Nothing in online gaming quite matches the incandescent drama of a dying alliance. It is only in EVE that player organizations routinely exceed 1000 members. If you think World of Warcraft guilds are a hotbed of infighting and ridiculous squabbles over loot, imagine the consequences of a 2500-man alliance tearing itself to pieces. When you take into account the ancillary impacts of alliance death, the carnage gets all the more satisfying: the political destabilization and power vacuums, the birth of new and bitter grudges between formerly friendly corporations, and the absolute and unrecoverable material losses. If you have a thing for chaos and destruction - or even a mild sadistic streak - nothing beats watching the death throes of an alliance. Actually, that's not entirely true. Participating is more fun than watching, and being the architect of an alliance's death is a sublime experience that you simply cannot get in any other game.
One of the many fascinating things about human behavior is that while individual people show great variation and unpredictability, groups of people show striking similarities in the way in which they behave. The more people in an organization, the less variation, and the more predictable they become. Once a certain threshold of membership is passed, alliances in EVE begin exhibiting a somewhat standardized death-cycle. This is unsurprising, because alliances are demographically similar (barring certain ethnic exceptions, they are made of geeks from the 'West'), they exist in the same bitterly contested arena of conquerable 0.0, and all are forced to jump through the various gameplay hoops which CCP dubs 'sov mechanics'.
In the past four years, I've witnessed more than twenty alliances disintegrate, and I've had the benefit of directly orchestrating a good number of those. Better than just hammering at them from the outside, I've been able to witness the process from the inside concurrently by way of our spy network. While watching Lotka Volterra explode in much the same way Veritas Immortalis had, I dubbed this death-cycle the "failure cascade" due to the way in which the process rapidly accelerates from small errors into thermonuclear drama. Delightfully enough, even our most vehement enemies have chosen to use my nomenclature, as well as the more traditional media. Apparently the term has some preexisting engineering use, but googling it now turns up mostly yammering about spaceships.
Despite the popularity of the term, a failure cascade never been properly defined. Like the obscenity standard before the Supreme Court, you 'know it when you see it'. While it is impossible to bring scientific clarity to bear on how exactly one goes about ruining the day of thousands of people in a spaceship game, based on my purely anecdotal experience and the unique access our spy network has afforded me, I am going to do my best to define a failure cascade in formal terms, explain how and why I think this kind of implosion happens, and show how you can identify an alliance in the midst of a cascade, and how to possibly structure your own alliance to defend against it.
A failure cascade is the disintegration of an alliance caused by collective helplessness in the face of sustained and unrationalizable adversity through a process of pilot attribution shifting from the alliance to the corporation or the individual.
Failure cascades follow a predictable five-stage causal chain.
Sustained Adversity -> Failure of Rationalization -> Collective Helplessness -> Change in Identification -> Collapse and Recovery
When we say that an alliance is in the 'early stages' of a cascade, this often means that they are reacting poorly to sustained adversity. "Late Stage" cascade frequently refers to the helplessness phase, because changes in identification are rapidly followed by collapse.
Phase One: Sustained Adversity
This is both the most obvious and the least well understood phase of the cascade. Adversity can take many forms, all of which amount to "bad things happening". An alliance whose towers are being sieged by an enemy force is experiencing adversity. So is an alliance whose ratters and miners are being ganked, or whose jump bridges are being camped or disabled, or who loses a capital fleet in a dramatic fashion. When thinking about adversity, commanders often assume that massive, crushing loss is the most effective way to send an alliance into a cascade. Taking out a capital fleet or a titan is the most commonly-cited method of sending an alliance down the tubes. It is also completely, utterly wrong.
Psychology has shown that humans have an incredible capacity to cope with great tragedy and personal adversity. When asked if they could imagine life as a paraplegic or after suffering some other sort of horrific event, most respondents confidently assert that it would devastate them and leave them permanently adrift. Yet in practice, victims of awful fate swiftly recover their 'set point' of personal happiness. If the ability to mentally cope with great loss did not exist, the species would have certainly died out by now. In alliance terms, collective 'great tragedies' are terribly ineffective at sending an alliance into a cascade. Innumerable Titans have been destroyed, capital fleets wiped out with shocking ease, and these events do not reliably correlate with cascade. It is true that a major loss can have an impact, but only as part of a larger context of sustained adversity.
