All Astronautical Evolution posts in 2019:
The Holy Grail of Space (October)
Return to the Moon, 50 Years On (August)
SpaceX Dragon 2 Success (April)
Killing the Doomsday Fallacy (Feb.)
All Astronautical Evolution posts in 2018:
The Atheism Question (Oct.)
The Religion Question (Sept.)
I, Starship (June)
Back to 2017:
Comments by Alex Tolley (Oct.)
Elon Musk’s “Great Martian” (Oct.)
What is a Supercivilisation? (Aug.)
Back to 2016:
New in 2020:
2022: What’s to do on Mars?…
2021: New space company Planetopolis…
2020: Cruising in Space…
2019: The Doomsday Fallacy, SpaceX successes…
2018: I, Starship, atheism versus religion, the Copernican principle…
2017: Mars, Supercivilisations, METI…
2016: Stragegic goal for manned spaceflight…
2015: The Pluto Controversy, Mars, SETI…
2014: Skylon, the Great Space Debate, exponential growth, the Fermi “paradox”…
2013: Manned spaceflight, sustainability, the Singularity, Voyager 1, philosophy, ET…
2012: Bulgakov vs. Clarke, starships, the Doomsday Argument…
2011: Manned spaceflight, evolution, worldships, battle for the future…
2010: Views on progress, the Great Sociology Dust-Up…
Index to essays – including:
The Great Sociology Debate (2011)
Building Selenopolis (2008)
|Site home||Chronological index||About AE|
Killing the Doomsday Fallacy
This subject came up again recently. I wrote the following three years ago, and it’s probably time I published it. I don’t have time to re-read the various Doomsday Argument articles again now, so I’m relying on readers of this post to point out any errors I may have made in interpreting what they originally wrote.
A spurious bit of logic-chopping
The Doomsday Argument was originally posed by astrophysicist Brandon Carter in 1983, independently by H.B. Nielsen (or Nielson) in 1989, and subsequently taken up by John A. Leslie, J. Richard Gott III, Nick Bostrom and others.
I first discussed it in AE issue 84 (July 2012), and would like to return to it now with a more economically worded analysis. For this purpose I have re-read two key papers, by Leslie and by Gott (referenced in full at the bottom of this page).
It is intuitively obvious that the Doomsday Argument is likely to be a fallacy, because, like an ancient Greek bit of logic-twisting such as Zeno’s Paradox, it bases a major claim about the physical world on a chain of abstract reasoning with no reference to relevant real-world observations. It is therefore an implausible source of new knowledge.
On closer examination, the Doomsday Argument is found to contain four specific errors of reasoning. It is therefore referred to henceforth as the Doomsday Fallacy (DF).
The assumption that intelligent observers are sharply bounded in time
Gott defines an intelligent species as “one that is self-conscious and has the cognitive abilities to reason abstractly, think about the future, create art and so forth” (p.315). He speculates that such species are currently being formed in the universe at a uniform rate, and adds that our species, Homo sapiens, is the only species on Earth to fit this description so far. Our species is roughly 200,000 years old, and from this he estimates a total longevity for our own plus any possible intelligent successor species (including hypothetical future machine species), from start to finish, of between 0.205 million and 8 million years, with 95% confidence (p.316).
The impression given is that rational intelligence switched on 200,000 years ago. This would not be correct. There was no act of special creation of human intelligence at one point in the past. My mental powers are roughly the same as those of my parents, which were roughly the same as those of their own parents, and so on back indefinitely into the past, back to Homo erectus, back to Australopithecus, back to our squirrel-like ancestors who lived alongside the dinosaurs, back further still to the simple information-processing powers of the first living cell.
Certainly there was a period of rapid brain growth in our ancestors a couple of million years ago, but it only looks to us like an evolutionary jump because we are seeing a compressed timescale: the reality at the time would have been that generation succeeded generation with no perceptible biological change. The evolutionary jump was in reality an accumulation of small changes over thousands of generations, only visible on a long-term view.
The abilities which Gott mentions are in any case hardly evenly distributed among the people living at any one time. Some individuals may exercise abstract reasoning, while at the same time others dismiss them as useless intellectuals. What proportion of a population would have to be interested in such things as discussing the Doomsday Fallacy in order for Gott to qualify that population as a whole as “intelligent”?
