-Gamification at is current state lacks methods for fast iteration and evaluation of game mechanics and game elements. Using UAREI modelling method to simulate gamified system and evaluate its effects users based on player types. In the experiment a simulated market of agents is used randomized motivation and behavior accounting. User behavior is expressed using mathematical formula and adjusted for case study.
I. INTRODUCTION
Gamification [
Gamification systems can be classified into these categories
as suggested by [
Behavior-changing gamification, aims to encourage people to make better choices thus increasing motivation.
Currently gamification domain lacks methods for modelling,
simulating and analyzing user behavior in gamified systems.
This paper offers a simulation and analysis method using
UAREI [
Several efforts exist at classifying and codifying recurring
gamification practices and common techniques such as (1)
Mechanics-Dynamics-Aesthetics (MDA) framework [
In game research, there is a strong separation between design
methodologies and usability evaluation tools, which are rarely
employed in the early stages of the design process. Although the
game developers use many often heuristically designed tools to
Copyright © 2017 held by the authors
assist the design, there is still very few existing methods
employed to connect design practices with gamification and
game design [
The design of serious games is a complex process. Two
opposing principles must be united: achievement of serious
objectives and meaningful gameplay. This can be achieved
using detailed technical modelling and implementation [
Based on Flow Theory [
For player type classification HEXAD player types [
They seek to create new social connections and relationships. Free spirits are motivated by autonomy and selfexpression. They like to explore. Achievers are motivated by mastery and overcoming game challenges. They continuously need to improve themselves. Philanthropists are driven by altruism helping others without any reward for themselves. Players are motivated by extrinsic rewards. They are playing the game only if they expect to be rewarded. Disruptors are motivated by changes. They are willing to ‘disrupt’ the game rather by plying by its rules.
To assess player types, we can employ the HEXAD
questionnaire [
II. CASE STUDY
OilTrader is a game developed to model the influence of reinforcement model on player decision to continue or leave the game. OilTrader is a market simulation game which allows to trade shares of oil to money or to buy oil shares. The game serves as an illustrating example how real world markets would behave if there were no external influences. The interface of the game is presented in Figure I.
FIGURE I.
OILTRADER GAME SCREEN
OilTrader is a simulator which allows for users to experience simplified market conditions while trading the digital shares of the fantasy company OilFund. It involves seeing historical game outcomes and trying to predict outcome of the next round. The game consists of rounds, each thereof takes 15 seconds. Each player starts with 500 shares and 500 dollars. In each round, a player decides to sell or buy the OilFund shares, or is not to place any trades in that round. Only a single trade can be done in a single round. The user sees four sections in the game. At the top, he sees his money and the OilFund shares. At the left column, the user sees trading controls and round timer. Below it, he sees trade history data and the impact on his money or shares the trade had.
The physical aspects of the game (Figure I) are comprised of tokens. Tokens are divided into the following three types: • •
Oil tokens: cylindrical markers representing the player’s ownership of oil.
Money tokens: sack-shaped markers representing the player’s ownership of money.
Each player starts by entering the game website. Next, he registers / logins to the game. From the beginning, the player needs to pick action for the current round. He can sustain, sell or buy oil. The player picks an action and enters how many oil shares he wants to sell or how much money he is willing to spend to buy oil shares. After his decision, he waits for the round to end. The trade is evaluated determining the seller to buyer ratio. Using this ratio, the player resources are redistributed based on the Minority Game logic. Finally, the player can decide to leave the game or continue to play the next round.
For simplification, assumption is that a player can only be affected by the elements of the game’s user interface which he/she can see. Experiment hypothesis is that it is possible to evaluate the influence of the reward mechanism (visually represented as a leaderboard table) on the duration of game playing depending on the different psychological types of player.
Users will be divided randomly into two groups: the main (experiment) group and the control group. The game’s user interface for the control group has the leaderboard which represents player achievement, and shows player position, net worth (shares + money) and win or lose state in the latest round of the game. The game’s user interface for the experiment group has additional three metrics (streak, biggest win, and biggest loss),
which represent player progress, and are aimed to incentivize the internal player reward (see Figure II). (a)
(b) model for formal specification of gamification, and the UAREI visual modeling language for graphical representation of game mechanics. The whole UAREI system can be used for full gamification development process.
The gamified systems can be described as a tuple: = { , , , , } (1) here: U – users, which are interacting with the system; A – actions, which trigger system behavior; R – rules, which encapsulate logic in the system; E – data entities; and I – interfaces which define data format.
The users are defined as a tuple = { , }, here: – a set of all outgoing links to other elements in the model; and – a selection function which defines how a user is selected from a collection in a simulation mode.
Actions are a collection
= { 1, 2, … , , … , }, here is a single action, the total number of actions. A single action is defined as = { , }, here: – a set of all outgoing links to other elements in the model, and – a selection function, which defines the way an action related data entity is selected from a collection. defined as:
Rules are a collection
= { 1, 2, … , , … , }, here is a single rule,
the total number of rules. A single rule is defined as = { , ( , )}, here: – a set of all outgoing links to other elements in the model, and ( , ) is a rule function ( , ) = { − −
(2) here: C – context of current execution path; M – a system model; y is a computed result value, and NULL is returned if rule doesn’t apply.
Rules are used to control context flow in the system. If a rule execution evaluates to an empty result the current execution path is continued. We can define the “else” path by using inversion “! ”. No data will be stored in storage and no other rules will execute if the previous rule failed or returned empty value, but system flow will continue giving feedback to the user node. Rules can update the context in anyway needed for the application. to other elements in the model.
Entity collection is a collection of all data entities in the system
= { 1, 2, … , , … , }, here is a single storage entity and k is the total number of storage entities. A single entity is defined as = { , , } , here: – entity scheme definition,
– data objects, and – a set of all outgoing links
Interface is a collection = { 1, 2, … , , … , }, here is a single interface and l is the total number of interfaces. A single interface is defined as = { , }, here: –a set of all outgoing links to other elements in the model, Q – data query, on which the data for the interface is selected.
In Table I we present the list of graphical symbols (graphemes) used in the UAREI model diagrams.
Minority game logic systems represents rational player decision making process based on game state picking the next action. To model minority game agents were introduced into UAREI model as part of user behavior. (m_networth, m_bloss). action was chosen by the user and saves amount which traded, in case of sustain – 0, else random value from zero to how much is currently owned by user. fields: Money, Oil, Networth, Win (did the user win last round), Streak (how many times in a row did a player win), BWin and BLoss (biggest win and loss), Round, name, and motivation seed = {{ }, }, here s
is defined by such m_position, m_win, m_streak, m_bwin, = {{ w }, } here sheme is defined: UserID, Round, Action and Amount.
= {{ , ℎ }, } - – computes which group won sellers or buyers are minority and computes buy to sell and sell to buy rations.
= {{ , L , }, has such fields: Round, Sell to buy, Buy to ℎ
}- ℎ sell and outcome. ℎ a
= {{ – selects all users and sorts by networth.
= {{ }, }, } - , } o - displays last 5 outcomes from user perspective.
III. EXPERIMENT
Hypothesis for this experiment is that simulated randomized agent behavior would be different for each psychological player type. This method allows comparing two gamified systems or games and evaluating their impact on user motivation by player types.
Simulation was done on two groups: experiment and control. Experiment group sees additional UI elements during game play. We will assume that users are only effected by elements which they can see. Simulation with the fallowing models where run based on such configuration: ( ) = −1( ) + ∑ , − −n −
(12) It is stated what there are 3 ( = 3) factors in control group and 6 (
= 6) factors in experiment group impacting how the user behavior will change. Three shared factors are networth, winning and position.
Additional factors introduced in experiment group are biggest win, biggest loss and streak. – each players factor is a random number between -1 and 1. , can be -1 if the impact of this factor is negative (losing money) experiment based on experiment group.
= ∑ −− = 6 −−44 3 = 2 (13) 75 75 0 0 75 75 , lizers -75 -75 -75 0 0 0 ,
= = simulation run until all players decide to stop playing. 0(
) is a random value between 30 and 60. The
Simulation is done using agent based simulation. Two simulations there run separately for control and experiment groups consisting of 1001 agents in each group. Control group has three motivational factors and experiment group has 6 motivational factors. Initial motivation 0 was randomly chosen. Also, for each player’s factor was randomly chosen.
In this model, there are six elements in gamified version which effect player behavior. Using logical reasoning for picking each weight, which will be used evaluating this system simulation. Three points are picked from -100 to 100 percent range, which are weight - -75, 0 ,75. The closer you are to 0 represents what a factor has almost no impact on user behavior. The closer user factor weight is to -75 be more negative outcome has the factor to user motivation. The closer you are to 75 the more positive impact has the factor to the user’s behavior. Table II shows how different factors should impact user motivation for different player types. Worth noting real experiments results should be used to justify the classification weights. motivation factors weights to the closest player category.
If we have player motivation factor weights as a vector ( √∑( ℎ = ( ℎ, ), ). , ℎ − ℎ, ,
, and
classification , ,
weight There 2 ) . , , , distances from each player type. Players type is the player type which is closest (min( )) to the player type in Table II.
= {
, } is a set of C. Simulation results 10 8 6 4 2 0
2 1,5 0,5 1 0 -0,5 -1
XP
Control
In Figure IV, we see the simulations results. In experiment, two groups participated – experiment group and control group. Looking at each group results by player type and with all of them together (Mixed type). Looking at averages without classification between player types there is no difference between control and experiment groups round counts. Looking at classified player types we see that there are clear differences for each player type behavior caused by the introduced changes to experiment group.
1,50 0,34 0,58
0,78 Change -0,58 -0,82 0,02
A method for simulating games from user motivation perspective was offered. The method built on top of UAREI modelling framework by introducing agent motivation. Results were analyzed based on suggested player type classification. It was found what motivation of gamified systems or games might be predictable if each system gamification element would have a known impact on each player type.
More research is needed on analyzing different psychological player types as the result of an experiment performed here show that it is difficult to clearly assign player types to real subject, as the qualities of different psychological types maybe mixed in the same person. Rather than defining crisp player types, a fuzzy-like approach to player typology is needed. This method allows us evaluate game patterns effects on players. These conclusions may spur the development of novel player classification taxonomies and motivation enhancing gamification simulation in the future.