Let us suppose for a moment that I’m running two NPC in a given 5E game. One is lying and the other being honest. The players suspect both of lying so they decide to roll an Insight check which could either be rolled in the open by the players or rolled in secret by the DM. Two characters are making this Insight check where one is naturally trusting and the other naturally suspicious. The matrix of results for such a roll would like like this:

Matrix of Insight checks for natually suspicious and trusting characters against lying and trustworthy NPCs.

	lying NPC		honest NPC
	failed check	successful check	failed check	successful check
sussy believes	lying	lying	lying	truth
trusty believes	truth	lying	truth	truth

The Insight check reveals what we would expect which is the bias of the characters in question. Before you argue that this is an over simplification let’s acknowledge that during the course of the game a DM would not necessarily treat the result of in Insight check as a lie-detector. Instead in the case of the honest NPC being honest it would be better for the DM to have the character recall some supporting fact and in the case of the lying NPC lying have the character recall a contrary fact.

Despite that, if the two PCs in question succeed on their checks and know they succeeded then there is clarity akin to a lie-detector. One success and one failure may also bring accidental clarity. For example the trusting character succeeds on a check against the lying NPC while the suspicious character fails. If the players are able to sort the characters natures they again achieve clarity akin to a lie-detector. This matrix would of course be further complicated in real-life by a more likely group of 4-7 characters each on the the simple trusting/suspicious scale above.

So far though we’ve avoided the inverse case. What happens when the trusting character fails a check against the lying NPC or the suspicious character fails a check against the honest NPC? Should the DM have the trusting character recall a fact in support of the lying NPC‘s lie? In other words should the DM in their capacity as DM lie to a player?

I don’t believe a DM should. But as the DM you’re in a tough spot. You could give some shorter explanation to the player in this case but the incongruency in your response will alert the player to their character’s failure even if the dice did not.

In the above example the DM may remind the player of some fact their character knows in support of the honest NPC‘s story but when the DM replies, “seems legit” to the failed check against the lying NPC the player knows their check failed.

The DM could of course shorten their positive response to something like:

Matrix of Insight checks for natually suspicious and trusting characters against lying and trustworthy NPCs with terse DM response.

	lying NPC		honest NPC
	failed check	successful check	failed check	successful check
sussy	seems false	seems false	seems false	seems true
trusty	seems true	seems false	seems true	seems true

Please don’t quibble with the table. These statement are easily replaced with something else like noticing the NPC‘s stutter which only happens when lying. But no matter the short hand we’re really back to the main fault of Insight. It is a lie detector and worse completely unnecessary.

If I invent an NPC that stutters when he lies then I’m going to stutter when I talk to my players as that NPC; rolling absolutely not required.

The problem ultimately is this that Insight either fails to bring any more clarity making it unnecessary or brings 100% clarity to the game removing all mystery. And that is incredibly boring.