Appropriateness of Adaptation

 

Appropriateness of Adaptation

There are lots and lots of subtopics to think about here!

For me, one of the biggest questions is the following: Is this dependent on game genre, game pace, or style of game play?

I hope that over the coming posts you’ll understand why I pose this question.

The world of computer games is both big and diverse, which suggests that adapting to user emotions may not be appropriate for all games, or indeed in all situations.

Let’s consider single player vs. multiplayer. In a single player situation the environment around the player can adapt to that one user. However, in a multiplayer game, where the players are on opposite teams the environment would have to adapt to two contradicting states. Is this feasible, and if so, how would you do it?

 In single player mode the AI of potential monsters can be adapted to the player's emotions, whereas in a multiplayer versus style game, the game cannot adapt one of the player's capabilities. Even attempting to steer the progress of the game through changing luck levels for the players is not advised. This could benefit one player, but the other would end up feeling cheated. This is an ethical issue, one of many. I will attempt to discuss ethics in later posts.

Moving on from single vs. multiplayer there is the question of style of game. I say style of game, because at the moment I'm not sure whether it is genre dependent or dependent on pace or whether the game play is about solving a puzzle, or killing some monster/other player's avatar say. Casual games, such as the game I've made, I think can greatly benefit from emotion detection and adaptation provided it's fast pace.

Consider a casual game such as Frozzd http://www.yoyogames.com/games/show/20523. In this game you "jump from planet to planet as you guide the Mubbly creatures, and use them as an army to defeat the Frozzd." This is an amazing casual game by the way! Definitely worth a try!
The game is not extremely fast pace; you can take your time running and jumping between planets. Having said that, if you encounter the Frozzd you need to be ready, and then things start happening fairly quickly. I think adding emotional adaptations to this game would benefit it as it would differ each time you played it, and in addition it would give you a sense of achievement on each level. More experienced players could then be made to feel challenged from the get go, and noobs could be allowed to finish the game (not necessarily without dying a few times, we’re not talking make it dead easy here..)

On the other hand, consider Mah Jong. I suppose this is also a casual game, but it is a totally different style. Mah Jong is more like a slow card game, matching pairs to make them disappear.



Playing Mah Jong you may well experience different emotions, however because the game play is slow, and it's a game based on thought and careful consideration, the physiological changes are likely to be so slow and small that they could just as well be because of something totally unrelated to the game (e.g. someone comes in to the room and tells you of some good news). This suggests a limitation of the technology and will be discussed in a later post. From a design point of view though, it's hard to see how a game like Mah Jong may benefit from adaptations, or indeed what you may adapt. The blocks are all laid out at the start of the game. As two and two vanish it depends on your own choices whether you will make it or not. In fact, if the game was adaptive and changed certain blocks (it would have to be hidden blocks, or the player would see it changing, which could lead to frustration, or at least boredom if they could see the 'cheating' taking place), the system could potentially change the outcome of the game in a negative fashion by messing up a winning strategy.

This leads me to think it’s not related to the genre of the game so much as pace and/or style of game play. What do you think?

5 Comments
6 Comments
Posted by Phewsie

 

Appropriateness of Adaptation

There are lots and lots of subtopics to think about here!

For me, one of the biggest questions is the following: Is this dependent on game genre, game pace, or style of game play?

I hope that over the coming posts you’ll understand why I pose this question.

The world of computer games is both big and diverse, which suggests that adapting to user emotions may not be appropriate for all games, or indeed in all situations.

Let’s consider single player vs. multiplayer. In a single player situation the environment around the player can adapt to that one user. However, in a multiplayer game, where the players are on opposite teams the environment would have to adapt to two contradicting states. Is this feasible, and if so, how would you do it?

 In single player mode the AI of potential monsters can be adapted to the player's emotions, whereas in a multiplayer versus style game, the game cannot adapt one of the player's capabilities. Even attempting to steer the progress of the game through changing luck levels for the players is not advised. This could benefit one player, but the other would end up feeling cheated. This is an ethical issue, one of many. I will attempt to discuss ethics in later posts.

Moving on from single vs. multiplayer there is the question of style of game. I say style of game, because at the moment I'm not sure whether it is genre dependent or dependent on pace or whether the game play is about solving a puzzle, or killing some monster/other player's avatar say. Casual games, such as the game I've made, I think can greatly benefit from emotion detection and adaptation provided it's fast pace.

Consider a casual game such as Frozzd http://www.yoyogames.com/games/show/20523. In this game you "jump from planet to planet as you guide the Mubbly creatures, and use them as an army to defeat the Frozzd." This is an amazing casual game by the way! Definitely worth a try!
The game is not extremely fast pace; you can take your time running and jumping between planets. Having said that, if you encounter the Frozzd you need to be ready, and then things start happening fairly quickly. I think adding emotional adaptations to this game would benefit it as it would differ each time you played it, and in addition it would give you a sense of achievement on each level. More experienced players could then be made to feel challenged from the get go, and noobs could be allowed to finish the game (not necessarily without dying a few times, we’re not talking make it dead easy here..)

