Here’s another challenge. Your testers are going to be heavily influenced by their expectations. For example, I can suspend my disbelief for a cartoon as long as I didn’t expect live action. Setting up the test is going to be quite challenging as you make sure you don’t bias the participants right off the bat.

I think consideration of this issue will help address your questions as well.

http://slate.msn.com/id/2102086

So it sounds like an ambitious problem to tackle, but will make a very good thesis when it’s done.

To (1), I would say that they *can* be related, but are not necessarily so. Within a certain genre of game that aims to create a world like our own, graphic realism can play a strong role in that experience. But within video games there are many games which create alternate worlds with no such expectations.

An extreme example would be a game where you just play with moving colors — like an interactive satori screensaver. This could be quite fun (and extremely immersive), but there is nothing real about the world. In fact there are not even representation objects or characters to interpret as real or fake. Regardless of its lack of realness, one thing that can be said about this type of game is that it appeals to our visual aesthetic sensibilities.

To deal with (1), you could constrain your work to particular gaming genres, but you’ll note that in many cases things which appeal to our aestethic sensibilities will trump realness.

For example, in an FPS where the character enters a burning building, visibility *should* be near zero (i.e. the screen should be nearly solid grey or black). The character’s health realistically should start dropping from toxic inhalation, etc. But taking a less realistic approach by making visibility higher and filling the space with amazing flame visuals would probably be more immersive. In such cases realism works against immersion because realism would actual point to a reduction in sensory perception and situational awareness.

Moving on, (2) is something of a wierd way to measure expectations vs. outcomes. It doesn’t strike me as valid because losing SoD could be due to so many non-graphic factors. For example, something doesn’t sound right, or the level-design is ridiculous, or you simply have a non-game thought about taking out the garbage. While there may be some relationship between SoD and immersion, SoD seems like it would be too fraught with noise to get a good signal to noise ratio.

Whoops. I got a meeting.

The part I’m really testing is: does graphic realism generate a level of expectation in behavioural realism? I argue the question is important because I think if there’s an expected behaviour that’s not met, SoD is broken. I don’t claim that meeting these expectations is sufficient for SoD either though.

So the problem is, how do I measure whether something is matching the expectation level? In your example, the graphics are unrealistic in that they don’t portray anything near the real world. Just like Tetris. Hence, there ARE no expectations to be met and anything you do with them is accepted. Try a Tetris game with people in contorted forms (actually I’ve seen this) – it looks bizarre.

a) Disappointed when a realistic game has goofy game physics, OR
b) Extremely impressed that you are able to an amazing move which clearly defies sensibility

I think in the thesis as you’ve stated, you’ve mostly been trying to talk about (a), but (b) can jar SoD just as hard, but would actually *improve* immersion.

The issue is that expectation is not a fixed variable. As I move through the game and learn more about the gamespace, my expectations also change. Perhaps for levels 1-3 you have very basic abilities which all fit neatly within the realm of realism, but in level 4 you get spells. Those new abilities change your expectations and quite possibly in a good way (i.e. increased immersion).

In a sense what I’m saying that SoD is neither necessary or sufficient for immersion. Clearly the class of game that I described in my first comment is free from expectation and requires no SoD — still it can be immersive. And in the class of games where SoD does help (games that use a world based ostensibly on representational realism), I see the breaking of SoD as sometimes being positive and othertimes negative.

Perhaps the questions I should look to answer are better stated as:

i) given a set of graphics, are there a set of initial expectations on how these graphics will behave

ii) if so, when/how does it affect the user (e.g., positively or negatively?)

The other piece you are investigating is whether a break in expectations effects immersion. I’m sure that every such break has *some* sort of effect, but as per (ii), the effect could be positive or negative.

Now as for testability, it is still going to be hard to come up with a good system for comparing expectation to experience. You could run a study like this:

1) Show the user several still screenshots of a game.
2) Give them a survey which asks questions as what they would expect to be possible
3) Give users a play script (e.g. complete the first level, or even just: explore the second floor of this building)
4) Have some users play the normal game
5) Have some users play the game where physics or skinning have been tweaked in surprising ways or places.
6) Videotape responses
7) Have an exit survey which asks about their impressions, and also to describe how they felt about surprises.

I would posit that if there are enough surprises, and those surprises have self-consistent behavior, that the user expectations may change mid-game and immersion can be achieved using that newly updated worldview.

Perhaps for example, flash bombs in Halo bounce off walls in an over elastic way. The first time this happens, I’m might say: “woah! that’s wierd”. But as I play the game I might find it to be a really useful way to surprise people around corners, or bounce the bombs into spaces where I know enemies are hiding. Once the behavior that was originally surprising becomes integrated into my updated expectations, it can become part of my arsenal of gameplay techniques and once again improve immersion.

The only negative I can see is if the break has no useful purpose to extend gameplay in the long run. For example, tetris with contorted people doesn’t play better. There is nothing that I can do with that surprise, and it just seems odd. But if having the pieces be people meant that I could change their position to get a different piece (but perhaps person A can only become B, B only C, etc), then it might become part of an immersive gameplay.

Video taping is something I’d toyed with but until now, it didn’t occur to me to tape the user’s expressions. At CHI, a presentation talked about how you could discern a lot from the player’s expressions and that may be a good way to go.

Exit surveys were initially planned for each room where I had them do a task but I think doing a full level may be more appropriate.