Crafting Moral Dilemmas: How to Make Players Truly Think
You know a campaign has worked when a player logs off still turning over a decision they made. Combat gets forgotten. Political intrigue fades, too. A genuine moral dilemma, though, the kind where they honestly weren't sure what the right answer was, tends to stick around. Players bring those moments up years later and argue about them with friends. Some people even write whole essays about a single choice. A few of those choices quietly shape how a player handles later stories, and once in a while, how they handle their actual life.
Sid Meier famously described games as "a series of interesting decisions," and in his GDC 2012 talk he offered a handy test for spotting a dull one: if a player always takes the first option, the decision isn't interesting. Moral dilemmas are where that test bites hardest. A real dilemma is the most valuable thing a campaign can produce, and it's also the easiest to fake. The internet is full of "morality systems" that dress themselves up as deep choices when they're really checking whether you read the dialogue carefully.
This post is my guide to building dilemmas that aren't tests and can't be optimised, the sort that make a thoughtful player genuinely struggle. I use these principles when writing a Game Bible for a custom EchoQuest world, and they hold up just as well at a kitchen table with friends. I've had a personal reason to think hard about this, too. EchoQuest keeps real game state, story flags and NPC relationships included, so every choice a player makes has somewhere to live afterwards. That made me ask what a choice worth remembering actually looks like.
The Anatomy of a Real Dilemma
A fake dilemma: "Do you save the village or let the bandits burn it?"
Nobody picks "let the bandits burn it." There's no tension at all. It's a good-or-evil test, and most players pass it without blinking. That's backed up by research, incidentally. When Andrew Weaver and Nicky Lewis recorded 75 people playing the first act of Fallout 3, they found that most players made moral decisions and treated the game's characters "as if these were actual interpersonal interactions." People are kind to NPCs by default. So a good-or-evil fork barely registers, and worse, the player senses they're being tested and resents the question. Fake dilemmas also damage a campaign's tone, because they tell the player that the GM thinks moral choices have obvious answers.
A real dilemma: "The village Elder helped the bandits in exchange for protection of his daughter, who is now wanted by the crown. The bandits are gone now, but the daughter is hiding in the village under a false name. Do you turn the Elder in, which will legally protect the village from charges of harbouring criminals but condemn his daughter as well, or protect his secret, which means the crown will eventually find a different scapegoat in the village instead?"
Now there's tension, and no clean answer anywhere. Whatever the player chooses, they lose something, and they're choosing who pays. The Elder did something wrong. He also did it for love. The crown's law is legitimate, even though its punishment is wildly out of proportion. And the village deserves protection, but protection costs people. No algorithm solves this, which is exactly the point.
The Four Ingredients
1. Legitimate values on both sides. Both options have to serve something the player actually cares about. If one option is clearly better, you haven't written a dilemma. You've written a puzzle with a right answer. The hardest dilemmas set two of the player's own stated values against each other. A character who prizes both loyalty and justice will be torn when those values demand opposite things. Likewise, a character who values mercy and also the safety of innocents will be torn when sparing one person puts many at risk.
2. No third option (initially). Players will always hunt for the clever solution that costs nothing. That's fine. Let them try, and let your AI GM honour creative attempts. In EchoQuest, the player can always ignore the suggested choices and type or say anything, and the GM is told to follow the player's creativity instead of herding them back onto a path. Still, the default state of the dilemma should offer no clean escape. If they find a clever third path that genuinely solves both sides, reward it wholeheartedly; that's their moment of virtuosity. Design the scenario assuming they won't, though, because most of the time, life-shaped problems don't come with third options. The player's job is to choose under real constraint.
3. Personal stakes. Abstract dilemmas ("save one person or five") are philosophy-class exercises. That one has a famous pedigree: the philosopher Philippa Foot introduced what became known as the trolley problem in 1967. Puzzles like it have a "correct" utilitarian answer and produce intellectual engagement, not emotional engagement. Neuroscience points the same way. In a 2001 brain-imaging study, Joshua Greene and colleagues found that personal moral dilemmas engage emotional processing far more than impersonal ones; pushing someone off a footbridge with your own hands lights up very different territory from flipping a switch. So dilemmas turn emotionally real when the people involved are named characters the player has already met and grown fond of. The Elder isn't "an Elder." He's the man who let the players sleep in his barn and told them about his late wife's apple pies. Twice, he slipped them rations they were too proud to ask for. Now they have to decide whether to turn him in. The same dilemma with strangers is barely a dilemma at all, but with the Elder it's a wound. (EchoQuest tracks every named NPC's standing with you on a scale from -100 to +100, with a short note on why it changed, so the GM remembers the barn and the rations too.)
