Thursday, August 28, 2008

Long-term predictions and wishful thinking

I was trying to understand the ideological framework that leads so many people in the West to dismiss Russia as a "country in decline" exactly at the moment when it's rising from its knees, regaining its lost self-confidence, and surpassing the world in the GDP growth as well as in its military flexibility and readiness.

It seems that this ideological framework is a conglomerate of at least five general fallacies that also appear in many other contexts and that I would like to analyze. The only "objective" comment against the impressively good and promising condition of current Russia that I have fully understood is its negative population growth. Well, in 2007, there were 11.3 births per 1,000 people in Russia, more than 10.7 in the U.K. but less than 14.2 in the U.S. At any rate, there's not much difference here. Because of the lower life expectancy, there are more deaths per 1,000 citizens in Russia. But all these numbers, especially the birth rate, are flexible and can change.

However, the critics of Russia seem to be certain that in 2050 or so, Russia will be going down. Moreover, this idea influences their opinions about the "right" behavior that the Western politicians should adopt today. Fine, so let me first enumerate the (corrected) fallacies, before I will discuss them in detail, together with many other examples of these fallacies:
  1. It is irrational to try to predict demographic subtleties in the very long run, e.g. in 2050
  2. Even if it were possible to predict these things, they shouldn't significantly influence rational decisions in 2008, especially not our "perception of justice"
  3. Even if the trends in 2050 could be predicted, it is irrational to assign them with positive or negative moral labels
  4. Even if a quantity could be expected to be in "decline", it surely doesn't mean that you can imagine that it will be zero any time soon
  5. When a rational person applies a certain kind of predictive methods and assumptions to one nation (or other subjects or objects), he should do so consistently with other nations (or subjects or objects), too.
Fine. So let me discuss these issues in detail.

1. Very long-term predictions are irrational

The most obvious example of this fallacy is global warming. Certain people want to plan the society up to the year 2050. In fact, the next IPCC report will want to prepare plans up to the year 2300. The latter number displays lunacy that is 70 times higher than the lunacy of megalomaniacs such as Iosif Vissarionovich Dzhugashvili who had 5-year plans only.

The society is so complex that it is unreasonable to make reliable or even semi-reliable predictions into a future that is as distant as the year 2050. What will be the relevant questions for the society in 2050 is almost completely unknown today. To see that it is the case, just look 50 or 100 years into the past: could the people who lived then be wisely planning our lives or to realistically expect what we would be talking or worried about?

I am convinced that every sane person must be able to see that the answer is clearly No. The typical time scale where the key "fashionable" problems of the society mostly change is comparable to 5-10 years. There are many other questions that don't change so quickly - for example, the qualitative patterns of planetary orbits don't change for billions of years - but the humans are rather unstable, constantly changing physical systems. There are general aspects of the human behavior that do not change for centuries; but there are many more fashionable things that change in a few years.

2. Very long-term predictions have nothing to do with current decisions, anyway

Even if it were possible to predict how the society is going to look like in 2050 or 2300, such a vision cannot influence the decisions in 2008 as long as the latter are rational. The impact of a particular decision made in 2008 on events in 2050 or 2300 is both uncertain and tiny in average. The reason why it is uncertain is that the society is a very complex system and the consequences of a decision that we can make right now will be "thermalized" - i.e. challenged, compensated, and overcompensated by many, often deliberately opposite decisions in the future.

From a human perspective, 50 years or 300 years are extremely long time intervals. Many more decisions and their revisions will be made between 2008 and 2050 and choosing decisions in 2008 according to their impact on the life in 2050 is a demonstration of someone's inability to see the difference between different time scales. Decisions about XY should only consider the future YZ years where YZ is the typical time scale at which XY is changing - for example measured by the time scale at which things similar to XY were changing in the past. For political things, the time scale is comparable to 5 years and is not far from the time scale at which a politician has the credentials to change something.

