. The rest of the group scrambled away to safety, barely.
The head of the group had done an error. He had failed to detect the presence of the tiger despite the ruffling of the leaves. This type of error is called a ‘false negative’ — failing to recognize a true effect. This is also called as a type I error. People who performed type I errors mostly ‘died’ in prehistoric times.
Now imagine another group of people resting at another edge of a Savannah and forest. They also contain people from the group who witnessed their leader being eaten alive by a tiger. Now suddenly they hear a rustle of the leaves and are promptly reminded of the fearsome tiger. They scramble and leave the resting place. However, the rustle of the leaves this time was just because of the wind. This time, there was no tiger.
The group had done an error. They had detected the presence of a tiger, despite the fact that there was no tiger. This type of error is called a ‘false positive’ — recognizing an effect that is not present. This is also called as a type II error. People who performed type II errors ‘did not die’ in prehistoric times.
We are all descendants of people who performed type II errors, basically because it helped us. Let’s imagine that out of 100 times ruffling of the leaves happened, maybe once it was because of a tiger. People who performed type I errors were not affected 99% of the time, but ‘died’ when there was a tiger. People who performed type II errors were affected 100% of the time, as they decided to leave, but they lived! Though this is an oversimplification of the idea of type I and type II errors, the overall picture is quite the same.
We subconsciously perform type II errors every day. We do not trust new people, because it is possible that they might cause us ill, even though the probability of that happening is very low. Similarly, we do not trust new products. We do not trust new leaders. Most importantly, we do not trust new ideas, because they might go horribly wrong. However, most of the time we are performing a type II error. Most of the things we fear would happen do not happen at all. Also, we often over exaggerate the risk of something happening. Type II errors prevent us from exploring beyond our comfort bubble. In prehistoric times, when the risk of performing a type I error was huge (usually death, because you didn’t detect the tiger), type II errors served us well (you ran away even though there was no tiger. You made a mistake, but lived and reproduced). However, in the modern world, where circumstances and risks have largely reduced (we usually do not die, that is), type II errors are just a huge hindrance to make the world a better place.