On the other hand, consider Mah Jong. I suppose this is also a casual game, but it is a totally different style. Mah Jong is more like a slow card game, matching pairs to make them disappear.



Playing Mah Jong you may well experience different emotions, however because the game play is slow, and it's a game based on thought and careful consideration, the physiological changes are likely to be so slow and small that they could just as well be because of something totally unrelated to the game (e.g. someone comes in to the room and tells you of some good news). This suggests a limitation of the technology and will be discussed in a later post. From a design point of view though, it's hard to see how a game like Mah Jong may benefit from adaptations, or indeed what you may adapt. The blocks are all laid out at the start of the game. As two and two vanish it depends on your own choices whether you will make it or not. In fact, if the game was adaptive and changed certain blocks (it would have to be hidden blocks, or the player would see it changing, which could lead to frustration, or at least boredom if they could see the 'cheating' taking place), the system could potentially change the outcome of the game in a negative fashion by messing up a winning strategy.

This leads me to think it’s not related to the genre of the game so much as pace and/or style of game play. What do you think?

Posted by ahoodedfigure

When I think about the single player vs. the multi-player experience, I often liken human players, especially those who aren't present, as often being a surrogate AI.  Especially in games where there is minimal interaction between players, this is strongly the case.  As player interaction allowance increases, this decreases the "AI" usage of a player, and increases potential distractions.  As most players are capable of de-facto Turing tests as part of our normal routine, we're eventually able to tell the difference between a computer contributor and a genuine one, especially as the level of interactivity and the diversity of potential modes of interactivity increase.
 
So it's more HOW a multi-player game operates that determines the differences, rather than whether or not there is a multi-player mode or not.  You then have the single player version of this, with the player being the sole influence on the environment (time seems to stop for the player, allowing her or him to do pretty much what they want at their own pace), or if there are other actors in the environment which affect things independent of player wishes, or if these same actors are partly or wholly reactive to player behavior, but still seem to have a will of their own, and thus complexify the relationship of player to game.
 
If you limit interaction in a multi-player environment, and you make clear the goals, you decrease this disruption of the feedback element for the game, while if you increase the interaction in a single player environment, you may wind up creating unforseen physical states in the player as a result of possible incongruous or human-like behavior on the part of the AI.
 
Addressing your end question directly, I wonder if the potential for different games and how different people relate to them is too broad to be able to nail down a specific idea on why this is so.  
 
One "frustration" idea came to mind; perhaps you've heard of it.  There's a Tetris clone which actually picks tetrads that are the WORST possible tetrad for a given situation, based upon an algorithm.  When you need a long piece to make a Tetris, it doesn't come up, possible EVER. The goal of the game changes into trying to fool the algorithm into giving you at least adequate tetrads, instead of expecting the eventual correct one to come down.  The converse of this would be to have a Tetris that reacted to a player's poor performance by providing helpful block formations.  This could still be sabotaged, though, by a player stacking everything up in the middle and not even bothering to play correctly.
 
The game has no direct understanding of what the player is doing, but the player's behavior is still manifested on screen in a way that he machine reacts to it.
 
Pace, then, is absolutely important in understanding, especially in rudimentary stages, what a player is actually reacting to, as a game that moves too quickly may cause multiple states which run over each other and collide, creating confusing readings.  A slower-paced game will be easier to understand WHAT the player is reacting to.  An accurate understanding of player state may also hinge on repetition, where the game tests to see if what they're reacting to is something onscreen, or if it's just a coincidental event outside of the sensors' purview (such as the good news).   Should a faster-paced game be used, it may benefit from simplicity, where the stimuli are limited to a few separate incidents which the player reacts to, and thus can be repeated and won't confuse the readings.
 
Then again, a player's reaction to the first of a type of incident is not necessarily going to be the same as the second.  The player could be delighted at the overwhelming force that crushes their tiny force, because the odds are so ridiculous, but repeated exposure to the same sorts of odds may lead to frustration.  It helps to repeat incidents to continually gather calibration data over time, maybe, to sort of check the hypothesis of the starting calibration, if that makes sense.  The pacing of these repeated incidents may have to be spaced out over several play sessions, rather than hammering it into the player again and again, to avoid another reaction, that of frustration and disgust, and possibly irritation that the game simplistically hammers in the same event over and over.
 
It should also be noted, and perhaps you'll get into this in the ethical part, that a game that reacts to a player may be seen as invasive.  A player suddenly doesn't have the freedom, within the context of the game, of having whatever emotion she or he wants to have.  The player must be mindful of their emotional state, as readable through physical sensors, lest they affect the game in an undesirable way.  For a home version, then, pause would be absolutely necessary, as I would hate to be punished by the game for not having the expected emotional state.  I realize that's outside the scope of your study, but it's something to consider.

