Alex This next talk is by Gwaredd Mountain, who is the technical director at Climax Studios. SILENT HILL, the latest iteration, does some pretty interesting stuff getting inside the players' minds, so continuing on the same theme. Take it away Gwaredd!
Overview of the Game
Gwaredd Whereas Baylor's talk was all about how players perceive the AI, this is kind of the other way round; it's about how the AI perceives the player. I'm sure you're familiar with SILENT HILL, this is a game we released, actually it fells really weird because even though it was out in March, to me it feels like we were doing it last year. SHATTERED MEMORIES is the seventh version of the SILENT HILL franchise and just out of interest how many of you have played it? A few... More than Havok which is always good. So for those of us who have played it, and maybe for those who haven't I should have a little movie if it works.
SHATTERED MEMORIES Movie
Psychology Behind the Franchise
Gwaredd I'm sure you're all familiar with the franchise. SILENT HILL's a very different game from other games in the horror genre, in that it's not about gore, blood and shooting things like Resident Evil and those kind of games. SILENT HILL is more about Psychology, the getting inside of players' heads, the eerie, creepy, uneasy atmosphere the town SILENT HILL represents.
Research & Development
This is something early on that the designers sat and threw some ideas around about. They came out one day and said 'we've got this really crazy idea', kind of one of the things that some of the designers were familiar with earlier on in the panel: "What if we can get inside the players head?" Okay, tell me more. "Well, what we want to do is to make a game which can read the player and then we can change the game to use their fears and phobias" which is hard as it turns out. It's quite hard for even a human being to do these things; we can't sit down with a player on a couch and talk about the mother for half an hour and try to use that, so it's quite an interesting challenge. We thought: "Okay, we don't know anything about Psychology, we're game programmers", so we did some research.
Firstly, we read lots of books and when we ran out of books we read blogs and magazines and pretty much anything we could get our hands on. Largely this made us confused. We had lots of questions, we didn't really know a lot about Psychology, we had lots of ideas but trying to make sense of this mass of material was quite hard. Eventually we went up to a local university and we said "Look, we're trying to make this game where we get inside the players' heads", and they laughed and said "Well okay, you're silly but we're going to help you out and point you in the right direction". That was great, we took the material and assembled it in some kind of useful way, revisiting the idea of Psychology and psychologists getting inside the players' heads and using their phobias against them and all that kind of stuff, and we built the game around this concept. We had a psychologist who thought it would be great if we had therapy sessions and swapped between therapy sessions and gameplay, so we took some of the team and we put them into therapy, which in hindsight maybe wasn't the best idea, in fact some of them are still there.
The Psychology Theory
After all the research we had done we had some questions, some ideas about what we could do and we thought "Okay, what we need now is a model of a human being we can plug into the Wii and off we go." It turns out that psychologists have many models of human beings, there wasn't one convenient one we could just plug into a computer and get to work. Some of the different models of human beings we had included a DISC Personality Model.
If you open up any book on HR or Management of People you come across this stuff as it's quite a simple thing to get a handle on. Along the horizontal line we have on the left 'Get it Right' and on the right 'Get it Done', then of course 'Task Focused' and 'People Focused' on the vertical line. You can map your personality onto one of those quadrants and work out who you are. Programmers typically tend to be in the blue in that they are 'Get it Right' and 'Task Focused', game programmers tend to be slightly more towards the red. It turns out there are more than four personality types.
I have called it Myers-Briggs, aside it was Carl Jung who came up with these dichotomies. These are four dimensions of personalities, which obviously gives 2^4 or 16 personality types. Myers-Briggs were actually a couple of psychologists that came up with a test for identifying these types. What else have we got? Neuro-linguistic programming (NLP). I like NLP mainly because it's got programming in the title and one of the guys who developed it was a mathematician. NLP is great by the way if you want to re-program your brain. You can program your brain just like you can program a computer, you can change your own personality. NLP gives you tools to do that, if that's something you want to do. In NLP they have the idea of Meta-programs, which is a similar thing to the Carl Jung dichotomies. They keep changing their mind, currently there are 14 Meta-programs, these are the first 6 that they think are the most useful in the workplace.
Psychology And Gameplay
So you get the idea that these different personality models work in this kind of way: you have some dimension of a personality, you track whatever dimensions you're interested in and you use that as your model. The question for SILENT HILL is "What are the personality dimensions we want to track?".
It turns out that psychologists have 5 that they consider the most important, so we thought "We'll have those". For those who can count, there's four. It turns out neuroticism is quite hard, so we'll just sweep it under the carpet and not bother with that one. We took the big 5 and also because we think we understand this now we'll add our own; the kind of moralities, the tastes which are related to the story of SILENT HILL. SILENT HILL is a story-driven game so these things tied into the theme of the game; the theme of the story; the narrative.
