Today was my graduation from Cornell, but since I’m not a fan of ceremony, the topic for today is completely different: a subset of selection bias known as observer selection.
Selection bias in general is selecting particular data points out of a larger set to distort the data. For example, using the government’s own NOAA website (National Oceanic and Atmospheric Administration), I could point out that the average temperature in 1934 was 54.10 degrees Fahrenheit, while in 2008 it was 52.29. Clearly from these data points, the US must be cooling over time. The problem with the argument is, of course, that the two years 1934 and 2008 were chosen very carefully: 1934 was the hottest year in the earlier time period, and 2008 was the coolest year in recent times. Comparing these two points is quite meaningless, as the overall trend is up.
Observer selection is when the selection bias comes from the fact that someone must exist in a particular setting to do the observation. For instance, we only know of one universe, and there is life in our universe—us. Could it have been possible that our universe had no life?
The issue with trying to answer this question is that if our universe indeed had no life, then we wouldn’t exist to witness that.
“The anthropic principle: given that we are observing the universe, the universe must have properties that support intelligent life. It addresses the question “Why is our universe suitable for life?” by noting that if our universe were not suitable for life, then we wouldn’t be here making that observation. That is, the alternative question, “Why is our universe not suitable for life,” cannot physically be asked. We must observe a universe compatible with intelligent life.”
The point is, there may be millions, billions, or even an infinite number of universes. But even if only one in a trillion were suitable for life, we must exist in one of those. So our universe is not “fine tuned” for life, but rather, our existence means we must be in a universe that supports us.
A list of observer effects:
- The anthropic principle, as above. Our universe must be suitable for life.
- A planet-oriented version of the anthropic principle: Earth has abundant natural resources, is in the habitable zone, has a strong magnetic field, etc.
- A species-oriented version of the anthropic principle: Our species is very well adapted to survive. If we weren’t, then we wouldn’t be thinking about this.
- There are no recent catastrophic asteroid impacts (the last one being 65 million years ago). If there were, then we again wouldn’t be observing that.
- The same goes for all natural disasters. No catastropic volcano eruptions, no nearby supernovae or black holes, etc.
- The same goes for apocalyptic man-made disasters. Had the Cold War led to a nuclear exchange that wiped out humanity, we would not be able to observe a headline that said, “Nuclear Weapons Make Humans Extinct.” Thus, we must observe non-catastrophic events in the past.
- Individual life follows this as well. Say you had a life-threatening illness or accident in the past, but you’re alive now (of course, given that you’re reading this). Given that you’re alive now, you must have survived it, so to the question, “Are you alive?,” you can only answer yes.
All of these are strong observer effects, in that they are absolute statements and not probabilistic ones, i.e. “Our universe must have life,” and not “Our universe probably has life.”
There are numerous other observer effects that are probabilistic but can be still very significant. For example, given that you are reading this, you are more likely in a literate country than in less literate one. Moreover, the probability would be higher than that if I did not know anything about you.
In this post, I mentioned the example of democracy in political science. In summary, political science has a lot more to say on democracy than on any other form of government. Is this because we are personally biased towards democracy? Not necessarily. In a less open system, fields like political science might be forbidden from research (or academia is rated less important), and hence there are no (or few) pro-totalitarian political scientists. Hence, we end up seeming to favor democracy.
We also know that history is written by the victors. But a related historical example is the rise of strong states combined with the rise of liberalism and progressive thoughts in the Modern era. Namely, states in which liberalism arose (England, France) tended to be strong states. A weak state adopting progressive measures would be wiped out by a stronger one. Hence, history is also analyzed by the victors.
So what can you do about observer selection? All we can do is try to be aware of it and introduce corrections to study a full set of possibilities rather than the subset we are in by being a particular observer. For instance, if we were just using historical data of natural disasters, we would be underestimating the actual probability of a catastrophic disaster, as we live in a time where none could have occurred for a while.