Pixel-Bitching: L.A. Noire and the Art of Conversation

It didn’t take me long to realise that something wasn’t right.

As a devotee of noir fiction and a long-time admirer of both James Ellroy’s LA Quartet and Curtis Hanson’s L.A. Confidential (1997), I was more than looking forward to Team Bondi’s attempt to recreate 1950s Los Angeles using the Grand Theft Auto sandbox template. However, as soon as Ken Cosgrove was shoved into an interview room with a suspect and told to extract a confession, I knew that something was desperately wrong – not just with L.A. Noire, but with video games as a whole. After decades of investment in realistic graphics and physics engines, modern video games can perfectly recreate what it is like to shoot someone in the face… but ask them to recreate a believable conversation between two humans and they are at a complete loss. What we need is a revolution in the way that games portray social interaction.

A Case Study

L.A. Noire follows the career of tortured war veteran and social cripple Cole Phelps as he works his way up through the LAPD, going from rookie flatfoot to celebrated detective and player in the political games that rage out of sight in the offices of LA’s rich and powerful. Right from the start, the game paints Phelps as a man who is difficult to like, because he simply does not understand other people. In a series of flashbacks to Phelps’ time in basic training we see him gradually alienating the people around him as he wages petty vendettas against those who are more popular and capable than him. Phelps’ fundamental disconnection from his fellow humans is made abundantly clear when you are first confronted with one of the game’s three basic mechanics.

L.A. Noire is an attempt to create a video game take on the police procedural genre. The police procedural differs from traditional crime novels, in so far as the emphasis is placed not so much upon who committed a crime or on foiling the baddies, but more upon the day-to-day realities of what it is like to be a working copper. In order to recreate the complexities of this existence, L.A. Noire fractures the gameplay into three distinct modes:

  • The first is the traditional runnin’, gunning’ and drivin’ template that makes up the Grand Theft Auto franchise.
  • The second is a crime-scene analysis sub-game faintly reminiscent of that of Quantic Dream’s Heavy Rain (2010).
  • The third is an interview sub-game where you use the clues collected in the second sub-game to attempt to squeeze information out of a series of witnesses and suspects.

While the second and third sub-games are quite effectively linked, L.A. Noire also links the interviews back to the GTA aspects of the game by making it possible for you to spend ‘insight points’ that you gain from working action-based cases in order to make the interview sub-game easier. One of the more bizarre subtexts of this relationship between sub-games is that the more people you shoot in the face, the better you get at knowing when people are lying to you. In effect, Phelps’ capacity to be inhuman to his fellow man helps him to understand his fellow humans better… thereby raising the possibility that Phelps is in fact a sort of autistic Colonel Kurtz whose willingness to commit acts of terrible violence is a form of spiritual strength. The road to Nirvana is easy to walk when you are wearing jack-boots.

Colonel Kurtz

L.A. Noire features an interview system heavily inspired by that of Capcom’s Phoenix Wright: Ace Attorney series. When interrogating a suspect you can assume that they are telling the truth, express scepticism or call them a liar; when you accuse them of being a liar, you need to be able to point to the clue that allows you to make that claim.

Phoenix Wright: Ace AttorneyThis system proves problematic for a number of reasons:

Firstly, you only really get one choice of response so if you express scepticism when a suspect is telling the truth you then lose the chance to gain that piece of information forever. While this sort of mechanic might have made sense in Ace Attorney’s courtroom setting, it is notable that Ace Attorney does allow you multiple bites at the apple whereas L.A. Noire does not. You may only be sitting in someone’s front room asking them questions but you will never EVER get the chance to row back from expressing mild scepticism about one of their responses or by suggesting they might be lying whereas in fact a clue only suggests but does not prove the lie. There is no ‘What I meant was…” or “I’m really sorry, I didn’t mean to offend you but…” button. This inability to row backwards from any social faux pas you happen to make is amplified by the interface itself.

Secondly, while the game may present you with three options, there is really no telling how Phelps will interpret your instructions. Confronted by a grieving widow who might be glossing over the facts because she can’t bear to face them, a good detective might gently probe a particular fact or ask them to rephrase their answer and then ask for clarification, but there is no telling beforehand whether Phelps will probe with sensitivity or whether he will accuse the grieving widow of having murdered her husband by way of some misguided plan to ‘shock’ her into telling the truth. Team Bondi have invested a lot of money in L.A. Noire’s facial graphics, but while it is easy enough to read the suspects, it is difficult to read your own character and how he will respond to your instructions. But it’s not just moods that the game struggles to interpret.

Thirdly, the game’s interface requires that you substantiate any accusations you make by clicking on the clue that proves that the suspect has just lied to you. The problem is that while many of the lies are lies by implication (a shoe does not in and of itself prove that someone was lying, but a shoe that is the same size and model as a footprint found at a crime scene does suggest that a suspect might be lying when they claimed not to have been to a scene), the game stores the clues in a very cut and dried manner, meaning that you often find yourself wanting to confront a verbal statement with a clue that implies that the suspect is lying. However, because the game does not track implications, I frequently found myself knowing that a suspect is lying but without having clear evidence. Ace Attorney also suffers from this problem, but gets around it by allowing you to put numerous possibilities to the defendant. By denying you the possibility to take multiple bites at the apple, L.A. Noire forces you to commit to an accusation without allowing you to tease out the implication in your own way.

Screenshot from L.A. Noire

Taken together, these three quirks of L.A. Noire’s interrogation system makes for a stilted parody of real human interaction, a mini-game that does not so much recreate human interaction in digital form as crudely parody it in a way that makes you doubt your own capacity for human speech and understanding. I mean… can anyone actually make any sense of what I am writing right now, or are you just going to accuse me of being a lying whore who murdered my husband in order to flee the state with a well-dressed fancy man? L.A. Noire is to the noble art of conversation as Space Invaders was to flying the space shuttle.

How Games Do It

In fairness, one of the reasons why L.A. Noire’s interrogation system is so horrifying is that its attempt at emulating real human conversation is so ambitious that it gets sufficiently close to the real thing that it falls into a sort of social “uncanny valley”, a territory that games have thus far managed to evade due to their reliance upon quite stylised and artificial renderings of human interactions. These approaches fall into three broad categories:

The first, which I shall refer to as Pixel-Bitching, has its roots in old text-based adventure games but features most prominently in the series of point-and-click adventure games published in the late 1980s and early 1990s by Sierra On-Line (which included the Police Quest franchise, the only other examples of video game police procedurals). Under this approach, characters waited for players to input the right combination of verb and keyword before allowing the game to progress. In games such as Leisure Suit Larry, you could babble away to characters and they would simply not react until you hit upon the right combination of words. Some particularly sophisticated games would only allow you so many bites at the apple or would include combinations of words that would alienate the character but mostly this sort of human interaction assumed a similar attitude to that of most point-and-click adventures, namely expecting players to hit upon the right combination of symbols to advance the game.

The second approach, which I shall refer to as Abstraction, features quite prominently in dating sims as well as in high-level strategy games such as Europa Universalis 3 and Civilization. Under this approach, the actual cut-and-thrust of human relationships is represented by a score that measures the extent to which the other character or nation likes you. If you give them money or buy them stuff or do stuff for them then the score goes up but if you attack them, insult them or lie to them, the score goes down. More sophisticated versions of the game allowed for the fact that being generally well disposed towards you does not necessarily preclude the fact that the character will sometimes act against you. After all, it’s nice that you give them money but they have their own interests and their own needs.

The third approach, which I shall refer to as Mechanical, allows for the fact that the player only ever has access to an abstract representation of the social situation that he is in. You may not be able to read the person you are talking to or win them round with your dazzling personality, but your character (who is supposed to actually be there in the room) might. Over the years, this approach has become the standard system for social interaction in video game RPGs, and can be found in games such as Baldur’s Gate, Knights of the Old Republic and Dragon Age. The more sophisticated variants of this approach allowed you to unlock certain conversational options on the basis of things your character had previously done or seen, thereby both personalising the conversational experience and creating an impression that what you do and say really has an impact upon how you interact with other people.

Screenshot from Knights of the Old Republic

Most games function by combining elements from all three approaches. For example, Dragon Age leans most heavily upon the third approach, but the options available to you in conversation are frequently impacted by how much the character likes you… a factor governed by a system influenced by Abstraction in the form of gift-giving. Similarly, strategy games such as Europa Universalis rely upon you making other nations like you, but the existence of core territories that different nations value in particular mean that it is possible to Pixel-Bitch your way into an easy rivalry. Even L.A. Noire itself relies upon a hybrid system, in so far as it uses a combination of Pixel-Bitching and Mechanical approaches, but its spin on the Mechanical approach is so innovative that it is doomed to failure almost from the start.

Aside from Pixel-Bitching in an intensely irritating manner, L.A. Noire’s interrogation system relies heavily upon the use of social mechanics. However, rather than these social mechanics residing in the person of Phelps in the form of skill-ratings, the mechanics reside instead in the players’ capacity to intuit people’s inner states on the basis of their facial expressions. In other words, it is not Phelps’ social skills that determine whether or not we successfully ‘read’ the suspect, but our own. This is undeniably a giant leap forward in video game social interaction, and the boldness of this decision stands as a testament to Team Bondi’s fully justifiable faith in their capacity to recreate complex and yet recognisable human emotions on the faces of their characters.

However, while the technology is without a doubt incredible (and one has to salute Team Bondi’s desire to break with tradition), L.A. Noire is absolutely crippled by virtue of the fact that we are left with only half a game. Indeed, many of the game’s frustrations flow directly from the fact that we, as players, can look at a character’s face and know that they are lying, only to find ourselves held hostage by a shaky interface and a wildly unpredictable and potentially autistic protagonist whose capacity to read human emotions and tailor his response to meet a situation is nowhere near as developed as our own. Every time Phelps bullies a witness or threatens a suspect who has nothing to lose, a part of me dies – because I know that I could have handled that situation so much better had I been the one in the room. If I’m the one who is being asked to read the situation, why am I not the one who is putting that ‘reading’ to work? However, while L.A. Noire’s hybrid system is so bad that one could happily call it a catastrophic failure, it does go some way to pointing out how the next generation of games should approach the emulation of human social interaction.

How Games Should Do It

The first beat ‘em up I ever played was the arcade version of Kung-Fu Master, in which you played a character that moved from one side of the screen to the other, punching and kicking (and occasionally ducking and jumping over) enemies as they came towards you. This meant that the game reduced the process of fighting to the question of timing: can you hit an enemy in the time between the moment when he comes into range and the moment when he hits and kills you?

Over time, the beat ‘em up genre evolved along the twin trajectories of the one-to-one confrontation of International Karate and the scrolling brawlfests of Double Dragon. As environments, enemies and character moves became more complex, games came more and more to resemble real physical confrontations in which speed, strength, balance, positioning, endurance and skill all play vital roles in determining who manages to win the fight. Indeed, pick up one of the more recently released fighting games from the Fight Night or Virtua Fighter franchises and you will discover games that require players to master multiple dimensions of play based upon the constellations of skills that affect the outcome of real physical confrontations. Fights are not resolved by a single punch; they flow, they have rhythm, they have a life of their own and their own set of rules. This is how social interaction should be handled in video games. This is how Cole Phelps should have taken on his suspects.

Screenshot from Fight Night

The problem with this noble endeavour is that, while the processes involved in hitting someone are fairly well understood, the study of communication is still very much in its infancy. For example, in a landmark paper entitled “Communication Theory as a Field” (1999), Robert T. Craig identified seven different approaches to understanding basic human interaction:

  1. Rhetorical: Treats communication as the practical art of discourse.
  2. Semiotic: Treats communication as mediated through signs.
  3. Phenomenological: Treats communication as the experience of dialogue.
  4. Cybernetic: Treats communication as the flow of information.
  5. Socio-psychological: Treats communication as the interaction between people.
  6. Socio-cultural: Treats communication as the means of producing a particular social order.
  7. Critical: Treats communication as the process of challenging assumptions.

Each of these broad traditions has its own vision of the nature of human social interaction and a different perspective upon the rules governing it, thus implying that there would be at least seven approaches to a properly developed communication engine. However, as L.A. Noire suggests, many of the complexities of this sort of modelling can be resolved by using improved graphics to effectively mimic social interaction. In other words, characters do not need to be able to communicate as effectively as real people; they simply need to make their utterances appear sufficiently human to enable us to suspend our disbelief. L.A. Noire’s facial modelling software is sophisticated enough to coax us into suspending our disbelief long enough to ‘read’ a situation… the next step is to allow us to respond to the characters without forcing us to do so through the medium of a character as socially maladjusted as Cole Phelps.


Jonathan McCalmontJonathan McCalmont is a freelance critic living in the UK.  His work has appeared in such places as Strange Horizons, The Escapist, Salon Futura, Gestalt Mash, The New York Review of Science Fiction and Vector.  He also maintains a blog named Ruthless Culture where he writes about films, literature and the fundamental hostility of the universe.

2 thoughts on “Pixel-Bitching: L.A. Noire and the Art of Conversation”

  1. After decades of investment in realistic graphics and physics engines, modern video games can perfectly recreate what it is like to shoot someone in the face… but ask them to recreate a believable conversation between two humans and they are at a complete loss.

    That may be true on the high end of the consumer scale, but there are genres of game that handle conversation better than L.A. Noire. In particular, the more or less underground community of the interactive fiction genre have been working for some time on what you might call the “technology of talk.” For a particularly interesting (and, by now, canonical) example, consider Emily Short’s Galatea.

Comments are closed.