Why is this? Loss in isolation is something that people are wired to rationalize away. A titan loss can be written off to pilot error, lag, or the fault of the pilot lighting the titan's cyno field. The loss of a capital fleet becomes a one-time affair, the error of a bad fc, or made irrelevant by proclaiming it 'already replaced'. Despite this, people continue to assert that if only a titan or a capital massacre can be achieved, the enemy will surely cascade. This is because of the inherent flaws in human prospection; we're terrible at imagining how we and others will emotionally react to events in the future, so we imagine how we would react to a titan loss, come up with 'Oh god, that would be horrible, I'd be completely demoralized,' and assume that our enemies would react the same way. Wrong!
Some examples: Band of Brothers lost several capital fleets in laughably one-sided battles as well as multiple titans through its history, yet these 'major losses' did not dent them. Nor did having all of their sovereignty removed by a spy in their ranks. In cases where BoB gave up and failed, such as the fight for DG- in Detorid and the Querious campaign, it was after a long period of sustained attrition and many small failures, rather than major losses. Goonswarm has lost more capital fleets in embarrassing ways than I can count, including losing our first titan. In the first invasion of Delve in early 2008, Red Alliance lost a titan but this did not impact their participation in the offensive at all. In fact, the whole coalition suffered what remains one of the largest losses in EVE history (before the disbanding and purging of BoB) during the attack on a BoB CSAA in F-T. Not only did the whole coalition lose 100+ capitals in order to destroy the CSAA, the CSAA was completely empty! Yet no failure cascades ensued.
Rather than relying on shocking incidents, adversity must be sustained and mundane to the point of being banal. Regular fleet losses, regular ganks, the boring stuff which no one bothers to report which still sucks. Unglamorous, everyday loss. A war which takes place in intense bursts does put an alliance at risk of cascade. In Feythabolis and Esoteria, RED.Overlord, Against All Authorities and Stain Empire launched attacks on Goonswarm stations at an extremely slow pace. A station would be attacked and captured, but then there would be weeks of time before a new attack occurred. Because the adversity was not sustained in time, even though Goonswarm was undeniably losing stations, the process had little impact. Had those losses been part of a nonstop campaign of conquest, things would have been very different.
Adversity must also be inescapable. If an assault only impacts one system at a time, and there is very little hostile activity elsewhere within the victim alliance's territory, there is no great risk of cascade. Pilots can choose to put themselves at danger or avoid it at will, so they feel little pressure to change their views of their alliance. This is one advantage of having a geographically large empire; it is difficult for an enemy to put inescapable pressure on your pilots until you have lost more space. Smaller alliances with little territory are often rapidly crushed because it is simple for a hostile force to ensure that every pilot in the victim alliance is under pressure at all times. Empire does not count as an escape; not only do attacking alliances often declare war on their intended victims, a pilot hiding in Empire is not a pilot fighting for his alliance's territory in nullsec.
Phase Two: Failure of Rationalization
Rationalization is a critical psychological defense. It allows pilots and alliances to face adversity and keep fighting - and it is amazing how powerful rationalization can be in the face of evidence (cf. creationism). In internet space there are a bewildering variety of rationalizations for the pilot facing failure. By far the most common is to blame lag, followed by blaming CCP and accusing the enemy of exploits or hacking. These are so common in PvP games that they become universally understood jokes, and despite their humorous connotations they are still earnestly used in a time of need. One moment a pilot will be laughing about how his enemy 'didn't want that ship anyway' and how it has 'already been replaced'; after he explodes, suddenly he feels the need to brag in local about how many other of those ships he has fitted and ready to go, and so that loss didn't matter. Or it didn't count because his enemy 'blobbed', or behaved without following the precepts of e-honor, etc. These rationalizations tend to be hilarious when we aren't the ones making them, but our belief in them is genuine when we're the ones spouting them.