Again, where does one place the end-point of an intelligent species? Clearly, with so much talk about existential risk nowadays, the assumption is of a fairly sudden wipe-out of civilisation and/or of the human species itself. But a look at our evolutionary past shows that when nature introduces an innovation such as simple microbial life, multicellularity or animal life, that new capability tends to persist over evolutionary time and act as a launch pad for the next innovation down the line.
It is logically impossible to rule out a scenario in which biological intelligence will in due course follow the same pattern. The reasoning of the DF would then produce the prediction that some new form of life currently unknown but growing out of the unique intellectual capabilities of Homo sapiens will appear between 5 thousand and 8 million years in the future, with 95% confidence (as if one could ever be confident in the face of such unknowns!). How do Brandon Carter and the others measure the relative probabilities of their own scenario vis-à-vis this one? Clearly, any numerical estimate would be meaningless.
If the new form of life in our evolutionary scenario were to consist of machine intelligence, or of a highly networked synthesis of biological and machine intelligence (as, say, our bodies are a synthesis of eucaryotic and procaryotic cells), then the future boundary of human reason in its current purely biological form might come well before the 5 thousand year point.
Of course Gott rejects this kind of hierarchical structure to evolutionary reality, and only recognises one form of intelligence: our own – or more accurately his own.
The assumption that we are typical
So let us grant the supposition that the beginning and the end of the history of intelligence can be unambiguously bounded. Possible moments to do that might be the time when symbolic spoken language first appeared, and a speculative future time when humans die out without leaving any intelligent successor species, thus breaking the pattern of cumulative evolutionary development. The DF then makes the assumption that the present time in history is a typical one. But real-world observation proves the opposite: that the present time is extremely unusual.
Thus in the late 20th to early 21st century the expansion of human civilisation through an explosion of science, technology, population and the economy reached a unique point of crisis, where continued growth on Earth alone was seen to be impossible, but that same growth opened up a new possibility of expansion into the rest of the Solar System. This immensely significant decision point, unique in the history of any species, has now been arrived at for the first time in human history, the first time in the evolution of terrestrial life, and possibly the first time in the history of the astronomical universe.
Furthermore, as discussed earlier, physics and astronomy indicate that life will continue to be possible for a future period on the order of ten trillion years or more. Since the universe is at present only a little over ten billion years old, our present location in time is within the first one-thousandth of the history of the universe. Unless life is to be forbidden for almost the entire future history of the universe in order to satisfy the logical demands of a small number of people living on one infinitesimal speck of cosmic matter (science knows of no physical or biological reasons for such a ban), our general situation in cosmological time is highly untypical: a major failure of the much-vaunted Copernican Principle.
The assumption of time travel
But let us grant the supposition, however, that new manuscripts are discovered which prove that the ancient Roman, Mayan, Chinese and other empires also landed on the Moon, invented nuclear power, developed a global economy and so on, as did many other civilisations of other species before humans ever evolved. And let it be assumed that new laws of physics are discovered which make it certain that any kind of life will become physically impossible anywhere in the universe after another few billion years. The DF makes a third error when it implicitly assumes that people are prenatal time travellers with potential access to the entire span of human history (however that span may have been defined).
Gott: “Knowing only that you are an intelligent observer, you should consider yourself picked at random from the set of all intelligent observers (past, present and future) any one of whom you could have been” (p.316); “This paper only points out and defends the hypothesis that you are a random intelligent observer” (p.319).
This hypothesis models the selection of a person’s date of birth, or equivalently their position on a chronologically numbered list of all humans ever born, by that person putting their hand into an urn containing numbered balls and pulling one out at random.
But in reality nobody chooses their date of birth; rather the circumstances of the time determine the people who are born at that time. There is no disembodied spirit outside time which must pick up a numbered ball at random in order to materialise and live at a particular time in the world; rather the balls in the urn metaphor are themselves people who have no power to choose to be other than what they are.
