Suppose an NPC approach the PCs and tell them that some bandits stole a trinket precious to him. He would be very eternally grateful to have it back, but have nothing to offer as a reward. Is it railroading? Of course not : the players can decide to do it or not. Let’s call that a free choice : the only thing that matters is the PCs own motivation to act (be it altruism, love of adventure, or even, why not, boredom).
Now suppose instead that this NPC ask the same thing, but offer a substantial reward for it (money, magic, influence, whatever). Does it becomes railroading? Sure, one choice looks more appealing than the other, but there is still a choice. So no railroading. Let’s call that a rewarded choice : at least one choice comes with a reward that the PCs want, so it nudges them toward it.
Now suppose the same situation, but this time, the NPC comes with a small party of mean-looking and dangerous allies. It’s clear that if the PCs still refuse, blood will be shed. It’s also quite clear that the PCs will have a hard time fighting this one : death is very possible. Does that becomes railroading? The choice is still there, and just like the second time around, one of the choice seems much more interesting. Is it railroading? The cost of one choice is quite high, but a choice can be made. So, no railroading…? Let’s call that a costly choice : at least one choice comes with a price that the PCs would prefer not want to pay (possessions, blood, reputation, etc.), so it nudges them to avoid it.
Now suppose one last time the same situation, but this time, the NPC comes with a powerful magical device that, at least with the means the PCs have, makes him impossible to harm and able to deal extremely high damage. It’s clear that if the PCs refuse to help him, they will die. Again, the choice is there : help him, or die. So, no railroading… That can’t be right. Let’s call that a do-or-die choice : at least one choice comes with a price so high that paying it would mean the end of some PCs (death, but also class powers, maybe will to live), so it almost force them to avoid it.
I sometimes am under the impression that for some DMs, giving the PCs choices affecting the narrative to make amounts to not railroading (for the purpose of this article, “choice” means specifically that, and not, say, tactical ones). But I think the examples just presented show that having a choice may be necessary, but is clearly not enough. Clearly, if all the choices are “do-or-die” ones, you have a (thinly-disguised) railroad. It’s like saying that the first Mario Bros. is open-ended because you can always choose to jump in a hole. On the other hand, sometimes, the fiction demands a “do-or-die” choice to be coherent and feel reasonably “real”, so presenting them is not automatically railroading your players. Actually, not presenting them could sometimes feel like railroading. For example, suppose you insult the king to his face in front of the court. “Apologize and quit or be sentenced to death” is perfectly fine, even if not apologizing clearly means the PCs death (because the exgremely deadly king’s guards are in great number); anything less and it feels like the DM refuse to let the PCs be in bad terms with the king. That becomes very clear when we think about classic dungeoncrawling or hexcrawling games where a world of adventures awaits, but one that does not guarantees that the stuff you’ll encounter will be suited for your power level.
But the type of choice does not stop there. There is also the element of optimality to take into account. A choice where there is an optimal choice is not really a choice; it’s more a question of calculation. So a game that presents choices with one predetermined optimal decision veers more into the railroad territory. An open-ended game DM make sure that a decision is not “punished” just because it’s not the one that was planned as the optimal one. For example, if the DM planed on the PCs paying fisherman to cross the sea, but they decide to steal the boat instead, the DM should offer a reasonable fight there, but not send the kingdom’s armada after them or punishing them with very dangerous weather.
Moreover, choices exist even when the DM did not present one. A railroad can offer some predetermined choices at predetermined points, but an open-ended game is always open to the creative input of the players. That does not mean that everything tried will work, but this at least means that those choices will not always be brushed aside or heavily penalized.
Lastly, I think that hooks to adventure (or plot-hooks) should never be do-or-die ones (at least, not unless the players are fine with it). An open-ended game lets the players decide what they want to do, and a hook is meant as a way to let the players (and PCs) know that there is something to do. A hook that forces one choice is the very definition of putting the players on a set of rails. In the same vein, hooks should generally avoid being costly choices where the cost is paid when refusing to take the hook. It can sometimes make sense and can even underscore important facts about the people in the game world (will you help this person or lose its friendship?), but it quickly starts to feel like the DM just don’t want the players to choose their adventures if used regularly. Other kind of choices make great hooks.
So presenting choices to players does not make a game open-ended; it takes more than that. Every choice presented should let multiple decisions have a reasonable chance of success; nudges should not always clearly tip the scale; and these choices should be implicitely given all the time, even when the DMs thinks that there is nothing to choose. All of this is part of an open-ended game : these are meaningful choices.