It is in this stage of the cascade that the propaganda war takes a deeper significance to all parties. It is usually during this phase, in the face of mounting failures, that there is the most forum blustering from both sides. At the beginning of a conflict, the aggressor is often restrained in their bragging in case the attacks do not go as planned; the defender has not yet experienced sustained adversity, so it is here that we see the most "good fight" rhetoric with each side congratulating the other. As soon as things turn bad for the victim, however, rationalizations are mustered with alarming vehemence. Accusations fly on the part of the victims as they try to explain their failures away. Similarly, the aggressor does his level best to force the victim to confront the cognitive dissonance between the facts and their defensive rationalizations. This is why there is almost never a 'clean' war, without accusations of impropriety - those accusations are part of a critical psychological defense mechanism that is ingrained in all of us. This is also why using espionage to leak negative commentary from within a victim alliance is such effective a propaganda method, as the dissident voice of a fellow alliance member is much more difficult to explain away.
Yet regardless of propaganda, if sustained and inescapable adversity is applied to an alliance it becomes difficult for the victims to rationalize their losses as failures mount. Say you trip and fall down the stairs once and break your leg. Perhaps you were merely unlucky, had a bad day, slipped on a banana peel, were startled by a loud noise, or were wearing uncomfortable shoes. The broken leg hurts like hell and is certainly bad, but it doesn't mean that you are a clumsy person. However, if you stubbed your toe on something every single day, you would find it hard to deny that you were a klutz - even though your stubbed toe is far less painful than the broken leg. So it goes with alliances: dramatic, painful losses are easily written off, but repeated more mundane failures are difficult to rationalize. If your alliance keeps losing fleet fights and your small gangs are getting pasted, and you lost yet another R64 moon, it becomes difficult to maintain that your alliance is doing well. When rationalization fails, helplessness sets in.
Phase Three: Collective Helplessness
Helplessness is a state where a pilot comes to believe that his actions on behalf of the alliance are pointless, impotent, or irrelevant in the face of adversity. This can be because he cannot ignore the failings of his alliance and must acknowledge them: "This alliance sucks, what's the point." It can also be because he feels that no action he can take will make an impact on the situation: "I love my alliance, but I can't do anything to keep us from getting rolled." Regardless of the reasoning, helplessness takes the pilot out of the war until the helpless state is overcome, which will depending on the pilot's explanatory style. For some, helplessness may be swiftly overcome. Some pilots never give up in the first place, others keep fighting with only short breaks until the bitter end. Others throw in the towel after the first loss.
Individual variation is minimized when dealing with a group of more than a thousand pilots, however. It doesn't matter much to the alliance if any given random pilot has given up and needs some time out to recuperate, but if the level of trauma reaches the point that the alliance cannot muster an effective military defense, a state of collective helplessness results. There is a certain military threshold of effective resistance, and once the victim alliance loses the ability to resist, adversity accelerates rapidly. Failure after undeniable failure knocks increasing numbers of pilots into a state of helplessness, further degrading the alliance's ability to defend itself.
It is difficult to come up with an example of an alliance which has recovered from this late stage of the cascade. But there are some examples of alliances avoiding this state altogether, even after the failure of rationalization. One of the significant mysteries of modern 0.0 warfare is Red Alliance's incredible 2006 defense of C-J6 against the combined assault of Lotka Volterra, Ascendant Frontier, and the rest of the Southern Coalition. RA was outnumbered six to one, more so if you consider that the average RA pilot in that battle was multiboxing. They had no allies to help them (as this was months before Goonswarm and Tau Ceti Federation joined with them) and were considered pariahs in the 'EVE Community' on account of RA's adoption of 'dishonorable' tactics in the face of these overwhelming odds. Red Alliance had lost a massive amount of territory and had one single station left. By most theories of alliance disintegration, Red Alliance should have been rubble. But somehow they resisted and turned the tide. How?