Building & Using the Profile
The problem is, now that we've got a personality model, how can we build this up? These are some of the things the guys were alluding to this morning. It's very difficult when the interface to the Wii is a controller to take what the player does with it and map it to their personality. Do we try and work out how they wiggle the stick, how they move the joystick, how they press the button? It's very difficult to do. We have one thing that sets us aside: a psychologist. We have the idea of Dr. K, the psychologist and the therapy sessions. We thought, "We've got therapy sessions, let's give the players real psychological tests." Which is what we did.
The first thing you do is fill out a questionnaire, the first thing in the game and the only one you'll see. It's based on the Myers-Briggs questionnaire and the weighting's the same for the questions, we added one question of our own. This gives us a first stab of what the player's personality might be. We also track everything else that they do in the game: where they go, what they look at, how long they spend in specific areas. The actions that the players do are the things that we record and the things that go into and influence our model of their profile, and as you can see, up in the top right is the current profile we think it is. A different action will affect those numbers.
The other thing we do is what's called self-fulfilling prophecies, which are our own psychological test we've constructed in the game. This is the second one in the game, you've found a note earlier telling you the key is under one of these jackets and it's up to the player to decide what jacket it's under and like other psychology tests it doesn't matter what answer it is, it matters what thought processes went into choosing the answer. The truth is whichever jacket you look under the key will be under that. It's the jacket that you choose first and whether you look under all the jackets and zip them back up that matters. We track all these numbers and metrics and put them into the profile.
Personalizing The Experience
Once you've got a profile, the next step is to do something with it. You've got to change the world, the experience the players have. The things we change aren't the big things, we've still got to get a game and hit our dates. We change a lot of details of the game, we change a lot of little things. We change the characters, under one profile you'll get the same story but you'll be told in different ways, the discourse will change, the characters will change. Although the constituent narrative elements of this scene are the same, with a different character you'll get a different experience.
It's also interesting to note, if you look at the background, this is sexy Cop and has a red thematic background and this plays into the themes of the story. There are various different engines based on your profile, some can be based on the idea of drug abuse, alcohol abuse, girls being abused by her father etc. The background and the theme of everything in this scene plays into whichever particular narrative ark you happen to be on based on your profile. We change the monsters, this is... I'm not quite sure what this is, but you're chased round by some kind of abstract geometric entity, compared to the ones that you saw in the trailer, and we also changed the details of the world, so here's one scene where you've got a billboard which has contents based on your profile and the telephone number that you have. If you come across any telephone number in the game you can phone it and the answer that you get is again based on the profile. Of course with a different profile you get different content. So this is it all put together working.
Accuracy Of The Profiling
That's kind of in a nutshell what's going on under the hood and how we changed the game. The question is, does it work? We set ourselves this lofty goal of being able to profile the player and getting inside their head and all the rest of it. The first question is, are the profiles accurate? Do we actually build up an accurate profile of the player based on these seemingly arbitrary events we throw in? At the end of the game in the credit sequence you actually get a snippet of your psychological report.
Is Random Good Enough?
Here this is obviously pieced together with bits of text pulled in depending on your profile, and it's difficult to say as I've played it through with my wife and kids so the profile would have been mixed with all of us. We've had reports online from people saying that it is actually quite accurate, certainly in the testing we did in development it was coming out as being reasonably accurate. There was a certain amount of randomness in there and this could either be the fact that it is accurate or it could well be the fact that people read what they want to into the text and think "That sounds roughly like me", like they do with horoscopes.
My gut feeling is, knowing how it works and the different systems involved, that actually there's no reason why it wouldn't be reasonably accurate, the reason it that obviously the different things we do in the game have different weights associated with them. The things that have the most weights are the things you do with the psychologist that are based on genuine psychological evidence and even if everything else is random and there is a question about how well those things transfer to the medium of computer games, there's still no reason why there wouldn't be a modicum of accuracy within that.
More Accurate Than You'd Like
There are other things we do for example there is a bit where we track where the self-fulfilling prophecies are. If you go to one place it will close off another place and we determine which place you go into. In the town if you go into the electrical store it will make some judgment about your profile whereas if you go into the clothes store it will make a different one. The problem there is that the players don't consciously make those decisions, they tend to come across that scene and they'll go in the first one that they see or the most obvious door.
Those self-fulfilling prophecies that affect the profile are random and based on the things to do with how the scene was set up much rather than anything else. But because of the weighting of the psychological tests that we do know are reasonably accurate there's no reason it shouldn't have a fair stab of guessing your personality correctly. The other question is: Is it scary? All this stuff that we did originally like getting inside the players' heads etc.? I have another little video here because I'm a lazy presenter which was posted up to YouTube, search for SILENT HILL Take 1 and Take 2 and this is cut together from both.
