Michael KreponFailures of Intelligence

US intelligence agencies have been guilty of four general types of error — or more, depending on counting rules. The most costly type of error is a failure to recognize and pull together “actionable” intelligence in time to foil bad surprises. Type I errors are typically abetted by compartmentalization within the intelligence community. The failure “to connect the dots” prior to 9/11 wasn’t new; it used to be characterized as failing to detect signals in noise, as particularized in Roberta Wohlstetter’s masterful book, Pearl Harbor: Warning and Decision. A subset of this type of error is being unable to detect well-established, covert nuclear weapon programs, like Saddam Hussein’s prior to the first Gulf war.

Type II errors are the reluctance or inability to predict the possibility of pleasant surprises, which become possible when leaders change, when they realize dead-end policies, or when they mess up. It’s hard to predict good news. Leaders open to changing course remain surrounded by those who have survived by not taking risks or by engaging in bad behavior. Besides, strategic culture doesn’t change on a dime, so risk-takers have to send mixed messages, signaling new possibilities while protecting their flanks.

A subset of this type of error is assessing the existence of a well-established covert nuclear weapon program when it may have been halted by decision, incompetence, or dysfunction. Examples include Saddam Hussein’s Iraq prior to the second Gulf War and Muammar Qaddafi’s Libya.

Predicting more of the same is always a safer course than predicting the possibility of diplomatic breakthroughs – especially since the latter can brand the intelligence analyst as being hopelessly naïve. Did the US intelligence community see the course-correction in Burma coming, or was it as wrong as the Kremlinologists who misread Gorbachev? As noted earlier in this space, US Iran watchers and proliferation cynics may have gotten Rouhani wrong, as well – we’ll see.

Type III errors occur with the zealous collection of data that does far more harm to US diplomacy and international standing than might be gained from preventing bad surprises. This error is now a daily occurrence, enabled by new information-retrieval technologies and the lingering effects of suffering massive attacks on the US homeland over a decade ago. One measure of how much 9/11 remains with us is the reluctance of a President steeped in Constitutional Law to pare back the National Security Agency’s appetite for collecting and storing metadata on American citizens.

Type IV errors relate to discounting the possibility of paranoid behavior by bunkered adversaries. The presumption at work here, sometimes tested by those seeking to burnish their hardline credentials, is that bad actors won’t over-act, even when the US jerks their chains. This type of error was most egregiously on display in 1983, when some in the Reagan Administration were oblivious to the possible consequences of pursuing initiatives to keep the Kremlin off-balance. The US intelligence community, which found it difficult to distinguish between paranoid and aggressive Soviet behavior, failed to raise cautionary flags. For those interested in the particulars of this Year of Living Dangerously, check out the National Security Archives’ treasure trove of files about Able Archer.

The Obama Administration, chastened by ill-conceived, costly, and poorly executed wars in Afghanistan and Iraq — the surge in Iraq being successful only on a timescale irrelevant to larger consequences — is unlikely to engage in Type IV errors anytime soon, whether or not the intelligence community speaks up.

The commingling of Type IV and Type II errors makes it hard to explore whether the potential for positive developments exists below layers of paranoia and pathology. Making overtures to reduce nuclear dangers in such cases requires heroic efforts to establish footholds for those not locked into enemy images, to relax the grip of those who are, and to assure allies and friends dubious of risky diplomacy. This would be as hard to do for North Korea in the future as it is now for Iran. Leaders who accept this challenge can expect very little help from intelligence agencies.


  1. Michael Moser (History)

    There might be a type V error: failure to consider unintended consequences. here is a recent example:

    The US threatened to intervene in Syria, and then backed down – I guess it might have been smarter not to raise the threat in the first place.

    So on September 14, 2013 – US and Russia agree on the deal to eliminate Syrian chemical weapons; this signals that the US does not have the intentions/means to back up threats of military force;

    23 November 2013, China sends air force jets to Senkaku islands;

    I am assuming that the two events are related; interesting what comes next.

    • Magpie (History)

      Yeah, no, not related. The Senkuku Push has been on the cards for a long time.

      I’ll try not to write my usual epic, but in summary: China has (I think) enough confidence in their anti-ship ballistic missile program to push out eastwards. Add a growing nationalistic middle-class (nationalistic as middle-classes are everywhere), and the risk-reward balance is tilting a lot more towards territorial expansion. This Chinese emergence is running into a revival of jingoism in both Japan and South Korea, which is making the game a lot more dangerous than it should be, but that won’t deter them. They know they’ve won The War That Won’t Happen (On Account Of China Already Won).