The conventional wisd m says that this is due to the ethnic nature of Red Alliance; they did not cascade because they possessed a unique, out of game culture. Yet there were multiple ethnic-Russian alliances in EVE at the time (RISK and Against All Authorities) and other ethnic Russian alliances have since hit failure cascade. Why did Red Alliance survive? Because they were not helpless, even in the face of sustained, undeniable adversity. The average skill level of the RA pilots was far beyond that of their enemies. While RA did lose territory, in the battle for C-J6 and elsewhere the Red Alliance battleship sniping groups devastated their opponents, even if they lost battle after battle. Red Alliance pilots repeatedly destroyed Southern Coalition capitals in an era where capitals were far more valuable than they are today. Adding further confirmation that they were not helpless, RA pilots could see evidence of the impact their efforts made when Lotka Volterra and Southern Coalition pilots would whine on the forums and accuse RA of 'not dying fast enough'.
Phase Four: Change in Pilot Identification
A change in pilot identification is the primary method people use to escape helplessness in a failure cascade. Identification is how a pilot views and describes himself in the context of the game. When a pilot begins the game, he is alone in Empire and friendless, so his identification is only with himself, adrift as an individual and concerned only with his own interests. In time, he will find and join a corporation, and become integrated into that corporation's social milieu. If asked "Who are you?", many pilots at this stage would say "I am Pilotname, a member of Blah corporation," much like how people who live on the East Coast of the US identify themselves with their professions (I hate that). Should the corporation join an alliance with a strong identity, the pilot may come to label himself as a member of the alliance, particularly if there are frequent alliance events (fleets, battles, mining ops) where members from different corporations can interact and bond.
There is a natural tension between the interests of a corporation's CEO and the alliance leadership. An alliance's resistance to adversity is directly linked to its ability to retain pilot identification, so it is in the interests of the alliance to convince its membership that they are members of an alliance first and foremost. But the power of CEOs comes from their leadership of a corporation and the obedience of the membership to them, not to the alliance. Even in an alliance with extremely strong identification, member corporations will often have corp-only activities to retain a unique identity. Most alliances are run by a lose oligarchy of the CEOs, but some corporations - usually those with the strongest military - have more influence than others. These leader corporations tend to be the ones pushing weaker corps towards ever-increasing identification with the alliance, while maintaining a separate corporate identity. This tension becomes exacerbated in a time of war as the stronger corporations call upon the weaker to shoulder the burdens of the conflict. In a cascade, the push-pull of corporate vs alliance identification explodes as CEOs and pilots begin to stop thinking of themselves as members of an alliance and begin to look out for their own corporate or individual interests.
A shift in identification happens because it is one of the only easy escapes from a state of helplessness, besides quitting the game entirely. A pilot who identifies himself with a helpless alliance in the midst of a cascade experiences helplessness himself, and to get out of it all he has to do is change the way he thinks about himself. Rather than being a member of a failing alliance, he thinks of himself as a member of a perfectly effective corporation in an alliance full of failures. In one moment of rationalization, he absolves himself of the helplessness and reassures himself of his superiority over everyone else not in his corporation.
As more pilots are knocked into a state of helplessness by sustained adversity, more shift their identification away from the alliance. Pressure on the CEO grows from this disaffected membership to get out of the ugly situation and abandon the alliance; that pressure is exacerbated by the tension between CEOs at the leadership level. Collapse becomes inevitable as personal animosity grows and corporations begin quietly evacuating their assets unbeknownst to the rest of the alliance. This is the phase where open infighting within the alliance becomes common, as corporations blame each other (or their allies, another popular scapegoat) for the alliance's failures. For example, when announcing their loss of the Great War, Band of Brothers claimed that they had been a 'sacrificial lamb' for their allies.
Phase Five: Collapse and Recovery
There is a significant stigma against being a corporation who flees an alliance 'too soon', so the collapse of an alliance at the terminus of a failure cascade resembles an avalanche. No corporation wants to be the first to go, so when the first corporation finally screws up the courage (or, more usually, finds a likely-seeming excuse for how they've been wronged by the alliance leadership) to leave, they are rapidly followed by those who were holding back. The collapse has an incredible inertia, as it provides its own motive to leave for those corporations who did not originally intend to. Remaining in an alliance which has cascaded is extremely dangerous as it has no more military power, and since the majority of the member corps already left there is no shame in exiting. Often there will be a small core of corporations who stick with the collapsed alliance, usually those corps who wielded the most power in the alliance.
The post-collapse alliance has two options remaining besides disbanding. It can try to make a life for itself in non-conquerable 0.0 space, claiming to be 'free of the shackles of POS warfare.' History is full of cascaded alliances moving to NPC 0.0 vowing to return, only to dissolve within a matter of months. The raison d'etre of the nullsec alliance is to hold space, and after being stripped of this common interest the core corporations of the failed alliance often find they have nothing to hold them together any longer. In addition, there is a significant public stigma associated with cascaded alliances, and usually the process of the cascade has seeded significant animosity between the remaining corporations.
The other option is to become an alliance of temporary refugees, with the post-cascade core of the alliance moving into the territory of a stronger, stable ally. Depending on how patient the caretaker alliance is, in these circumstances it is actually possible to recover from a cascade. This is not only because the refugee alliance is no longer suffering from inescapable adversity but also because having allies who will aid a defeated alliance is enormously reassuring (and rare).
In either situation, the impact on the former corporations and pilots in the aftermath of a cascade is minimal. Most alliances in EVE end up collapsing due to hostile invasion, and if the experience was that traumatic, the losers would stop playing the game en masse. In truth, the experience -is- traumatic, judging by the posts made by the helpless pilots in the moments of their suffering. But as soon as they have escaped from their helpless state and their failed alliance, the soothing balm of rationalization comes to the rescue, and memories are helpfully adjusted. Post-cascade pilots remember that they put up a great fight and never lost hope, while it was those other guys in that other corporation who ruined everything. Regardless of the circumstances, a cascade is always the other guy's fault.
Defenses Against Cascade:
The best defense against a cascade that I have discovered is having a monolithic alliance structure, where the 'alliance' amounts to one corporation. It is much harder to put a corporation into a cascade than it is an alliance, because CCP has structured corporations as dictatorships with one absolute authority. Alliances have competing CEOs and a variety of different interests; the monolithic corporation may still have some amount of infighting and scheming, but much less than the more common oligarchy. The only example of a monolithic alliance that I know of is Goonswarm, which is the Goonfleet corporation, plus several much smaller member corporations. Rather than having an oligarchy and shared power between a number of member corporation CEOs, the monolithic alliance has only one leader, and the consent or membership the ancillary corporations (while helpful) is not strategically necessary.
It is also helpful to have an ethnic, national or cultural identity outside of EVE to bind the pilots to identifying with the alliance. There are a number of ethnic organizations in EVE, and these tend to be more resistant to cascade than non-ethnic organizations. The smaller the ethnic group or more isolated the language, the better a defense against cascade there is. Right now, Russian and German players have a number of alliances which they could join, so the need to stick to a particular alliance is minimal. There is only one Polish alliance, by contrast, and perhaps two Hungarian alliances. Those pilots are very invested in their alliance's health.
In practice, this means that a monolithic or ethnic/national/cultural alliance can be repeatedly brought to a state of helplessness, but it is very hard to shift pilot identification and induce collapse.
What the Hell is All This Psychology Crap:
Does this all sound like unscientific bullshit? It is! This is a purely anecdotal theory that I've pulled out of my ass based purely on extensive observation, and you might wish to discredit it merely because I'm an obviously biased Goonswarm partisan. That said, everything else I included about learned helplessness, rationalization, failures in cognition and the inability of humans to predict their future states accurately has been tested rigorously and is taken from behavioral economics or psychology. Here's some links about this kind of research to help you waste time at work and/or avoid common errors, such as a persistent belief in classical economics and/or libertarianism.
While this has nothing explicitly to do with the theory of the failure cascade, if you're interested in psychology or espionage or just really enjoy manipulating people, you should probably read Robert Cialdini's 'Influence'.
This column may not be easy to read, but it distills much of what I've learned about failure cascades into one place; the identificaiton of phases is particularly useful. It is gratifying that the game has adopted my nomenclature to describe the collapse of an alliance; 'failure cascade', 'cascading', 'failscading', etc are by 2012 the unquestioned terms of art.
Even three years later, there's not much to add; the theory, definition and phases have stood the test of time, with only ever-more examples to support it.