4. Irreversibility. The decision should feel permanent. If the player can always undo it later, there's no weight to making it. The old gaming trick of "save before the choice, see both outcomes, keep the one you like" defeats moral choice entirely. I'll be honest about a tension in my own design here. In May I added single-step undo (the U key), because a misheard voice command or a misclicked choice shouldn't wreck someone's character. It only reaches back one turn, and the turn you undo has already used its AI credit. So it can technically rewind a dilemma, but my position is firm: undo is for slips, not regrets. If you undo a hard choice just to peek at the other branch, you've turned a dilemma back into a puzzle. The choice you made is the choice you made, and knowing that going in changes how you sit with the decision.
Types of Moral Dilemmas That Work
The lesser evil. Both options cause harm, so which harm is more acceptable? This is the shape most people picture when they hear "dilemma," but it isn't the only one. The risk is leaving both harms abstract. Ground them in specific named people and they'll hit with full force.
The betrayal. Keeping a promise or a loyalty collides with doing the right thing. The player promised to protect the prince, and the prince turns out to be corrupt. Or the player swore an oath to the temple, and now the temple is asking for something the player believes is wrong. These are especially powerful because they put the player's relationship with their own integrity on the line.
Certainty vs. hope. A guaranteed bad outcome against a chance at a good one, with catastrophe as the risk. Accept the certain loss of one person's life, or gamble on saving everyone and possibly losing them all. Mathematicians can argue about expected value forever. Characters who care about the people involved can't.
Justice vs. mercy. Punishing someone who deserves it, or giving them grace they haven't earned. The villain has been captured and sits at the player's mercy. They've done unforgivable things, and the player can have them executed with full legal cover. They can also let them go, knowing that mercy might mean the villain hurts someone else, or might mean the villain becomes someone different. There's no formula for this one.
The individual vs. the many. One person's wellbeing set against a larger group's. "Sacrifice one to save many" is the classic frame. The subtler version comes when the "individual" is someone the player loves and the "many" stay faceless. Brains don't actually run utility calculations when love is on the table, and Greene's footbridge results suggest as much.
The cost of staying. This is a dilemma the player doesn't realise they're in. They keep backing an ally who's slowly turning into someone they don't recognise. Walking away might save the player but condemn the ally. Staying might condemn the player and barely help the ally. These are the slowest dilemmas of all, and sometimes they unfold across an entire campaign.
Timing and Delivery
Dilemmas land hardest after investment. So don't put a moral choice in the opening scene, because the player hasn't met anyone yet and doesn't care. Plant the seeds early instead (introduce the characters, establish the clashing values) and bring the choice to a head once the player is emotionally committed. Ideally, a dilemma uses people the player has spent at least three or four sessions with. For more on making those people worth caring about, see my post on writing compelling NPCs.
Presentation matters as well. A great dilemma doesn't arrive with "you must choose." It arrives as consequences: the crown's investigators ride into town, someone recognises the daughter, and the Elder asks to speak with the player privately. The player realises they're at a crossroads through events, not announcements. As a result, by the time the GM asks "What do you do?", the player has been chewing on it for ten minutes already. EchoQuest's GM ends every scene on a direct question like that, which suits a slow-burning dilemma nicely.
One more thing: don't telegraph the right answer. If your campaign has tells (NPCs nodding approvingly when the player gets it "right"), players stop wrestling with the choice and start optimising for the reaction. The GM should treat both major options in a real dilemma with respect. Neither should be punished as obviously wrong, and neither should be celebrated as obviously right. Both should simply be lived with. When I rewrote the GM's instructions, I put "summing up or moralising at the end of a scene" on its list of things to never do, partly for this exact reason. A lecture after a hard choice ruins it.
Have you ever made a call in a game that you still aren't sure about? That itch is what you're designing for.
After the Choice
The most important phase of a dilemma is what happens after the player has chosen. The world should react. If the Elder is turned in, he should be executed off-screen, and his daughter should turn up later, alone, with a face the player can't forget. If he's protected, he should keep finding small ways to show his gratitude, each one quietly reminding the player what they did. The choice doesn't end at the moment of decision. It ripples forward.
In EchoQuest, you set this up in your Game Bible by seeding the relevant relationships and tensions, and the GM builds toward the dilemma from there. Afterwards, the choice gets recorded in structured game state. Story flags hold the decision itself, and each NPC's standing (plus its little note) holds how they feel about it. On top of that, the GM reads a running campaign summary every turn. The GM's instructions are blunt about using all that: past choices shape what happens now, and old threads should come back "when it hurts or helps the most." Character progress is also saved on the server, so a consequence can catch up with you even if you resume on a different device. The more clearly you establish who the players will care about and why those people's interests collide, the stronger the eventual choice will be. And the longer it'll linger once the session ends.
Want to test your dilemma-writing on a world that already exists? Pick one from the Adventure Library and pay attention to who you start caring about.
