Are the New Levels of Australia’s National Terrorism Threat Advisory System Seriously Flawed?

Level of Discussion

The Council of Australian Governments (COAG), held its 40th meeting in Sydney, 23 July 2015. Amongst the topics of discussion, was the proposed changes to Australia’s National Terrorism Public Alert System in an effort to ensure that the threat levels “provide more precision, and provide better information to the public on changes to the national terrorism threat level”.

Under the existing National Terrorism Public Alert System as used to “communicate an assessed risk of terrorist threat to Australia” there are four alert levels:

  • Low: Terrorist attack is not expected.
  • Medium: Terrorist attack could occur.
  • High: Terrorist attack is likely.
  • Extreme: Terrorist attack is imminent or has occurred.

As stated within the levels, the current system is conveying the expectation of a terrorist attack.

In addition to being a source of public awareness, the system also aims to guide preparation and planning, inclusive of precaution and vigilance undertaken to minimise the occurrence of terrorist incidents.

The newer system, expected to be introduced later this year after public consultation, suggests moving to five alert levels:

  • Not Expected
  • Possible
  • Probable
  • Expected
  • Certain

As reported in the AFR, Prime Minister Tony Abbott said the new system had the endorsement of Federal, State, and Territory leaders; and he had the following to say:

“Any changes to the threat level under the new system will be accompanied by a statement providing more information on what the new threat level means, where the threat is coming from, what the potential targets are and the means by which an attack might be perpetrated.”

The Need for a Sufficient Statistic

While the increased number of threat levels (from four to five) may be seen as an improvement, as may the increased disclosure of the potential targets and means of an attack, that is not to suggest the proposed system is superior. It still carries with it the similar flaws of the existing system, but manages to shuffle around some wording and simultaneously ignore the related problems.

Or mistakenly believe they have been solved.

Without knowing the specifics, it can make full consideration of the matter difficult. However, even before it is released for public consultation, numerous elements may be addressed.

A major challenge with attempting a risk management system in such a context, is that there are aspects of both art and science to it. With that comes the falling back onto worn practices, which although some have their uses, others are little more than questionable lore and mythology.

Another critical element, is that of the information set. That is, what is known by someone at the time. With due consideration to the fact that what one person knows may not be the same as what another person knows.

Consider first, the simple example of a fuel gauge in a car. The purpose of the fuel gauge, is naturally to tell someone how much fuel is in their car, in the knowledge that when the fuel gauge shows as empty, the car will (typically) cease to run.

Of course there are many factors which will go into this, such as: type and size of the engine; the amount of work the engine is doing; how much load is being carried in the vehicle; speed of the vehicle; weather conditions; road conditions; traffic conditions; style of driving; shape of the fuel tank; the accuracy of the fuel gauge itself and how well it is working; amongst others.

Yet the question is:

Is the fuel gauge sufficient to make an informed decision?

Yes, is the likely answer for the most part. Provided that the fuel gauge is working with reasonable accuracy. This answer, however, will be subject to the driver’s consideration of other factors though, regarding how much they want to risk running out of fuel, and what may be the implications if they do.

The driver may add additional layers of risk management and redundancies as a precaution, such as tracking how many kilometres the car has travelled since it was last refuelled and carrying a jerrycan of extra fuel.

Now turn attention to the National Terrorism Public Alert System. Again, the important question is:

Is the alert system sufficient to make an informed decision?

Akin to the fuel gauge example, the alert system is attempting to communicate and compress many factors into the one gauge. Factors are likely to include: threat credibility, detectability, expectation, nature, and number; impact of threat; impact and severity should the threat become reality; the ability of the threat and its consequent actions to adapt; amongst others.

This is admittedly quite a difficult problem. One with a solution that may not be quite as convenient as a fuel gauge. Suggesting a threat matrix (or a distribution), such as threat impact versus threat frequency, may not be as readily understandable by the general public as the more simple levels of alert.

However, this also makes the risk management system prone to being little more than a system of colour coding. Particularly if it is lacking in qualification and quantification, with some sense of probability and scale attached to it.

Levels, Order, Probability, and Confusion

A major concern is the wording of the proposed new system’s levels, which fails to be an improvement on the old and existing system.

The key here, is the ordinality, or the order of the threat levels.

Under the existing system, this is quite clear:

Low is less than Medium, is less than High, is less than Extreme.

There is no overlapping of the levels (save for the explanations accompanying them). Although notably there is a lack qualification and quantification of such levels.

Now, further consider the proposed new system’s levels:

  • Not Expected
  • Possible
  • Probable
  • Expected
  • Certain

Foremost is that the existing system is attempting to convey the expectation of a terrorist attack; the new system is apparently trying to do likewise, but it is not suggesting that it is truly any more precise however. It is merely shifting some words, and omitting others. Also note that there is no impact guidance here.

Is a “Certain” terrorist threat to 10 people the same as another equally certain terrorist threat to 10,000 people?

Weathering the Storm of Probabilities

The language used on the new levels is also highly problematic in several ways. While there are differences and allowances made with respect to everyday grammar and layperson’s terms versus technical terminology, there seems little to clarify the meaning of the new levels between either spectrum.

Basic probability is bounded between 0 and 1, where a probability of 0 implies that the outcome is impossible and a probability of 1 implies that the outcome is certain. Whilst there are differences between academic and textbook discussion compared to the real world as probability may experienced and applied, within reason, notions of probability tend to be basically grasped.

For example, most understand the statement “there’s a 20 percent chance of rain tomorrow” to imply a lower probability of rain than “there’s a 90 percent chance of rain tomorrow” with probabilities of 0.2 and 0.9 respectively.

Yet is this concept so easily applied to threats?

Is talking of the threat of rain the same as talking of the chance of rain?

That the weather is not sentient, makes the question rather tautological. Although it’s also in the interpretation of what a rain event means to different people. Some may want rain and others not. In the former case, rain is not a threat but the chance of something favourable happening. Weather reports may also attempt to qualify the type of rain, such as it being showers or drizzle, in addition to giving a quantified probabilistic statement.

Importantly though, the weather is not going to fight back any more than it is going to plan ahead. Nor is it going to alter itself because it has learnt of the weather forecast regarding it.

However, with regards to terrorism, it is not quite as simple, because the threat to engage in an act of terrorism may be rather different to actually engaging in that act. Furthermore, the notion of sentience and adaption, combined with “Expected” shall be further addressed shortly.

Imagine the Overlapping Possibilities

Taking the proposed wording of the new levels, first consider “Possible”. In everyday use, “possible” is typically used to convey the remote, but highly unlikely sense that something could happen. However, to tighten up the language, “possible” is a non-statement. Provided that something is not impossible (with a probability of 0), then by definition, it must be possible because of the simple fact that it could happen. The reality of the matter, is that a terrorism threat will always be possible simply because it could happen.

Next is the “Probable” level, and again in everyday use “probable” implies it could happen, and that vaguely it is more likely to happen than is the “possible”. However, this is yet again another non-statement. All probabilities must by definition be “probable”. And again, without an attached probability or chance statement (however misguided and inaccurate), a terrorism threat will always be probable simply because it could happen.

“Expected” falls into exactly the same definitional trap. In no way does it make a statement of probability, such as there’s a 90 percent chance of an expected terrorist threat. It is equally valid to make the statement that there’s a 0 percent chance of an expected terrorist threat. Granted, everyday language implies “expected” can happen, it might even have greater chance of happening than not. However, it is still considerably vague.

The three levels of Possible, Probable, and Expected provide no further guidance, and are more convoluted, than the existing Low, Medium, and High levels, whether in common everyday or technical language. Essentially these new levels are falling into the colour coding trap of apparent risk management, which actually does little other than give the appearance of something being done.

