Expert: James Gort - 1/22/2009

Question
Thank you for your time.

I had this article delivered in my mail today and I'm rather concerned on what is stated here!
Would you please take a look and give me your honest opinion.

http://arxiv.org/ftp/arxiv/papers/0810/0810.5515.pdf

Thanks
Phil

Answer
Hi Phil,

First, thanks very much for that link.  My first impression is that there's a lot of pages to state the obvious.  I find nothing wrong with the methodology of computing probabilities, except this is all well-known.  Any computed probability MUST take into account uncertainties in the models or factors which go into estimating the probabilities.  So any computed probability MUST have a "delta" probability to show the uncertainty.  Many public news reports (and even scientific journals) leave out the important "delta".

As an example, I do a fair amount of work on IT security risks.  The next few paragraphs talk about IT risks (but relates them to the paper).   

Risk is a probability that an asset (of a given sensitivity) will be compromised by a threat agent (considering all possible threat vectors), given a set of system vulnerabilities and deployed safeguards.  It is based only on the following factors:

•   probability of an attack attempt (note:  this implies some knowledge or assumption about the threat agent’s capability, motivation, and effort, and some knowledge of new vulnerability definitions, since the threat agent could be motivated by the knowledge that some types of vulnerabilities could exist).  NOTE:  In this case, an estimate of capability, motivation, and effort all have uncertainties (their own deltas) which would have to be included to determine the delta for the probability of an attack.

•   the damage that could result due to a successful asset compromise (note:  this requires an asset value, which may depend on the type of threat agent – a hacker may cause less damage by a compromise of troop movements than an enemy agent might).  NOTE:  The paper talks about catastrophic consequences like the earth being destroyed, so the potential "damage" estimate would be very, very high (but still would have its own delta).

•   the probability a vulnerability (or vulnerabilities, considering a threat vector) is successfully exploited IF attacked (note:  this takes into account all deployed safeguards).  NOTE:  Relating this to the paper, this takes into account LHC magnetic field containment, cooling mechanisms, shutoff mechanisms, etc. which tend to reduce the probability of an "event".

The first factor is termed “Likelihood”, or simply “L”.  The second factor is termed “Impact” or simply “I”.  And the third factor is termed “Vulnerability Severity”, or simply “VS”.

Risk = L x I x VS

Some people (incorrectly) equate "Risk" to "Likelihood", but any correct calculation of Risk must take into account the potential consequences.

As I mentioned, Risk is not just a number, but must include its uncertainty.  Simple calculus gives the uncertainty in Risk, based on the uncertainties of the other factors.

delta (Risk) = I X VS X delta (L) + L X VS X delta (I) + L X I X delta (VS)

So the above is all well known, and any model of how particles are produced and interact in the LHC (or any other model) has uncertainties which MUST be included in the calculation.  In the case of the LHC, some of the uncertainties arise because of the incompleteness of the Standard Model (the LHC is trying to fill in some of those holes), the complete absence of a quantum gravity model (which would help model the behavior of small black holes), and the uncertainty inherent in quantum behavior (the uncertainty principle).

Now - my comments about the paper itself.  First, the paper talks about very small probabilities (e.g., one in a billion).  One has to be very careful about what is meant.  If there is a one in a billion chance of a meteorite wiping out a continent, does that mean that for every billion meteoroids hitting the earth, one would wipe out a continent?  Or does it mean that a continent is likely to be destroyed every billion days?  Or something else?

When we're dealing with very small probabilities, statistical theory does not hold very well.  On page 9 of the paper, the authors state "the tendency to reach the desired answer rather than the correct one".  I would argue that there is no "correct" one for small probabilities, since billions of trials would be necessary to verify the result.  Probabilities are not about "correctness", but about the most probable outcome given large numbers of trials.

More criticism of the paper - Section 4 shows a lack of knowledge of the Standard Model of particle physics, using key references from the 1980's.  Ordinary matter is not made up of "...two types of quarks and the strange quark".  If we ignore anti-particles, ordinary matter is made up of 15 types of quarks and 3 types of strange quarks (among aother particles).  And I'm not sure why the authors are concerned that the earth may be converted to "strange particles".  Strange particles also constitute ordinary matter, but they have extremely short lives by themselves.

One risk I will comment on - that the LHC may produce small black holes that "swallow the earth".  Although small black holes MAY be created, black holes are not the "universal suckers" that some people believe.  First, there is every reason to believe that small black holes would be very unstable and short-lived.  OK - this is based on an uncertain model and COULD be wrong.  But black holes ONLY "suck in" irretrievably (more or less) matter which is closer than its "event horizon".  The "event horizon" of mini black holes is VERY small, so the probability of interacting with any matter would be VERY small.  And at the quantum level, other forces are MUCH stronger than the gravitational force.  Again, this does not negate the possibility that a catastrophic event COULD happen, but (based on known models), it's VERY unlikely.

As an example, when you put a teakettle on the stove, there is a finite probability that half the water in the kettle would turn to ice, while the other half would freeze.  That probability can be calculated accurately (based on a known number of water molecules).

The LHC example is reminiscent of the first A-bomb testing.  Some people thought that there was a finite probability that the earth would blow up in a large chain reaction.  And that was before the Standard Model was as well developed as it is today.  Perhaps it was foolhardy, but known physics indicated it was a very small probability.

So, in the main part, the paper is largely correct.  Risks are based on models, and our Standard Model still has holes.  And quantum gravity is a big hole.  That said, I think the risks are still VERY small (but I could be wrong!).

Cheers,

Prof. James Gort