Okay, so there it clearly can be scary. The truth is the reason they found that scary is probably because they are 13 year old girls and secondly everything they found scary has nothing to do with their psychological profiling. That scene would be the same regardless of their profiling, the only difference would be the type of monsters that chase you, maybe they found abstract nothingness monster particularly scary I don't know. Largely it was reviewed as being not a particularly scary game, but it does mess with you psychologically, which is interesting. These comments come from Metacritic, they're not the reviews they're what people posted, I'd much rather listen to players than critics. "The much-hyped psychological profiling, subtle yet meaningful, truly makes you feel as if you are being watched". "The story is top notch and the psychological profiling lends itself to multiple replays", which is good. "The psychological elements are insane, it changes everything and you'll play through the game many times before you even see half of what changes."
Messing With Their Heads, Yes
It's interesting that the psychology makes the players feel like their being watched and messed with. The truth is it's not really, they do have a different experience when they play through with different psychological profiles, this is what games are about. They're about make believe, getting people to fantasize and regardless of what the technical aspects are behind the psychology the thing that really works is that we play up on this, it was all in the marketing. Right at the start of the game we had a psychological warning telling you this game profiles you and gets inside your head. We're building this suspense and the players play along with it and get involved with it, so even though they don't understand all the very limited things we can do with the computer behind the scenes, they believe in it and that's the experience they get.
I couldn't leave the critical reception without taking out a piece just on the psychological profiling from a very well-known English critic living in Australia, whom I'm sure a lot of you are familiar with.
Scaring Them, No
Why wasn't it scary? The truth is that those big ideas in psychology about people's personality are not really anything to do with what makes people scared. It all comes back to the amygdale part of the brain, the primal part of the brain, the primitive bits that make people scared are nothing to do with personality and everything to do with how we react to stimuli. What we should have done is Physiometric Testing which is where you hook little kids up to strange machines and try and scare them and work out what's scary based on their heart-rate, perspiration, pupil dilation and all those physiological things we could test.
We could have done this at design-time if we'd had access to that kind of facility. It would be very difficult to do at run-time, though on the Wii and Kinect we may have been able to read body language or detect stress is peoples voices using microphones, however it would be a crude and noisy signal and difficult to pull any useful metric from. But that's not really what we set out to do, we set out to create an experience, which we did.
Profiling in Other Games
So, would it be applicable to other games, would you be able to use it in your own game? That would be quite difficult because not all games have a psychologist that can sit there and deliver psychological tests. You could certainly create situations within a game and model what the player does and try to infer some kind of meaning from that, but the thing is that we built our game around this concept. Everything that went into the game was there to support the experience the player should have rather than for the technical aspect on its own. It would be interesting, I would certainly like to see more modeling of the player in more sophisticated ways and it's tending this way if you look at some of the things we used to do in old games, you pushed forward on the stick, the player moved forward, but now if you push forward on the stick and if it's Assassins Creed or Uncharted there's a whole world of interpretation about what the player means rather than what they actually do. To get better at that we need to try and better understand what the player is getting at and I think ultimately if we want to move into interactive narrative we need to be better on working on that human-computer interaction.
This is related to other things that we're going to talk about later on in the panel which is to do with metric-driven design and metric-driven experiences - I made these terms up, they're probably accurate - which are two things: one is done at design-time and one is done at run-time. I don't really want to talk too much about these kind of things because we'll talk about them later and other people have done it much better than we have.
Challenges in Production
Some of the problems we had in development: first was content production, one of the things we were criticized about was the length, about 8 hours from start to finish, and we had to create at least three times the amount of content to support that because each scene has at least 3 variations. We also needed to create custom-cut scenes that could swap in and out with different content based on the profile and also we tried to mitigate that by building up a layer system where we could have the same thing but swap out things here and there to give us more variation. Obviously testing was difficult, this was largely ad hoc, it wasn't very precise. Most people on a voluntary basis, honest, took a psychology test at the start of production and we used that as kind of a way to validate our assumptions moving forward. We also had various different tools that was could inspect and change data.
Here are some of the challenges we faced. In terms of metric-driven approaches obviously LEFT 4 DEAD is the most famous and certainly got a lot of press, and they had a similar idea and built up a metric of the emotional intensity which is based on how many people are dying around you, how much damage you are taking and they have various models for decaying this over time. They used this metric to change the population density. I won't talk too much about this because there are release papers on it which can do a better job of it than I can.
TOMB RAIDER is an interesting one for metric-driven design. They took this and they took a number of specific bespoke game metrics they were interested in: Causes of death, number of deaths, completion time, which can be applied to most games. Help and demand was specific for Tomb Raider Underworld, this was a hint system for their puzzles. They took those, had a load of people play the game, took the metrics and used a neural network to try and infer player types based on those kind of metrics. Personally I wouldn't have used neural network I would have used linear-discriminate analysis, but that's because I'm a mathematician. Based on that they worked out four different types of players: veterans, people who completed it fast, had low death rates and didn't use a help and demand feature; solvers who were great at the puzzles but not too great at the environment; pacifists who were great at advocating the world but weren't so hot at other stuff and runners who were people who ran through, died lots but got to the end quickly. You could use this, the paper that I read on this didn't go on to say whether they actually used this in terms of balancing the game, which you could to favor one type of player or another or improve the experience for all of them.
