L.A. Noire: Meaningful Conversation

We’ve been playing LA Noire quite a bit at home, and enjoying it.  It’s also driving me batty.  I’ve been a long running fan of the story telling dialogue tree genre, going all the way to it’s heyday in the 90s – KOTOR back to Blade Runner, Laura Bow, even Star Control 2.   I absolutely love it as a coach game that’s more focused on dialog choices.  It works really well as a group game.

The core of this genre is the conversation system.  LA Noire is  a quantum leap forward here.  The facial performance is really remarkable.  We’ve still got a long way to go, but LA Noire marks a turning point, the first time human facial animation itself drove the gameplay of an entire 10+ hour game.

I think the designers made 3 key mistakes though on the fundamentals, mistakes that have been made in the past and I hope won’t be made again.  They’re easy to make.  It’s just a dialog system, right?  But dialog and morality systems are incredibly tricky to get right.

1.  Be clear what the player’s choices mean.

This is probably the easiest mistake to make, and one of the hardest to get right, because every single choice you write has different context.  In LA Noire, you can answer the suspect with “Truth”, “Doubt”, or “Lying, and I have evidence to prove it”.  These sound clear, but sometimes it means:

  • “I, as the detective, think the suspect is being untruthful”, or often
  • “I, as the player, saw the suspect do shifty eyes”, or
  • “I, as the detective, think that is a factually incorrect statement (whether or not the suspect believe it)”, or sometimes
  • “I, as the player, think the designer is trying to trick me and I should doubt them”, and sometimes
  • “I, as the player, think the suspect believes that is true, but could be lying to me, and I need to doubt them to determine that.”

It’s not at all clear which of these is the right interpretation of the system, and it often changes from question to question!  I have this weird sense that the different designers didn’t know either.  This is equivalent to telling the player “A jumps, now go jump over walls”, and then having jumping over walls sometimes succeed, sometimes fail, and sometimes outright kill you.  It makes the game frustrating.  Compounded by…

2. Correct and Incorrect make each question a game

The game judges you, and judges you immediately.  Each question has a particular right answer, and it informs you immediately (with a great musical cue) if you got it.  The interrogation becomes not about the success of any conversation or the case, but the success of each individual question.  In normal system design, this would be good.  Immediate feedback!  But in LA Noire it makes the previous problem worse.  It makes the confusion around what your choice meant immediate and obvious. It becomes a game – “which type of choice is this one?  Ooooh, got it wrong, shoot.”  I am now more directly trying to guess what the heck you (the designer) meant.  And if I guess wrong, egg on my face, irrespective of if it’s actually the wrong call for the larger case.  Maybe I thought they were telling the truth, but wanted to push the suspect just in case I am wrong?  That’s perfectly good detective work, but in LA Noire that’s w.r.o.n.g.  Any deviation is punished, and I have to trust the designer that playing their way gets the best result, breaking the 4th wall.  But I already have learned not to trust the designer (see problem 1).  Oh no!

Remember how I said this was notoriously tricky?  Player intent is scary stuff to try and predict.  I think it’s possible, but it’s hard, and you have to be a heck of a lot more accurate then LA Noire is.  Immediate Correct/Incorrect feedback has to be 100% clear on why it was correct or incorrect, or it will backfire.  And furthermore…

3. You only get one shot

In LA Noire, if I get it wrong, I can’t go back.  Each subject is a one-line, high stress ticket.  It’s completely yes or no.  Succeed or fail.  No other option.  I can’t learn from my mistakes.  I can’t see the rest of the story I’ve missed.  I can’t interact.  All because each question is unattached to the next, and case outcome for each suspect is often completely determined by one answer.  Even sillier, I’ve several times tried to show damning evidence to a lying suspect, only to be told Wrong.  But on the next topic, the subject clearly lies to your face as if they just hadn’t seen the evidence.  And, of course, if you present the evidence a second time you get Right.

This one’s particularly frustrating to me as a designer because there is an obvious (if difficult) solution to pull from reality.  We don’t have one-shot conversations.  If I don’t get the answer I want, I ask again.  I’ll push as far as I can go.  This goes triple for police work, particularly in an interrogation room.  All in a game that’s deliberately exploring  the meaning of false confessions and imperfect evidence!  Plus negotiations frequently hinge and flow around the presentation of key facts like evidence – it’s a great way to get the same sort of “No going back” they are looking for, while not making the player feel screwed by the designer.

These systems in games are in a weird places.  On one hand, they are the gameplay.  They are the game.  The less game-y they are, the more you have a movie.  But on the other hand, they are trying to tell a story.  They are trying to be vague. LA Noire is about a time and a place and a set of people.  And I think that’s right.  The immersion, the context, the cutscene is more important then the system here.  The key is there are lots of different kinds of game systems on lots of different time scales.  Dialog interaction doesn’t have to use the same type of feedback as a Mario jump.

If I was going to make the next game in this genre, I would try 3 things first:

1.  Make the options clearer, and stick to them.  Pick a point of view, and make sure the script revolves around it.  My gut is “the player’s reading of the face” was a good shot, but it’s not consistent enough.  Go with controls for “the player’s belief about what the suspect is doing” – are they being helpful, holding back, or need a hard dose of evidence-reminding.  Describe the avatar’s goal, not the suspect’s goal.  Let the facial animation inform instead of drive.

2. Outcome is never final.  Succeed and fail on a larger scale.  What you learn is what determines your success.  How you get there, the individual steps, are not worth grading.

3. Conversations loop and flow.  Conversational failure isn’t a 1-step process.  People can be aggressive multiple times, loop the conversation back and forth, react to suspects getting more (or less) defensive.  Evidence can be shown multiple times – the first can be to shock, the later ones are to rub the suspect’s face in their inconsistencies.  Record the extra dialog here.  This is your whole game.  It’s worth it.

The best thing about LA Noire is it’s human-ness.  In LA Noire, I can pick up a pencil.  In LA Noire, I can read your face.  Chris Hecker would be proud.  This is the kind of game we should want to make.  To explore being human on a social level.  Talk about things like morality and psychology and life.  But that’s hard, and it’s not because we can’t do it technically.  But the systems still fall short.  It’s because reality is a bad game.  It’s hard to make clear, it’s hard to make repeatable, it’s hard to make structured but natural.  I’m looking forward to our next stab at it.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s