Posted by Phewsie

Wow, this is an awesome reply! Thanks so much for this, lots here that can go into my thesis ;-) 
 
I think as a game designer you need to focus on how the game is meant to be played. That's not to say that they shouldn't consider how people may play it differently, and account for it, but in the case of Tetris, well if someone can't be bothered moving the blocks to achieve best results, they shouldn't really be playing it. And neither normal games nor adaptive games can really do anything to combat this sort of behaviour. This is not to say that extensive testing should be done to avoid the player destroying the game (getting stuck bugs, glitches in collision detection etc.)! I'm just saying that if the player only wants to wreck havoc then an adaptive game should not be expected to perform any better than a normal game. An adaptive game will attempt to understand the emotional changes in the player, but if he is refusing to play there isn't much the game can do about it. 
 
Regarding the pace issue I believe that most fast pace games are either fairly simple in design, such as tetris or street fighter, or they have waves of fast pace, shooters for instance. Either way, the sensors detect physiological changes as a minimum of 32Hz (32 measurements pr. second), and we know that depending on the sensor it takes up to 4 seconds for the change to take place, and generally it dissipates again after about 6 seconds. Depending on the design on the emotional classification system then, it should not be a problem to determine what triggered the emotional change, should this be required.  Not all games need necessarily care what triggered the emotional change. 
 
Though my dream is for these systems to be user independent (which mine is, simple though it may be), I appreciate that allowing the game to learn as it goes along will most definitely benefit the classifier, allowing for more accurate detections, and thus adaptations. However, this requires that each person who ever plays the game has they're own profile, one can never share (like we're currently doing on DJ Hero, playing one song each.. :P). Whether to use machine learning as part of the classifier algorithm really would depend on the individual game, what the designers are trying to achieve etc.  
 
Your ethical observation is astute, it is invasive. However, from my own experiments I find that very few people said they would not consider using such a system if it was affordable, due to invasion of privacy. This is of course a choice, and I'm sure there are people out there who would choose to not play such a game, but I think the majority wouldn't worry about it. Although I agree that it is invasive, I do not agree with your reasoning for it. I suppose it depends on the system, but certainly my own system is not 'monitoring' you emotional states in any ominous way. It simply looks for a change, up or down in arousal, attempts to understand it within the context of the game, and then adapts to it. It does not require a specific emotional state at any one time, it does not expect any states, or changes. The player is free to feel whatever they want to feel, and act however they want to act. Additionally once the change has been determined and adapted to it is not stored in the system, it does not keep tract of your 'emotional journey'. Having said that, should the player wish to, they could attempt to control they bodily functions so as to control the game using physiology as input. The way I look at this though is that if that's what they want to do, that's kinda cool. They choose to use their body, consciously, as an input device to the game. Funky! :)
Posted by ahriman22

Oh god it's an essay thread.

Edited by vidiot

Yeah, it's 2:30 in the morning....I'll tackle this.
 
From a design standpoint everything can be broken down to a simple "push, pull" concept. Give the player moments of progress, versus moments challenge. This can be of course associated from either a narrative or game design point of view, but for the focus of you argument, design point of view. This very basic theory covers everything to difficulty and even general pace to your game.
 
ahoodedfigure brings up the concept that adaptation to difficulty can run into legitimate bouts of frustration, which is understandable. Does having a game of tetris you can't win translate into something fun, of course not, there is no visible threshold of a pull the impression of giving the player the ability to progress. As long as you have that threshold, the safer. 
So when ever that awesome space algorithm from the future that calculates and adapt how your multiplayer session is going, independent from pre-desired paths from a designer everything will be gravy. 
 
Understanding emotions psychologically from a player perspective, is a bit more difficult and is more up to the player in question than to the designer. What we can do is try our hardest to guess, or impose what we think the player might be feeling. Even with a cool funky future space algorithm, everything will still be greatly subjective to the individual user.
 
Regardless, you guys covered almost everything that I could think of. 

Posted by Phewsie
@vidiot:
Indeed, everything will be greatly subjective. The question is how good can we get at guessing/deducing how the player is feeling..? I'm guessing with time we can probably get pretty good, certainly through use of machine learning algorithms, and accurate emotion classifiers. But I guess it'll take some time..  
I think using the drama curve used by films is a good place to start when attempting to induce 'fun' which you refer to. Continuous stress and nothing but is probably not fun for most people, nor is slow and easy. The trick is to find a good balance, allowing tension to build, build, build and then chill after a climax. Then start the process all over again. Perhaps a good idea would be to attempt to construct the ideal drama curve as part of the machine learning, rendering the system user-dependent, but potentially very clever at not only guessing what you are feeling, but ensuring the optimal level of enjoyment, challenge, pace and engagement for you..? 
 
BTW, I really like your push, pull theory. I've never really thought of games in that way, but you're definitely right! :)