3. If something is likely to happen, it doesn't mean that it's a good thing, and vice versa

Americans are the best examples of optimists. Optimism means to believe in a high positive correlation between the future events that we consider "good" and the future events that we consider "likely". Pessimists have the opposite correlation and the people who rationally realize that there's no clear correlation between "good" and "likely" are called realists. ;-)

How does an optimist achieve such a correlation? Well, there exist two basic methods. Either he looks at the things that are "likely" to happen, according to his estimates, and he decides that they are pretty good: if we exaggerate a bit, such an optimist simply accepts the future, whatever it is. These optimists suffer from "lethargy" but they can live with it happily.

The other possibility is that the optimist looks at the things that he considers "good" and decides that because they are good, they must also be "likely" to happen in the future. We say that this subgroup of the optimists suffers from a fallacy called "wishful thinking". For example, feminists may decide that it is "good" for the percentage of women in physics to increase (because of their irrational ideology), and because it is good, many of them also decide to believe that it is a likely outcome (which is a combination of two irrational steps).

Incidentally, the pessimists fall into these two basic groups, too - even though we usually don't speak about these subtleties. ;-) Recall that pessimism is a belief in a high correlation between events that are "bad" and events that are "likely". A pessimist may suffer from a "reverse wishful thinking": because something is bad, it is likely to happen, he thinks. Most of the pessimists who display reverse wishful thinking have an agenda because the "bad" thing is actually a very good thing for them personally: most typically, the "bad" outcome could help them to show that they were always right that something sucks! ;-)

There also exists the other type of pessimists. They think that because something is likely to happen, it must be bad. These people are "inherent grumpy sourballs". For example, Peter Woit knows that the percentage and relative importance of string theory in theoretical physics is going to increase, so it must be a bad thing. :-)

A realist doesn't believe either of these fake correlations: a future event can be likely or unlikely, good or bad, and there is no good a priori reason for a correlation between these two types of adjectives. The realist's position is clearly the most rational one but that doesn't quite mean that the realists always have the highest survival advantage. I believe that in the long run, it is probably healthy for individuals to be optimists. If you know about contexts where the pessimists have a systematic long-term advantage, I would love to hear about them.

Besides the optimism/pessimism question, I would like to stress that a falling population of Russia, even if it continued, is not a "moral vice".

4. An increase or a decline is something else than a catastrophic increase or a catastrophic decline

Another very frequent fallacy is to identify a "nonzero trend" with a "catastrophic trend". A nonzero positive and negative trend is something that is almost guaranteed to be observed for every quantity we can imagine. On the other hand, a catastrophic positive or negative trend is something very unlikely that only appears in exceptional cases. Many people believe that if they demonstrate that a quantity is not going to be constant, they have also proven an imminent catastrophe. Instead, they have only proven an inconsequential truism.

Again, global warming is the most obvious example. The fact that the average temperature on Earth has been changing and is likely to change by 2300 is often interpreted as a global catastrophe. Needless to say, these are very different things. There are often up to 123 orders of magnitude of difference in between these two possibilities. The vacuum energy density is positive and one could reasonably expect it to take the catastrophic, Planckian proportions that would be enough to kill even every single proton. ;-) However, the real, observed vacuum energy density is smaller by 123 orders of magnitude and it is harmless for life.

Some people only use the "continuous" thinking for a while but once they see that something is nonzero, they adopt a "discrete", black-and-white thinking in which a positive value is the same thing as a divergent catastrophe. It is not the same thing. For example, the 0.6 °C of warming that can be expected by 2100 will be as harmless as the same warming during the 20th century. To get a more decent weather on most of its territory, Russia would need a warming by 15 °C or so. ;-)

In the case of demographics of Russia, the opponents of Russia, after they (irrationally) decide that they have proved that Russia's population will be declining by 2050, seem to conclude that Russia will effectively absent and it shouldn't play any role, not even today. You see that this reasoning unifies most of the fallacies discussed in this article: irrational long-term predictions, a wishful thinking, a wrong identification of the interpretation of quantities in 2008 and 2050, and many others. One of the fallacies is that they think that if the population of Russia is going to decrease, it is effectively zero.

But even if the population of Russia decreased from 140 million to 120 million by 2050, it will clearly fail to change any qualitative features of the discussion about the role of Russia in the world. The GDP per capita or the military power per capita may change more dramatically (especially if the oil price is going to increase) so that the population decrease will be overcompensated. Moreover, you don't know what will happen with the population of other countries, either.

