The COIN theorists tend to have a more sophisticated argument (and my mood deteriorates when I think of TV pundits), so I'll address theirs:
The typical line of thought is like this (no quote):
The pro-Western powers (the government) needs to earn the local's support and allegiance by providing services and constructing objects for better quality of life. This construction work and the maintenance of public services can only succeed if protected properly against enemy (Taliban) attack (and blackmail in case of NGOs).That's nice in theory, but it fails my plausibility check.
Western troops need to move in, defeat (chase away) local insurgents and provide security to the pro-Western efforts to stabilize the area through popular support for the Western cause.
Western troops ride in armoured vehicles and live in guarded forts, yet they still suffer casualties.
How could Western troops - even assumed a high force density - hope to provide security for a population that outnumbers them 11 million (Afghan Pashtu) to much less than 600,000 (projected target strength Afghan National Army, Afghan National Police plus maximum imaginable quantity of pro-Western foreign combatants)?
This (future, dream) ratio doesn't look particularly terrible - until you consider more than half of the pro-Western troops are non-combat troops (much more among the foreigners) and remember that you would need to guard effectively every marketplace, every school, every isolated hamlet to provide real security and eliminate all safe havens.
The security problem isn't limited to the protection of pro-Western employees and institutions, after all. You would also need to protect the general population.
The "strongest tribe" idea means that locals ally with the strongest (and reliable) power and despise, even attack a weak or unreliable power.
It's a close relative of the "we must provide security" and "Afghan surge" concepts because it's all at least in part about having more forces in place.
I consider this "strongest tribe" idea to be very misled. It's not about strength or reliability. It's about threat value instead. There's little to no booty (the classic tribal warfare motivator) to gain in the Afghanistan conflict, therefore choosing sides is either about power (relevant only to a minority) or security.
The "strongest tribe" concept doesn't pass my plausibility check, a "most threatening power" concept could do so.
So there's our predicament: We are past the civilization stage where taking hostages and mass murder were considered viable tactics of warfare. We are not threatening (still dangerous, but not in a directed, useful and predictable way). We are not able to fully protect against other threats because such encompassing protection is impossible. The enemy will always find a way how to hit his targets in his own country/region.
This problem has its limits, of course. The Uzbek and Tajik communities are not as much inclined to bow to the Taliban as the Pashtu communities which often share culture, political goals and ethnicity with the Taliban.
The informed part of the pro-"we must provide security" crowd simply hopes that the difference between the Uzbeks/Tajiks on the one hand and the Pashtu on the other hand aren't that great, so imprefect security would suffice to even turn the tide in pro-Taliban communities.
I don't agree, for I do not support warfare that's critically based on hopes and dreams.
Some people assert that warfare should be continued until you found a way how to win. That may fit to those who remember their nation's history of first floundering and then winning in war after huge expenses. It doesn't fit my thinking, though: I expect wars to be only fought if waging war is the lesser evil in comparison to peace, which means that I don't accept high resource expenditures without having equally high expectations for the advantage gained by warfare.