Tigerhawk discusses the coercion and intelligence dilemma.

There is a tradeoff between intelligence and firepower. The less intelligence a counterinsurgent has, the more coercion he must use. Greater intelligence allows surgical use of force that kills more insurgents and fewer non-combatants. Restoring law and order to a warzone gains greater cooperation, because local populations respond negatively to arbitrary violence.

The US Army Counterinsurgency Manual:

1-119. The presence of the rule of law is a major factor in assuring voluntary acceptance of a government’s authority and therefore its legitimacy. A government’s respect for preexisting and impersonal legal rules can provide the key to gaining it widespread, enduring societal support. Such government respect for rules—ideally ones recorded in a constitution and in laws adopted through a credible, democratic process—is the essence of the rule of law. As such, it is a powerful potential tool for counterinsurgents.

The tradeoff between coercion and intelligence is simple. If you have excellent intel on an enemy leader, you can assassinate him without harming anyone else. If you lack intel, to kill him you must shell an entire neighborhood with artillery. More civilians will die, more property will be destroyed, and the civilians will be less likely to cooperate with those who randomly attack them.

Intelligence increases the productivity and efficiency of coercive acts. This determines whether the coercion will be very specific or very broad. Extremely accurate intelligence is highly effective – but so is extremely broad coercion. Police states and genocidal strategies work however morally repugnant they may be. If you cannot differentiate insurgents from civilians, you can kill everyone and let God sort them out.

Consider this model:
One of the problems in insurgency style wars is that both the insurgents and counterinsurgents sink to the middle of that curve where neither is effective. They sabotage one another’s efforts and prevent cooperation.

The change comes as one side wins cooperation from a growing number of civilians so it gets more intelligence; and the only way the other can stay as effective is by increasing its coercion, which only increases civilian cooperation with the other side. Then it becomes a race to see whether surgical strikes or brutality will win.

Coercion is naturally disadvantageous, since it the goal of both sides is to win control over the population and rule over them after the war ends.

Non-combatants are the key players in an insurgency. The combatants on both sides seek the cooperation of non-combatants. They do not need or want to win over “hearts and minds” – just cooperation. And this provides intelligence.

Non-combatants will cooperate with the side that represents normal law and order and does not use capricious violence.

a noncombatant will cooperate with the side that punishes noncooperation with the greatest specificity. If one side punishes capriciously, most rational noncombatants will decide that they are better off cooperating with the other side. Why? Because the more capricious side — lacking good intelligence about who is and is not cooperating — may punish noncombatants whether or not they cooperate with the other side. The side that punishes accurately, on the other hand, will only punish genuine noncooperation. Therefore, the smart noncombatant cooperates with the side that neither punishes too many actual cooperators or fails to punish too many actual non-cooperators, because he reduces his risk of punishment by the side that punishes efficiently without altering his risk at the hand of the side that punishes capriciously.

This is a reinforcing loop in the COIN system. A state that uses minimal coercion will gain more cooperation and intelligence. Intel allows surgical application of force to punish insurgents and reward cooperation.

The other side becomes trapped in a negative reinforcing loop. They lack cooperation and intelligence, so they must use broaders coercion and random violence to achieve their objectives. The extra violence further alienates civilians, further reducing cooperation and intel, and creating the need for greater coercion.

Since these loops reinforce the current pattern, it is very difficult for the losing side to change strategies. They become dependent on the very coercion that denies them intelligence.

The worst part of the dilemma is that civilians will cooperate or not based on their perception of capricious violence, not whether or not the perception is true.

This is why the Information War matters:

Finally, it is crucial to remember that noncombatants measure relative caprice, efficiency, and brutality in punishment according to their perceptions. If one side is perceived as more capricious than the other, fewer noncombatants will cooperate with it. If one side is perceived as more brutal than the other, more noncombatants will have to be coerced into cooperating with it — that is, the side that is less popular because of its perceived brutality (and other considerations of popularity) will have to coerce more successfully in order to achieve the same level of cooperation. To some important extent, therefore, perceptions become reality.

Because perceptions are so important in counterinsurgency, capricious acts and the publicity of those acts can actually hurt the war effort. When supporters of the Coalition and the government of Iraq object to the widespread and one-sided publicity of purported American war crimes, it is not that we think, a priori, that these events should be covered up or that we care about the political fortunes of the Bush administration. Rather, it is because we know that anything that increases the perception of the counterinsurgency as capricious will actually hurt the war effort insofar as it motivates noncombatants to cooperate with the other side. Similarly, relatively muted publicity of enemy atrocities artificially dims the perception that al Qaeda kills capriciously and brutally. Both problems would diminish if the press, which has an enormous capacity to magnify perceptions, applied the same moral standard to both sides.

News reporting must track the overall reality as closely as possible. When it focuses on one or two abnormal events at the expense of everything else, it creates false perceptions. This has a cumulative effect that increases the number of dead and lengthens the war.