For example, were the new proposed level names Blue, Green, and Red, they’d provide little by way of real information or represent an improvement.

Certain of What, Exactly?

Then, there is the “Certain” level. If something is “certain” there should be no doubt about its existence, or that it is a foregone conclusion. It is guaranteed. This is largely reflected is both everyday language and technical language, although it is not uncommon to refer to things with a level or degree of certainty. If a probability were to be attached to something that is certain, then a probability of 1 would apply because it’s supposedly impossible for it not to happen.

However, it must be noted that simply because a belief is certain, is not to suggest that the belief is correct. Or a fact. Someone may believe with absolute certainty that the world is flat, but it doesn’t mean they are correct.

Regarding a terrorist threat though, it’s still rather vague. It would not be incorrect to suggest that it’s certain a terrorist threat will occur in the future. Technically such a statement is bound to be right eventually, because it has put no limits on the time for the terrorist threat to occur; again making it a non-statement as far as conveying actual information is concerned.

Additionally, would not the certain terrorist threat also be realised because the threat had actually been received?

Or if a specific threat is certain, should it be stopped beforehand; perhaps in some ways, paradoxically making it not certain or uncertain?

Furthermore consider the following two hypothetical terrorist threats:

  • An off-the-cuff remark made by someone that they’d like to blow up a building.
  • A person armed with explosives standing in a building and threatening to blow it up.

Assuming both threats were received, then the threat level is “Certain” but does that level by itself really tell someone anything about the difference and severity between the two hypothetical situations?

An Unexpected Blank Swan

Which leads into the area of “Not Expected” and as already mentioned, the troublesome issue of attempting to qualify and quantify the extent of the “Certain” threat as it may impact human lives.

If a terrorist threat is “Not Expected” then that’s a probability of 0, but such an assessment could objectively be extremely wrong. If intelligence, counter-terrorism, and law enforcement agencies are correctly doing their jobs, a terrorist threat (and attack) should always be expected. (Further highlighting a flaw of the “Expected” level.) They should be running as many scenarios as they are able to, in an attempt to define what may happen, however improbable, and developing strategies and principles to deal with such expected outcomes and situations accordingly.

That does not mean that they can identify and plan for every situation – they simply can not – but it does mean that it should always be expected that some type of threat could occur.

About the only time that the phrase “not expected” should enter the vocabulary regarding potential acts of terrorism, should be when it is used honestly with the benefit of hindsight acknowledging that something was a “Black Swan”.

The concept of the Black Swan, as popularised by Nassim Nicholas Taleb, relates to the impact of the highly improbable. The event, which (subject to the information set) no-one saw coming, and save for the benefit of hindsight, it was completely unpredictable before it happened.

With this, in the aftermath, often comes the case of having the general still fighting the last war. (An example from World War II, would be the Marginot Line.) Efforts will be put into place to supposedly manage the risk which occurred, to prevent the same thing from happening again. Which although reasonable to not repeat past mistakes, the impact of the event was so great precisely because it was unexpected. Were it expected, it’s unlikely it would have had the same impact. In fact, if it was genuinely expected in advance it may have been prevented entirely.

Good luck may also be misinterpreted as something which was planned for, when in fact the outcome was unexpected and unforeseen before it occurred.

This is reflected in the rather chilling statement issued by the Irish Republican Army (IRA), following its failed 1984 attempt to assassinate British Prime Minister Margaret Thatcher:

“We only have to be lucky once, you will have to be lucky always.”

If Order Matters

Considering the above thus far, and the vagueness of the proposed new levels’ names, and the levels potentially overlapping, answer the following:

Risk Mismanagement

Furthermore, there are other issues related to the “Certain” and “Not Expected” levels, in particular with how they are, or aren’t, able to convey the information to the public. Where, because this information becomes public, the terrorist threat may adapt accordingly.

Whilst as outlined in the Australia’s Counter-Terrorism Strategy report, elements such as disruption are in line with concepts of fourth-generation warfare (4GW), the proposed new threat levels still exhibit flaws.

