Modifiers and bounds on possible results. A poor argument will mean negative modifiers to the roll, and an upper bound on the result you can achieve. You could think of arguments, setting, relations, etc. like the weapons and battlefield and the reaction you want to achieve as the target's defenses. If you want a tank-equivalent reaction, you need recoilless rifle-equivalent arguments. If you only have pistol-level arguments, no matter how well you roll will you overcome the target head on - but like you could sneak up on the tank and gun down the crew opens the hatch, maybe relations, blackmail etc. can get the job done. This explains it well to the players and makes both PC skills and player creativity matter, in the same way that to be effective in combat the PC needs skills and equipment while the player still needs to come up with the tactics. I would also note that it is one thing for the players to rationally see through the NPC's scheme, but quite another for the PCs to resist the charismatic NPC's con. Plenty of bright people have been had in con schemes that - when viewed from the outside - seem totally transparent, or been seduced (or themselves seduced someone!) and ended up in horrible trouble even though the outcome seemed obvious. It is too much of a roleplaying challenge for some though, and you might have to just make do with NPC reactions (basically, if the NPC "wins" and the PCs don't play suit, the NPCs find their behavior highly inappropriate or dishonest). Also, making most important social actions require several rolls is a good idea. If you wanted a combat-heavy campaign it would be a really bad idea to have a lethality level where one bad roll insta-gibs a PC. The same with social stuff. Don't just have a major interaction end up with a roll - roll regularly, and have the PCs and NPCs adjust their reactions accordingly.