Power-seeking behaviour is not, of course, universally a bad thing. It is entirely in character for many PCs to wish to have a measure of control over the game world, and to improve their personal abilities; indeed, most game systems are structured to allow this. However, in several cases this behaviour can cause problems: mainly, when it causes the character concept to be bent or stretched to accommodate a power gain.
The most significant case is that of loopholes. Every game system has them, although some are notorious for them; and some players will take relentless advantage, claiming if pressed that "it's in the rules". (Naturally, the solution is simple: "We're not playing Advanced Demons and Deviltry, we're playing my game.") This is often an early warning of potential problem players; however, it is only useful as a diagnostic test. Player who insist on playing the rules rather than the character will eventually cause difficulties, and, historically, have often had to be encouraged to leave the group.
This is a good example of asymmetrical information and goals. In a time when worlds usually come with rule packages attached, most GMs choose the world and make do with whatever the rules may be. However, the player sees both the world and the rules, and the rules are likely to have a greater effect when the world is still unfamiliar; so, for example, if the world book says expert martial artists are rare, but the rules say they're easy to play and unbeatable, players will want them as PCs. Conversely, if the GM likes alchemists but the rules say they're not survivable characters, no-one will play one. In essence, the rules guide players towards specific character concepts and styles of play; concepts at variance with this will be seen as inferior within the rules framework. This can cause extreme friction within a group, when some players are going by the rules (and taking advantage of loopholes) and others are attempting to role-play unusual characters that the rules may not handle well - or just normal characters that do not take advantage of new, high-powered rules. In cases I have observed, I have found that this division tends to be on the basis of experience, with the comparatively inexperienced players tending to rely heavily on rules.
As an extreme example of this clash, FASA's Shadowrun (first edition) presented a world in which people carried pistols and submachine guns for personal defence; thus the NPCs in published adventures tended to carry such weapons. So far, so good. In contrast to this, the rules system rendered any weapon smaller than a light machine gun almost useless for any purpose. Some PCs, naturally enough, started to carry around light machine guns.
At this point, the players whose characters weren't carrying them started to complain that they didn't have much to do in combat scenes. Rather than admit their mistake and fix the rules, FASA responded by bringing out a book of bigger and tougher weapons, trying to increase the lethality of small guns but actually starting an arms race when the same technologies were applied to the big guns; players of non-weapon-based characters then complained loudly that they were being left behind, and got their own power-increasing books (the infamous Grimoire being a prime example).
The original problem was left unsolved, as the NPCs in published adventures (and anyone else using equipment that made sense in a non-rules-mechanic context) fell further and further behind; it seems likely that this was a proximate cause of the munchkinism that developed in several Shadowrun-playing groups. After all, when the entire Élite Guard Force (TM) doesn't even slow you down, why shouldn't you try to rule the world?
A less severe example occurs in R. Talsorian's Cyberpunk 2020. Skill purchase during character generation is at constant cost/level, while improvement once play has begun is at increasing cost/level. Hence it makes game sense to specialise in one or two skills, and get the others at minimal levels, improving them later. This is worsened by the fact that a character's equipment budget is predicated on the level of a single skill...;
It is possible to bias the game such that the rules become less relevant, mainly by ignoring rules that do not fit the GM's conception of the game world; this is something I often do in my own games, but this can alienate less-experienced players who expect to be able to see the rules framework - and would quite justifiably be annoyed if, once they had generated characters to take advantage of the rules, I changed rules on them without warning. A helpful factor is that players are often unfamiliar with the rules I use, so that they are not aware of loopholes and do not follow the channels into which the rules might otherwise push them.
This is one of the main arguments for "realism" (i.e. a high degree of accurate simulation) in role-playing rules; if players can use their own skills and knowledge, even at a low level (e.g. "they're carrying rifles so they may well be fairly tough opponents" as opposed to "they're carrying rifles so they're practically unarmed"), they are less likely to be alienated by the world. This is certainly the basis on which I run games: whatever the system, a realistic world-view takes priority.
Even problems that are perhaps not large enough to qualify as "loopholes" can cause difficulty. One example that has arisen recently in my own games is the use of skill chips. These are a method for gaining levels of skill without taking time to train or spending accumulated experience: a chip or program, usually obtained with little difficulty, gives the character a specific skill, usually at a low level.
Clearly, this can lead to undesirable situations, especially in wide-ranging games where many low-level skills are more useful than a few high-level skills. Why should anyone learn a skill other than his main specialisation? Why doesn't everyone use them? Power-seeking players will rapidly invest in a large chip library, to the exclusion of players who don't wish to take this advantage for character reasons; and to some extent they cannot be faulted for it, as the rules specify that it is worth doing.
Of course, if one were to give out all this information during character generation, and change the game universe to take account of it, this would be fair enough. However, in my own experience the acquisitive behaviour first occurred several sessions into the game - and I had rapidly to find reasons to reduce the advantage of skillchips, without rendering them useless, or alternatively simply to write them out of the world.
My solution in this case was to add several restrictions on skillchip use: chipped skills can be recognised and outdone by someone who recognises the chip, conflicting chips lead to conflicting impulses, chips cannot be used continuously, and only limited skill levels are available. They are still useful - for example, for picking up languages and general tourism information - but not as overpowering as before.
I feel that several lessons can be taken from this. Firstly, I try never to tell players simply that something cannot be obtained; rather, I tell them why it is unavailable. Secondly, this case emphasises the necessity of knowing one's technologies; I already had a strong mental model of the way skillchips would work, so I could rule consistently on the new limitations. Thirdly, it is absolutely vital for the GM to retain mastery over the rules, so as to present a universe without major conflicts between world and rules.
In summary, a major cause of game-disruptive, power-seeking behaviour appears to be discrepancy between the GM's vision of the world and that presented by the rules; players, especially inexperienced ones, are likely to see and play according to a world-construct created by the rules, whereas the GM will see primarily his own vision, and mismatches can cause severe difficulties. A solution to this involves reconciling the world and rules, either explicitly (by changing or modifying the rules system) or implicitly (by fudging results on a case-by-case basis).
Roger Burton West