The only known possible way for a mad politician to be able to "forget" about a large nation altogether has been a "final solution". Unless you want to repeat something along these lines, be sure that Russia will be an important player on the international scene for the decades to come.

5. Analyses of situations in the world should be self-consistent and follow unified rules

This brings me to the last point: consistency. That's an immensely important value for (good) theoretical physicists. A theory should be both internally self-consistent, i.e. to predict a unique answer to each well-defined question that belongs to an a priori allowed interval, and the theory should also be used consistently for all relevant phenomena.

If someone proposes a new physical law, he must insist that the law should be good for the description and predictions of a whole, well-defined class of phenomena. By the adjective "well-defined", I mean that the boundaries between the phenomena where the law should be trusted and those where it doesn't have to work should be known, at least approximately.

If a hypothesis disagrees with an observation of a phenomenon in this class, it is falsified. Sometimes, it is enough to change the definition of the "class" where the hypothesis should hold i.e. to make a more modest claim. However, in many cases, it is impossible and the hypothesis must be simply abandoned. For example, a "theory of everything", i.e. a hypothesis about the dynamics that governs all fundamental forces and particles in the Universe, should be valid universally. It follows that every disagreement of such a candidate theory with a (correctly made) observation means that the candidate theory is instantly falsified.

If you can only choose a few hints, a few cases where your theory looks OK, but you have to neglect its other, bad predictions - typically predictions that are obviously wrong even without any detailed measurements (e.g. anomalous or divergent ones or negative probabilities) -, it is simply not good enough as a starting point for a candidate theory.

In the case of the population growth, the critics of Russia clearly don't look at the world consistently. China has had a one-child policy to regulate its population growth (which was still positive). However, it is clear that new conditions and policies can change many things. For example, the population growth in the Czech Republic used to be negative a few years ago, too. That's no longer the case. Long-term predictions of such dynamics is impossible because it will be affected by all kinds of factors, including the individual psychology of typical citizens, the national wealth, migration, and government's policies.

If we look at Russia from different angles than the population growth, we find many other examples of inconsistencies. For example, a commenter has hysterically criticized Russia for having signed the Kyoto protocol. That doesn't look like a consistent criticism given the fact that 181 other parties have ratified the protocol, too. ;-)

The same comment applies to Russia's recent successful operations in the Caucasus. It may look cute to someone to criticize Russia for these things except that NATO has done pretty much identical things in Serbia and its province Kosovo. And Russia's approach has arguably been more elegant, smooth, and peaceful. Also, the Ossetians and the Abkhazians are more justified to demand their own territory because they are authentic unique nations that have lived at those territories for centuries while "Kosovars" are just a fake nation, a redundant copy of the Albanian nation that was allowed (and, indeed, encouraged) to propagate and reproduce (by Mussolini, among others) more than what would have been appropriate. Legally, the three cases are virtually isomorphic.

When the violations of consistency are driven by someone's own personal interests or his image, the particular case of inconsistency is known as "hypocrisy".

Unfortunately, there are gigatons of it in the contemporary Western societies.

And that's the memo.


  1. Lubos,

    I think your classification of pessimists/optimists/realists deserves some elaboration. Your classification (and it may be just due to the desire to keep the text short) is unconditional. However, they should be evaluated conditionally on what we know and do.

    For instance, if I keep shooting at somebody with a gun, and assume a high probability of a bad event of the person dying, I can hardly be considered a pessimist.

    Similarly, the history of mankind gives us a lot of reason to assume that the living conditions of people will in the long run on average keep rising, as will for instance our understanding of science. If I consider this a good thing, I can hardly be labeled an optimist when I believe that these trends are likely to continue into the future.

    So the definition of an optimist/pessimist should hold regarding the correlations after I singled out the impacts of known observables.

    You are actually aware of this when you write "I believe that in the long run, it is probably healthy for individuals to be optimists." In the long-run, unconditional optimists according to your definition are actually on average pretty realistic, given mankind's average long run experience.

  2. I agree, Jarda, "good" or "bad" events in the future should be defined as "better" and "worse" than what can be deduced to be the average expected outcome in the future.