Wednesday 12 September 2012

A Future of Polygonal Friends: AI & Emotional Attachment (Part 1)

This piece is split into three parts (Part 2 is here, Final Comments are here), and was written for the Critical Distance Blogs of the Round Table. 


Some people read Playboy for the ‘articles’. Some people watch films for the celebrity roles. I play games for the story writing and characters. It’s my dirty secret.

The quality of script writing in games can barely even scratch the quality of what would constitute a ‘good’ film or novel, but I have a major soft spot for when a game wants to spin me a fable, or endear me to characters. Because writing a decent script for a game is apparently so damn hard, the industry focus is elsewhere; AAA titles push heavily for graphical fidelity and larger worlds to explore.

I feel that the next big step for games is improving how they tell a story – and that’s not just a plea for the industry to wean me off my addiction to charmingly ineffectual narratives. There are so many ways to go about stepping up the storytelling in games, critics could write (or argue) about it for months, but here I’m going to focus on character-driven storytelling.

To support a great character-driven narrative, you'll need... well, great characters. But if the only properly fleshed-out character is the protagonist, a player will have a hard time getting emotionally invested. While a well-written character in a story or a film can feel believable just by just watching them, video games are an interactive medium. We can prod a character, and we watch how they react with scrutiny.

In those cases, traditional writing can only take you so far – it’s the Artificial Intelligence of a game character that breathes more life into them.

But before I talk about what we can do with AI, there are some key topics aspects to cover first, more to do with writing skill than technical – and will use some examples from games where AI isn't a factor. It will also bring up potential writing and design problems that we'll come back to in the second part. All in good time.

Variety is the spice: Character design of the cast

It's very easy to formulate a cast of characters for any video game setting. The number of RPG characters with almost predestined personalities to fit in with their profession; the hardened crew of soldiers in any FPS (don't forget the token black one!); the full rainbow of 'sexy' feminine personalities the hero can bump uglies with. When you’re trying to get the character design process out of the way so other aspects can we worked on (which is a terrible idea by the way, but it definitely does happen), falling back on a familiar pattern to construct a cast is a tempting idea.

The laziest pattern is what I like to call 'Super Sentai casting'. Did you ever watch Power Rangers as a kid? Or maybe you're an adult child and are still watching it – whichever. In every series they have a cast of around five, and their characters are the result of some kind of MadLibs-by-committee, something like this:

___ is one of the ___ youths with attitude, representing the colour ___ and the legendary ___. Their general passion is ___, and they're the ___ of the group.

Pretty ridiculous. What's more ridiculous is that video games, even ones that are meant to be heavily character-focused, still write their cast like this. Take Katawa Shoujo (lit. ‘Cripple Girls’):

___ is one of the students at a school for the disabled. She suffers from ___ but wants to ___ in spite of this. Her ___ reveals her 'true beauty', and you help her with ___ so you can eventually get up in dem guts.

What these MadLibs designs lead to is a homogeneous cast. Everyone being of the same demographic leads to pigeonholed personality traits to set them apart. (The Japanese Otaku culture has a huge number of words to describe these personality tropes).

Even otherwise good games can fall victim to it. Persona 4’s main cast is all very similar in age, background, and role in the story – though that’s partially down to thematic reasons, and they are instead humanised by putting their insecurities at the forefront. ‘Tough, but Kind-on-the-inside’ Kanji isn’t an especially complex character, but the struggles he has with his sexuality are very human and identifiable.

It means that in the end, you’re asking people to empathise with an idea; a cluster of memes and implicated cues, rather than an honest attempt at making something human. Chicken Nuggets compared to an actual chicken. This works out okay in some situations – and for some people – but some stories just cannot be told effectively with a hollow or homogeneous cast.
The very best game casts have characters from all walks of life, so not only are they unique physically, but their experiences and world views are also in contrast. It can even help improve games that are otherwise lacking.
For example, take Lux-Pain for the DS. A Visual Novel, it wasn’t incredibly well-received due to haphazard mechanics ideas that don’t go anywhere, the cheesy premise of a high school-aged secret agent investigating mysterious goings on, and an absolutely god awful English translation. And yet, its wide and varied cast gave it some serious charm. You encounter students, teachers, shop keepers, even television personalities – and underneath the… unfortunate dialogue, there’s some simple but passionate backstory and character motivation.
Shunichi Inagaki is just your average nightclub owner – he's not even connected to the mystery Lux-Pain revolves around. But he's still appears, going about his daily business. As you talk to him, you find out about his family, his relationships with other members of the cast – and if you take the time to read the dossier files that slowly fill in as you play, he's rather ham-fistedly implied to be gay. Appreciated, though not artful. As friendly and relaxed as he is, not even Shunichi manages to escape the bad translation.
The translation team was apparently a group of around 10 in India. It shows.

Bioware’s Dragon Age and Mass Effect series get huge fan followings for their cast alone. In both the Grey Wardens and your interstellar crew, you are partnered with casts so diverse, they aren’t even all human. On a level, that feels like a slightly cheap way of adding cast diversity, but it does give scope to including perspectives that can’t be (safely) attributed to human culture.
The lack of diversity in video game casts is very much a social problem, and game designers limit themselves – or try to diversify and fail – for a few reasons. Maybe they’re not confident in writing characters of a demographic different to themselves. Maybe they don’t have the time and resources to design complicated casts (in which case they should probably manage their budget better). More likely, they’re fearful of what might happen if they don’t pander to the less-than-liberal consumer base that aren’t interested in characters that don’t fulfil their power- or sexual-fantasies. Or the TV Tropes crowd who can only process media that’s heavily compartmentalised.
The quality and diversity of a cast is only going to change significantly when social attitudes change, and the ETA relies heavily on your optimism.
Words, words, words (and actions): Dialogue style and variation
A problem that affects all media is dialogue diversity – it’s not uncommon to have a work where all the characters speak in a manner very similar to that of the writer. The writer knows how they sound, and so by writing in their own ‘voice’ as it were, it’s easy to write in a way that’s (superficially) convincing. But when too many characters sound similar, the suspension of disbelief falls apart, and no one sounds convincing.
In practice, this only really affects games without voice acting – with the dialogue voiced, characters will naturally vary linguistically (unless all characters are voiced by the same actor, but that would just be silly). So in a way, technological progress in game design has solved this one. But there are obviously still instances where games can’t afford – or just don’t need – voice acting, so bearing this in mind is still important. 9 Hours, 9 Persons, 9 Doors has a wonderfully varied cast – and a very good story in general – but due to the game’s philosophical (and pseudoscientific) themes, all the characters know a whole lot about prosopagnosia or Ice-9 for reasons not always explained, and sound very similar when discussing it.
Who would think that this lady had just explained 5 minutes worth of 'morphic resonance'?

Furthermore, it’s not just how a character speaks, but how often. It’s difficult to love a character who doesn’t communicate with you very often (not that it stops fans from writing smutty fanfiction about them, but still). ‘Communication’ need not require dialogue (an obligatory mention to Ico and Journey here), but even then, a limited range of expressions can leave conversations one-sided, and characters feeling flat.
A game could have a wide range of NPCs, and have them be story relevant, but if they say precious little, or worse, are prone to repeating the same thing often, then it drags you out of the experience; the grand play ruined by a cast member repeatedly yelling about mud crabs.
Catherine handles dialogue incredibly well. It has a lot of dialogue with characters that are incidental to the overall plot, and they continually change what they have to say over the course of the game. Obviously they have a limit, but the game sidesteps a situation where the player unintentionally causes dialogue to repeat by marking when a character has said all the useful information they have to offer.
Limited or repetitive dialogue is eased in many modern games by having dialogue trees in major conversations (which has definite flaws; I’ll elaborate in the second part). At the very least it ensures that the major characters don’t just exist to spout the minimum of dialogue to push the plot along, but it would be a total failure of pacing to employ this with every single character in a game, along with taking way too much effort to implement.
Okay, that covers the ground work. In the second part, I’ll be talking about AI as a concept, how it relates to the above, and where we can go from here.

Read on to Part 2.

1 comment:

  1. Brilliant article! I'm surprised to see how things like procedural storytelling and NPC AI haven't improved more over the years, but I think developers are starting to notice that that is what we really want instead of just better graphics.

    I also think that every single game should use the system from The Sims, letting characters act according to their personalities, following their needs...

    Thanks for sharing your words.