How Far Can an Unreliable Narrator Go in Interactive Fiction? Separate Story Judgments from Action Feedback
An unreliable narrator can insist that a door will never open again, but action feedback must tell the player whether the bolt has been removed and where it is now. A narrator-and-feedback worksheet preserves the narrator’s bias while letting players confirm and retrace their actions.
Introduction
An unreliable narrator can mislead players about motives, danger, and the future. Essential action feedback, however, must explain whether the last command was carried out, how far the action progressed, and what changed about the objects involved. The narrator can say, “That door will never open again,” while the feedback clearly states, “You have removed the bolt.” Suspense can wait behind the door. It should not leave players wondering whether they actually finished removing the bolt.
When revising, write the action facts independently first, then let the narrator interpret them. The following narrator-and-feedback worksheet takes that process from start to finish. The setting, characters, and action outcomes are fictional teaching examples, not real user cases, empirical findings, hands-on tests, or descriptions of DramaFork features.
1. Establish What Players Need to Know Next
Suppose the player stands before the door of an old house, with a bolt fitted horizontally across its inner face. The player chooses “Remove the bolt.” The action is complete: the bolt is in their right hand, the door remains closed, and they have not yet tried pushing it. The narrator is determined to stop them from leaving but will not explain why.
The original draft reads:
You grip the cold iron. Don’t bother. That door will never open again. You haven’t changed a thing.
The problem lies in the last sentence. “You haven’t changed a thing” could mean that fate remains unchanged, but it could also mean that removing the bolt failed. Players cannot use this information to decide whether to push the door or keep pulling at the bolt. The author wants them to doubt their chances of escape, but the wording erases the object’s state as well.
Start with a concrete question: after reading this passage, does the player know whether they need to try removing the bolt again? If the answer is uncertain, add the action facts. Whether there is a way out beyond the door, why the narrator is afraid, and whether pushing the door will invite danger can all remain hidden.
2. Divide Each Action Across a Four-Column Worksheet
Fill in “Actual action” first, then “Facts the player can confirm,” and write the narration last. This avoids crafting an atmospheric line first and then changing the action outcome to accommodate it. The rows below are consecutive steps in the same fictional scene. The last two test whether the revision can support subsequent actions.
| Actual action | Narrator’s judgment | Facts the player can confirm | What must not be conflated |
|---|---|---|---|
| Remove the bolt and hold it in the right hand | That door will never open again | The bolt is out of its slot; the door remains closed | Replace the result of removing the bolt with “futile”; imply that the bolt is still in its slot |
| Push the door with the left hand; it opens outward a crack | The door is only luring you closer | The push has been carried out; a gap has opened | Turn “should not have opened” into “did not open” |
| Place the bolt on the ground by the player’s feet | You have finally discarded your last protection | The bolt is no longer in the player’s hand and is by their feet | Say it “can never be found again” when it has not been lost |
These four columns are an author’s worksheet. Players do not need to see all of them. When reusing it, record only one action and its direct changes per row. Do not squeeze “remove the bolt, push the door, and walk outside” into a single row, or it becomes difficult to identify which steps are complete and which have not happened yet.
“Facts the player can confirm” is not a list of the author’s secrets, either. Include enough facts to support the next action. There is no need to disclose a mechanism behind the door, and removing the bolt certainly does not justify promising in advance that the door will open. Removing one obstacle and completing the whole objective are different outcomes.
3. Establish Two Voices Through Sentence Structure, Even in One Typeface
The easiest approach to apply consistently is to give feedback a fixed pattern: “object acted on + completion status + current location or state.” Leave evaluations, speculation, and predictions to the narrator. Even without different colors or typefaces, the wording can signal which kind of information the reader is receiving.
The revision can read:
[Action] You have removed the bolt and are holding it in your right hand. The door remains closed.
[Narrator] That door will never open again. All you’ve done is pluck an insignificant thorn from this house.
The narrator’s absolute claim is deliberately retained. An unreliable narrator need not qualify every sentence with “perhaps.” The boundary comes from consistent responsibilities, not a hesitant tone. Players can believe the prediction or challenge it by pushing the door next, without having to guess whether removing the bolt succeeded.
If explicit labels disrupt the narrative too much, consistently place a short feedback paragraph first and the narration in a separate paragraph afterward. Explain their respective roles when this format first appears. The trade-off is that recognition depends more heavily on reading habits. Do not swap the order midway through or suddenly introduce wording such as “you seem to have removed it” or “you apparently obtained it” into the feedback.
When sensory uncertainty matters, narrow the scope of what is confirmed. For example, while the player is feeling around in darkness, write, “You have reached into the bolt slot but have not yet identified the object inside,” rather than declaring that they have obtained the bolt. Feedback confirms the portion actually completed. It does not supply answers for what has not happened or has not been confirmed.
4. Distinguish Pending, Failed, and Completed Actions
“Command received” and “action completed” are not interchangeable. If the scene genuinely schedules the action for later, write, “You have selected ‘Remove the bolt’; the action has not yet been carried out.” Only after completion should the feedback say, “The bolt has been removed.” Do not offer reassurance through completed-action wording and later explain the unperformed action as the narrator’s deception.
In another fictional branch, the bolt is stuck and the attempt to remove it fails. You could write:
[Action] You pull on the bolt but cannot remove it; it remains in its slot.
[Narrator] See? It knows its master. And that isn’t you.
The failure is clear, while the narrator can still interpret a mechanical obstruction as rejection. Players can then choose to inspect the slot. If all they receive is “It rejected you,” they cannot distinguish between a command that was not carried out, an action that failed, and an author temporarily preventing the story from advancing.
Feedback also oversteps when it quietly adds a conclusion: “The bolt has been removed. You are safe now.” The first sentence confirms an action; the second guarantees a situation. One does not necessarily follow from the other. Unless safety is genuinely an established rule at this point that needs to be communicated, delete that claim or recast it as the narrator’s subjective assessment of safety.
5. Keep Actions Traceable When the Narrator Is Exposed
Now finish the sequence in the worksheet. The player pushes the door, and it opens a crack. The narrator changes tack: “The door is only luring you closer.” The player puts down the bolt, and the feedback states that it is by their feet. The scene has now disproved “will never open again,” but all three actions—removing the bolt, pushing the door, and putting the bolt down—still stand. Exposing the narrator changes the player’s understanding of that voice, not whether past actions occurred.
Before handing over the draft, temporarily hide the narration and read only the feedback. Answer, in order: What was done? Was it completed? Where is the object? Which step remains undone? Then reverse the exercise and read only the narration. Check that it conveys bias, manipulation, or concealment rather than providing the sole confirmation of an action. This is a manuscript review method; it does not justify claiming that the player experience has been validated.
This division suits interactive fiction and text adventures in which players need to track objects and actions. If a work deliberately distorts bodily perception or memory, retain a narrower stable boundary—for example, make clear that input has been received but that the narrated outcome remains unverified. Do not treat the same kind of feedback as fact at one moment and a lie at another without giving players a convention for distinguishing the two. Ultimately, players should be able to say: I did remove the bolt. Why the narrator keeps insisting that the door will not open is something I still don’t know.