      …if China doesn’t already have naval superiority right across the two China-Seas, they will soon enough to deal with any realistic escalation. With what I strongly believe they have gained from the DF21D program, China is now the world’s largest and least sinkable fleet. (Note I didn’t say China “has” the fleet. I said it “is” the fleet). As long as any war they get involved in is restricted to the west pacific, then they’ve already won. They know that – now it’s just a matter of pushing to get the rewards.

      Simply compare the cost of building and maintaining an Aegis platform in the west pacific, with the cost of overwhelming that platform with road-mobile anti-ship ballistic missiles. Even if they *haven’t* managed MIRVs, it’s not even close. Game over, man. Game over.

      Expect to see increasing Chinese support for Ryukyu independence, along with more activity in the Spratly Islands. They’ll go slow – there’s no profit or need to push the situation hard enough to end up in an actual war, and the Pivot is part of the pushback to try slowing them down a bit – but push they will. It’s just a matter of how far they can go while they have the upper hand – which, for now, really is “the foreseeable future”. Honestly, long term, I reckon Taiwan is in play.

      Damn, that ended up long. Sorry.

  2. Ryan (History)

    Type IV errors can still occur.

    North Korea exists you know.

  3. OT (History)

    Int agencies in general, not only in US, tend to err to the side which ensures continuity of their funding. I’d call that bomber gap type error. Or “North Korea has uranium enrichment capability type error”, if you like. As observed, this error type has almost no upper limit and is actually magnified by actions of fearful target, who does it’s darnest to scare the spook away with smoke and mirrors.

    Type II error is explained by unability to understand that target is not monolithic entity, but consists of multiple factions. There is allways hawks, who keep themselves fed by military spending, and doves, who hate watching all that money flowing by them. But they both are limited in ways they can express their true views. Leader might be forced to be ambiguous.

  4. krepon (History)
  5. krepon (History)

    For a jaw-dropping look into the reliance on tactical nuclear weapons, Cold War-style, see:


  6. Bradley Laing (History)


    Dr. Hutcheson reports that NRL’s SuperMISTI hybrid system has successfully detected, identified, and localized gamma and neutron sources in a maritime environment at operationally relevant distances.

    Since the completion of the Manta demonstration in 2012, NRL scientists have increased the number of HPGe detectors in the detection/identification subsystem from 24 to 48. This modification significantly enhances the gamma detection/identification capabilities of the system.

    The NRL team plans for further performance enhancements, including the use of a large-area BF3 detector array to increase neutron detection capabilities and the implementation of better localization algorithms for neutron sources.

    —What does this mean in non-technical language?

  7. Jeannick (History)

    Analysis is not values neutral ,
    it would be a very brave or foolish head of department who would present a report flatly contradicting the deeply held opinion of his leadership .

    another point is that if there isn’t enough dots ,
    the picture could be anything , ultimate intelligence is what is in the head of the leadership, for foreign actors
    the very mind set is foreign too
    hard intel doesn’t tell you of intent ,
    it only suggest possible outcome

    intelligence is perceived through one’s own colored glasses
    At a deeper level , one can only see what one try to see
    the search for facts ,already imply a mental selection of what one will look for

  8. j_kies (History)

    This appears to conflate issues of actual intelligence failures as well as poor judgment/resource application by the leadership in intelligence matters.

    The Wiki article is a nice starting point http://en.wikipedia.org/wiki/Cognitive_traps_for_intelligence_analysis

    The advantage of US intelligence was its dependence on technical means of collection (MASINT etc) where the analysis/exploitation attempt to reverse-engineer foreign systems. Together with appropriately conservative assessment approaches using engineering / scientific means, net assessments could be provided to the decision layer with confidence levels. The engineering/scientific expertise of the IC is trending downwards in military areas as personnel with direct experience in weapon issues are becoming rather scarce – in reaction the assessments are becoming permissive and the threats ‘grow taller’ to avoid surprise due to lack of US IC expertise.

    Our great failures occurred via conflation of human sources (CURVEBALL etc) with political agendas as a type of political confirmation bias.

    As to current events; broad collections are a waste of resources as pattern recognition inherently fails in the absence of pre-existing patterns.