The assumption of high probability
But let us grant the supposition, however, that studies into spiritualism reveal that each human does in fact have a soul which transcends time and may be born at any randomly chosen point in history. The DF makes a fourth error when it states that any individual may apply probabilistic reasoning to deduce that they are most probably living in the most heavily populated region of the human population graph. True, the majority of such people will be correct, but a minority must necessarily be wrong.
The question then becomes: do we here and now belong to the majority or to a minority? But probability and statistics are intrinsically unable to answer questions of this nature. They are completely impersonal: they are valid for an ensemble of cases but contain no information about any particular specific instance.
If therefore we are today living at a time of high relative population, then we represent a typical case, from the point of view of an observer outside time. But if we today are at a time of relatively low population, then we are by that same token outnumbered by a far larger group of people who are more typical than ourselves, from the point of view of that same observer. Probabilistic reasoning is equally satisfied either way.
What are the facts? People born into the present century see a rising population curve in their past, but do not know where that curve will go in the future. Examination of the actual physical circumstances of the human race reveals only that the population may collapse in the near future, or it could in principle level off close to its present value and maintain that level (on Earth alone) for a period of millions of years, or it may continue to rise (as interplanetary and galactic colonisation proceed) and ultimately maintain much higher levels (of various intelligent post-human species) for a period of billions to trillions of years. All these possibilities are certainly physically possible, and since the shape of the graph to date would be identical in all three cases it therefore contains no information about which future shape will in fact occur.
Repeatable versus unrepeatable experiments
The point just made is so critical to understanding the DF that we shall now rephrase it from a slightly different angle: we can say that the fundamental error behind the DF is to confuse a repeatable experiment with an unrepeatable one.
If I say: there is a 90% probability of pulling a green ball out of the urn, and a 10% probability of red, this has a definite practical application. If we pull, say, 100 balls out of the urn, we may expect that 90 of them will be green while the remaining 10 will be red. Of course we may get 91 to 9, or 89 to 11, or even 92 to 8, and so on, but the frequency with which such deviations from the mean may be expected are themselves statistically tractable with the Poisson distribution.
But suppose that I can select only one ball in my entire lifetime. One is tempted to say: the probability that I will get green is still 90%. But this is incorrect. This particular experiment only draws one single ball, therefore 100% of the balls drawn must be either green or red. The probability of green is either exactly one or exactly zero, and it is not known which in advance. A fractional probability only acquires meaning when multiple experiments are carried out.
The DF depends upon supression of this point. It achieves its effect by taking a plausible statement such as: “99% of all human observers over the span of history past, present and future live at neither the extreme beginning nor the extreme end of history, but somewhere in between”, and reinterpreting it to mean: “That 99% probability gives me the information that I do indeed live in a more typical period of history and do not in fact live at the extreme beginning or the extreme end of history.”
But no such information can be extracted. It is perfectly acceptable for one particular case (such as ourselves here and now) to have a low probability, so long as that case is heavily outnumbered by other cases which have a high probability. But this is a tautology: the one automatically implies the other.
Since those other cases, if they exist, exist by definition in the future, there is currently no way for us to tell whether our own state is high or low probability from the point of view of a future observer. From our present-day viewpoint the future has not yet been determined.
Whether we believe that intelligent life will die out within a short period on the cosmological timescale, or whether we believe it will live long and prosper, can only be gauged by examining physical reality. That reality is that sufficient natural resources exist for our heritage to last for trillions of years into the future. Therefore such a possible future is now open to us, and if our descendants do not in the end fulfil that early promise, that failure will be due to actual physical or social or economic factors, not to any logical requirement that the inhabitants of planet Earth in the early 21st century must find themselves in the middle of history.
Such a requirement is, as we have seen, completely illusory.
To sum up, the DF rests upon the following assertions:
Obviously, all four statements are incorrect, and the Doomsday Fallacy is therefore multiply refuted.
John Leslie, “Time and the Anthropic Principle”, Mind, New Series, vol.101, no.403 (July 1992), p.521-40.
J. Richard Gott III, “Implications of the Copernican Principle for our Future Prospects”, Nature, vol.363, 27 May 1993, p.315-19.
Please send in comments by e-mail.
Interesting and relevant comments may be added to this page.
|Site home||Chronological index||About AE|