Consider the hypothetical of a “Certain” threat to shoot people on Street A. Subject to how this information is made public, it may sensibly lead to people avoiding Street A. Yet, knowing and adapting to this, the terrorist action instead shifts to the unexpected outcome of bombing Street B.

Although the heightened threat level would be unlikely to shift from “Certain” to “Not Expected” under the circumstances, the “Certain” level has done little to prepare people for something unexpected (and also “Not Expected”) which adapted to the information set.

These unintended consequences of risk management need not be spoken of in purely hypothetical terms either, as the Germanwings disaster sadly demonstrated. The act of attempting to prevent planes from being hijacked following the events of September 11, by placing a door only able to be unlocked from inside the cockpit, created a new risk which was unforeseen by most.

As previously suggested, a major challenge to any alert levels system, is its ability to present itself as sufficient for the intended purpose.

There are considerable issues with the notion of the alert levels expressing a terrorist threat. Already discussed, is that not all threats are of equal magnitude, even if they are classified at the “Certain” level.

Back to Nature

Where the proposed National Terrorism Public Alert System could learn something from, would be the Australian bushfire danger ratings and warning systems. An example and explanation of such a system can be viewed here.

Consider that much like the fuel gauge, the bushfire danger ratings are compressing many factors into something useful and understandable by the general public. Importantly, it is dealing with “danger” instead of “threats”. The underlying factors, such as temperature, lack of rain, season, winds, lightning strikes, fuel sources and combustibility, may all be considered threatening so to speak. However, rather than talk of the threat as the end itself, it instead factors them into the concept of danger for the levels.

Danger, it seems, as a concept is better able to encapsulate and express in ways readily understandable and communicable to the general public, the notions of threat combined with impact.

It should not be lost either, that the bushfire danger levels are:

  • Low-Moderate
  • High
  • Very High
  • Severe
  • Extreme
  • Catastrophic

No messy mentions of amorphous and technical concepts like “Not Expected” and “Possible” and the rest.

That does not make the bushfire danger system perfect; and terrorism with human beings who are aware and adapting to the circumstance does make things more complicated. However, even with bushfires, notions of human interaction and the deliberate lighting of bushfires by pyromaniacs and copycats, the relevant focus should be on awareness and prevention. Understanding the danger they pose. Rather than referring them as threat levels as the end in and of themselves.

Conveying the information to the public regarding terrorism danger could also further take cues from bushfire and weather alert systems. As at the time of writing, Australia is at the High level of alert. The risk with this, is the potential complacency it may engender with such a blanket classification.

Using things such as topographical heat maps (see here for an example), inclusive of clustering analysis, may be a way to visually demonstrate terrorism alert levels across the country which vary with location.

This would be subject to some qualification. Such as more densely populated areas, like major cities, being more likely potential targets than sparsely populated bushland. However, the concept could be explored in the official consultation process with the public.

Consultation and Communication

In consulting with the public, it is also important to be able to communicate the terrorism alert levels in the most easily understood and unambiguous manner. While much of the above may appear as issues of semantics, if the very premise of the question is wrong it is unlikely to provide meaningful answers and should be reframed.

Suggestion of the new system providing information about the meaning of the levels is a positive, though it’s doubtful that the focus should be solely on the threat. Identification of, and warning potential targets, including potential methods of how attacks may be perpetrated, could also be considered a positive. Yet the adaptive nature of terrorism should not be discounted.

Due consideration should be given to how the alerts are interpreted, inclusive of notions of low-to-extreme versus not expected-to-certain, and threat versus danger. Consult with the public, ask them questions about such things, including how they interpret them.

Having made it this far, what do you think about proposed changes?

Does the new system really “redefine the threat levels to provide more precision” as claimed?

Is it not expected that some terrorist threats may possibly, probably be expected for certain?

Add your thoughts in the comments and answer the uthinki question below: