Maybe that's because we usually already know what a D&D campaign will be about. I mean, the only times groups engage heavily in that discussion (IME) is when they are planning on deviating significantly from D&D's "script." "Let's do an evil campaign" "Let's do an all-thieves campaign" or something similar. Otherwise, you're generally free to roll up your Dwarf Fighter start play...many times even without a name.
Interesting points.
I rarely (as in "can't remember ever seeing it") see a player choose a class or race in order to signal something to the DM about the game they want to see.
For me I can remember a fair bit of this. A player plays a dwarf because "dwarf" = "stern, no-nonsense but dependable do-gooder". A player plays a drow chaos sorcerer who worhsips Corellon because "chaos" = "go crazy wild" while "Corellon-worshipping drow rebel" tempers that with "but still somewhat sympathetic and ultimately well-meaning". A player plays a paladin because they want to see tests of faith.
Or, in Oriental Adventures, a player choose bushi rather than samurai because he wants to be a free-wheeling rather than a duty-bound warrior. And then there's the classic all-thief campaign: I think the thief is one of the more hook-heavy classes in classic D&D, and I think one of the reasons for its notoriety is that the standard dungeon adventure doesn't really pick up on those hooks.
What I'm a little less clear on, is what happens after a campaign gets going. Even in groups that start play fairly divorced from dramatic thinking and focus almost entirely on exploration, I regularly see that they develop attachments to locations, NPCs, etc. that become fine foci for dramatic play.
<snip>
Should the rules address or acknowledge that in some way?
I could imagine a version of a PrC/Paragon Path rule that tried to acknowledge or address this. For instance, when a certain trigger is activiated (an ingame event? reaching level 6? the GM deciding "it's time"?), then you get your "attachment" PC feature.
A really primitive version of the feature might be something like - when the thing you're attached to suffers, you take d10 psychic damage unless you step in to protect it, in which case you get advantage on your relevant d20 roll. (The reason I'm going for damage rather than an action penalty is that, in D&D at least, a penalty to action tends to be so significant that it's tantamount for coercion, which is not what you would be looking for here.)
It can be tricky. If a player builds a character with a a really high lockpick score, is it because he wants to engage thievery as a theme or because he doesn't want to deal with locks as a challenge?
I think this is an excellent point.
In BW, this is something where Beliefs and Instincts can do a bit of work. Suppose a roguish PC has a high lockpicking skill, and also a Belief that "All secrets should be laid bare, and all beauty out in the open for the world to see", that is practically instructing the GM to put locks in front of the character, which have things behind them that mere mortals were not meant to know!
Whereas if the PC has an Instict "Never shut a door without confirming I have a way out" then this might suggest that the player doesn't want escape scenes to be a big part of play, and a high lockpicking skill on this PC might send a similar signal: what this PC is about isn't picking locks, but rather not being trapped.
In the end, though, there is probably no substitute for asking - my own preference is not to talk about this meta stuff in the middle of play, where it can interfere with the immediacy of the ingame experience, but before or after the session proper. Or, if the player is not sure about his/her own preferences, test the water one way or another, see how things pan out in play, and follow the relevant cues.
One thing I'd like you to expand upon, if it's not too much trouble, is the part about Beliefs, Instincts, and Traits; what are they in the context of the game, and how do the players go about leveraging these things?
Others have said a bit about this. I've also given an example above of how a Belief and and Instinct might look, and the different signals they might send.
Beliefs are about stake-setting, by reference either to other character (other PCs, or NPCs with whom the player has purchased a relationship for his/her PC) or more abstract values. A player chooses Beliefs for his/her PC (three of them is the default number), and various PC features can mandate or leverage Beliefs in various ways: eg all Dwarves must have a relationship, and then must have an oath sworn to that relationship, and a Belief centred upon that oath. Faithful PCs must have a Belief that is a statement of their faith (and if they lose that Belief they lose their access to their Faith attribute).
A player can change Beliefs at any time, subject to GM delay on a change if the timing is viewed as too exploitative. Manifesting a Belief in play earns Fate Points (eg if my Faithful PC worships Pelor, and my Belief is "The light and the flame are my guide", then when I light a torch to make my way through a dungeon I earn a Fate Point; dramatically playing out your inner turmoil when a situation calls for you to act contrary to a Belief can earn you a Persona Point (a type of super-Fate Point); accomplishing a goal set out in a Belief earns you a Persona point. So the GM is expected to set up situations which will create room for all this (being led by the players, given that they choose their Belief; but the players in choosing their Beliefs should of course be having regard to the game concept); and the players then earn Fate and Persona Points by following those leads.
Bcause Beliefs do a range of things in the game, when I look for analogues in my 4e game I'm looking at a variety of different things. Goal-based Belief, for instance, in 4e correspond at least roughly to player-designed Quests.
Or if I think about a Belief based around a relationship, I think of the drow PC in my game who is a member of "The Order of the Bat", a secret society of Corellon worshippers who seek to undo the sundering of the Elves. This was all introduced into the game by the player of that PC, much as in BW you might build your PC with relationships to a secret society with that goal (in BW that's called purchasing an Affiliation) and then build a Belief around it ("Together with my fellow members of the Order of the Bat, I will see the Elves brought together once more!"). I use this hook in designing encounters - with elves, with drow, etc - and the player plays up to it also (eg constantly addressing the Wood Elf PC in the party as "My Elven brother", having secret little conversations in Elvish, etc). Unlike BW, however, there is no mechanical reward to the player for doing this - which also means the player of the Wood Elf has no mechanical incentive to pay it back (whereas in BW, even if he wouldn't earn Fate Points for his PC, he has a mechanical incentive to create situations that feed Fate Points to fellow players). This is part of what I have in mind when I talk about 4e narrativism being "vanilla" narrativism.
But when I think about the Raven Queen devotees in my party - the aforementioned Wood Elf, plus a paladin - who hate Orcus and hunt down undead, they get their radiant damage benefits against undead. So in that sense there is a mechanical pay off to manifesting their Belief in play. But there is nothing in 4e mechanically analogous to earning points for dramatically playing against your Belief - the closest I get to this is setting up conflicts (especially skill challenges, but also social negotiations more generally and some combats too) in which making progess requires choosing fairly clearly one way or the other. The players aren't forced to choose against their Beliefs, but they are at least forced to be made conscious of the need to choose for or against their Beliefs. The simplest examples of these are invitations to negotiate from NPCs opposed to the PCs' Beliefs, in circumstances in which thoe choice to fight rather than negotiation isn't an easy one, for whatever reason (eg balance of forces, social context) applies in the particular situation.
With Instincts and Traits, in addition to what's been said you get Fate Points when you trigger an Instinct so as to cause trouble (it is always the player's choice whether or not an Instinct trigger), or when you embellish your roleplay with a character Trait so as to make life more difficult. An example would be triggering your "Always draw steel when I see my nemesis" Instinct when you meet you nemesis at the court of the king; or suppose you have the Clumsy trait, and suppose also that you have got the key to the concealed rear entrance from the traitor, and you have just crept past the guards and are about to sneak in, then an example would be declaring that your PC is so clumsy that s/he drops the key down the sewer grate as s/he crosses the courtyard.
As the OP indicates, the GM (as well as other players) is meant to have regard to Instincts and Traits in framing scenes. An example in the rulebook is of a dwarf with the instinct "In a cave-in, push the young ones to safety" - if a player builds a PC with that Instinct then the GM
is obliged to have a cave in.
(Some Traits are not like Clumsy, and are closer to D&D-style feats or class abilities - the core rules aren't clear on how these other sorts of Traits relate to Fate Points, but later commentary from the designers suggests that when using a Trait can be its own mechanical reward (eg using a call-on Trait for a re-roll) then no Fate Points can be earned from that Trait.)