So that's it really, that's the psychological profiling in SILENT HILL for what it's worth. My key points are that this is a unique take, I don't know any other game that's tried this kind of thing so it's interesting in its own light and relates to the mindshare of what other games are doing. The other point has been made by other people, the fact that we did what worked for our game, we started with a game and we built the technology that supported our vision of the game rather than the other way round where we started with some technology and try to crowbar a game into it. That's it, thanks.
Question "I hope I didn't miss this in your talk, but it seems to me that the variations in the game are determined by the choices you make in the tests or how you respond to the Psychologist. I was wondering if you have thought about a way to give the player different experiences without having him playing actively in a different way? I don't know whether you see what I mean. For example, if I go in a game and the first time I play it I choose a main character and the second time I want a different story with a female character, so I have to choose a female character."
Gwaredd This is something we have to do because Cinematography is a big thing in films as they are shot in particular ways that they are not interactive. Games are interactive and the problem with interaction is that players tend to interact in ways that are not friendly to the game. We model things in terms of how we like scenes and how we present things to encourage the player. There's a whole language behind different games that encourage the players to do different things and we could certainly change that based on whatever metrics that we chose, whether we chose a Psychological metric or something else, then that's certainly something that's possible and would be interesting.
About Dishonest Players
Question "I've got a question, two questions actually. The first one being: Did players who were unable to be honest about something, were they typically not enjoying the game or did they just go through it and say 'this is cool technology'? The second one is: Did you learn anything odd about your co-workers out of all the Psychological testing?"
Gwaredd If you've worked with the co-workers I work with then: odd, yes. Surprised, no. You can certainly play through the game in any way that you chose and you'd get a different experience if you do different things, which is pretty much what we're aiming for. What players would infer from that, so say they could play the game in a way that wasn't themselves, they could say 'aha! I've got a game that doesn't understand me', then that's great for them. If they play through and think "well I've played through myself and I don't feel the game reflected that" then they'd be disappointed. So long as they believe in the illusion then they're going to have a good experience. If they deliberately want to play in a way that they don't go along as players then that's up to them.
The most recent game in the survival horror series, SILENT HILL: SHATTERED MEMORIES recently became famous by psychologically profiling its players. The game received acclaim from the press and players alike, in particular for getting into people's head — and adding replayability to the game. But what combination of tricks and techniques were used to pull this off?
In this presentation from the Paris Game AI Conference 2010, you'll hear Gwaredd Mountain (technical director at Climax) explain the player modelling used by the game to infer which of the Big Five personality profile is dominant in each player. He also explains what data drives the psychological profiling and how it affects the content in the game to deliver a customized experience to the player. He'll also share some insights on its integration within the game's design.
Gwaredd Mountain was Technical Director at Climax Studios during the development of games such as OVERLORD: DARK LEGEND and ROCKET NIGHT. Gwaredd previously worked at Argonaut Games, and is currently Lead Software Development Engineer at Microsoft Game Studios in London. You can read his blog here and follow him on Twitter as @gwareddm.
There are many different models of human personality, including DISC, the BIG 5, Myiers Brigs / Carl Jung, and neuro-linguistic programming.
The best you can do is decide which variables you want to model and track, and pick those — custom to the game itself.
SILENT HILL: SHATTERED MEMORIES starts with a questionnaire to build a psychological profile at the start of the game.
The game also tracks where the players go, what they look at, how long they spend in specific areas.
Different actions are annotated to have a different effect on the variables in the player model.
The game gives you mini-choices that change the content later on, in a self-fulfilling way.
As you play, the profile is used to serve different content: characters, monsters, objects, backgrounds and one of three narrative arcs.
Testing revealed the profiling to be quite accurate — given a certain noise.
Players also tended to identify with the psychological profile report, much like a horoscope.
The game "selling" the psychological profiling as a feature helped focus the players and reviewer's attention on it.
The psychological profiling doesn't make the game more scary; that's a much more basic reaction in the brain.
Here's a specialized format for you to download and play offline via a portable player:
Psychology Profiling in SILENT HILL: SHATTERED MEMORIES — Video Gwaredd Mountain Download MOV (QuickTime)
The MP3 file below is better quality than the streaming video above, and is a perfect candidate for listening to via a portable player (96 Kbps). The OGG file is the highest quality of all (128 KBps). You can download them both here:
The slides used during the presentation are available here: