Holistic player modeling for controlling adaptation in video games

HOLISTIC PLAYER MODELING FOR CONTROLING ADAPTATION

IN VIDEO GAMESBoyan Bontchev

Prof. at Dep. of Software Technologies,

Fac. of Mathematics and Informatics, Sofia Univ., Bulgaria

Marie Curie Fellow at Brainstorm Multimedia, Spain

Agenda

Introduction

Related works

The ADAPTIMES player model

Adaptation process

The “Rush for Gold” emotionally adaptive game

Case study

Preliminary results

Conclusions

2

9/04/2016Adaptation in Applied Video Games e-Society’2016

Player-centric adaptation

Player-centric adaptation in video games needs to answer some questions:

Who – player character structural and behavioral changes => requires player modelling

How – how adaptation will be realized => requires adaptive loop design

What – which game features can be adapted => requires adaptive gameplay design

Why – advantages of game adaptation => requires analysis of outcomes in playability and learning results


3

Player modelling

Player modelling is crucial for realization of an effective player-centric adaptation in both applied and entertainment video games. It implies:

Observation of individual player’s behavior “from a contextually omniscient view”

Construction and updates of a model of that player based on observation, and

Intelligent game adaptation for tailoring the gameplay to the individual player.


4

Source: Magerko, B. Adaptation in Digital Games, COMPUTER, July 2008, pp.87-89.

Player model types

Behavioral player models - describe dynamic processes of player behavior and deal mainly with flow, motivation and immersion

Organizational player models describe properties of player’s character, their attributes and facets and, as well, interconnections among them with their time and model space constrains


5

Behavioral player models 1/2

Mihaly Csikszentmihalyi (1990) developed the fundamental concept of flow as a process of optimal human experience. In digital games - a balance between the inherent challenge of the game activity and the player’s ability required for its execution.


6

Player skill

Ch

allen

ge

Boredom

Anxiety

Behavioral player models 2/2

Based on the flow theory, Gilleade and Dix (2004) stressed the importance of three issues:

motivation of the users: why they want to play

experience and skills: how able are they to play

detection: how to identify when change is necessary

Sweetser and Wyeth (2005) presented GameFlow model for assessing player enjoyment incl.: concentration, challenge, skills, control, clear goals, feedback, immersion, and social interaction.

The dynamic relation between the effort and the demand level was described as motivational intensity (Fairclough and Gilleade, 2012) going through boredom (low effort because of low demand), engagement (rising effort due to increasing demand), zone (peak of achievable effort at the highest level of demand), and overload – low effort due to excessive demand.


7

Organisational player models

Bridging the gap between game design and pedagogy (Kiili, 2006):

flow antecedents - challenge–skill balance, action–awareness merging, goals of an activity, unambiguous feedback, and sense of control

flow experience

flow consequences - increased learning, positive attitude, exploratory behavior, and perceived behavioral control

Game adaptation determine the level of playability (Sánchez et al, 2009): satisfaction, learnability, effectiveness, immersion, motivation, emotion, and socialization.

Besides playability, game adaptation may be used for increasing player competence in both playing and game-based learning context. User competence model (Silva and Behar, 2015):

psycho-cognitive abilities

attitudes

knowledge


8

Playing style models


9

Relationships between Honey and Mumford learning styles and playing styles of ADOPTA (Bontchev, 2011) and Bartle

Honey and

Mumford

Learns by: ADOPTA Bartle

Activist hand–eye coordination, planning andstrategizing, problem-solving, teamworkand the ability to think quickly

Competitor Killer

Theorist logically entering problems step-by-step,

with spatial awareness and verbal &

numeracy skills

Logician Achiever

Pragmatist planning, decision-making, testing

hypotheses, strategic thinking,

management skills

Strategist Explorer

Reflector Observing and watching reflectively Dreamer Socializer

The ADAPTIMES FP7 project ADAPTIve player-centric serious video gaMES

ADAPTIMES aims at investigating how

cognitive abilities (the cognitive part),

psycho-emotional status (the affective part) and

playing style (the conative part – that of “desire, volition, and striving”) (Schur, 1990)

can be used in a holistic approach for efficient and effective player-centric adaptation


10

Adaptation in


11

Workflow of emotion-based game adaptation control


12

Implicit Emotional

Indicators(via sensors & other devices)

Facial expr.GSR

Heart RhythmECGEEG

Pressure…

ChangeProcedures

TimeVelocity

AccelerationRouteSize

Defeat range Luminosity

Contrast…

Real-TimeExplicitPlayer

Feedbackthrough

AdaptationControl Panel

Analysis

Change

Visuali-zation

Triggering Events

on Behavior Templates

Case study – Rush for Gold

A 3D video game using adaptation based on player performance, efficiency and emotional state

Developed by using Brainstorm eStudio (http://www.brainstorm.es/products/estudio/)

Allows implicit recognition of ADOPTA playing styles in order to validate them to ones calculated by using self-report. The implicitly found style will be used for automatic selection of learning content appropriate to that style.

Goal: to collect 12 bars of gold (bullions), which are flying, hidden or inside logic puzzles needed to be solved, all located in a 3D Egypt temple with enhanced audio-visual effects being object of adaptation, as well.

The player can use a Strategy Management Table (SMT):

represents a table with three rows for planning numbers of bars of the three groups available in the game

contains data about average effectivity of performance for collecting the gold bars of each group useful for building a strategy for optimal way of play.


13

Rush for Gold


14

Player style measurement

Implicitly during play time of specific mini-game

Based on player’s metrics for specific game tasks:

Performance (result)

Efficiency (result/effort)

Task difficulty

Play time


15

Rush For Gold, part 2 -https://www.youtube.com/watch?v=aJe61bUDE40

Emotion-based adaptation in Rush for Gold

The game adaptation control is at run time and implicit for the player

The adaptation control makes the video game aligned to specific response patterns of individual players for bringing positive effect on playability and learning outcomes

The adaptation process runs in the context of the game and aligns adaptable game features to player-centric metrics showing:

player’s progress and performance

player’s emotions – emotions are inferred by face expression analysis using a Web service accessing Face Analysis Cloud Engine of SightCorp (http://sightcorp.com/) fear, surprise, sadness and disgust are applied for additional correction of

current difficulty of shooting and discovering tasks by using specific thresholds of their levels

when finding relatively high fear, sadness and surprise, task difficulty is decreased, while the opposite occurs in cases of high level of disgust


16

Adaptable game features in Rush for Gold 1/3

1. Adaptive automation of game tasks (and feedback for next experiments):

explicit tasks - game objectives (to shoot/discover gold bullions or to solve puzzles) adapt to the gameplay;

implicit tasks - not explicitly stated by the game interface but expected to be fulfilled:

maximize your skills in shooting, discovering, solving and strategic planning of activities

collect as many gold bullions as possible

player-driven tasks - created by the player thanks to his/her creativity within existing limitations of given game mechanics and leading to so called emergent gameplay

how to find best position for shooting (depends on gun high and shooting angle)

how to explore the temple with changing visual properties


17

Adaptable game features in Rush for Gold 2/3

2. Dynamic Difficulty Adjustment (DDA):

DDA by means of automatic level generation - uses methods for procedural content generation

Logic puzzles have increasing dynamically their difficulty with higher player results and vice versa

DDA by means of adjusting level content, i.e. game items for player interactions – means dynamic adaptation of level of inventory interacted by the player for specific game context, according to player’s skill acquisition:

Hidden billions are more difficult to be found with player progress

Flying bullions change velocity & acceleration plus striking force change with player efficiency as well with player´s emotions


18

Adaptable game features 3/3

3. Adaptation of audio-visual effects:

according player´s emotions and using thresholds for happiness level, there are adapted :

ambient illumination change

focus and contrast of statues with puzzle panels


19

Experiment

Executed at Sofia University, Bulgaria, with 34 volunteers (average aged 27, SD=10; 18 men and 16 women)

They passed group explanation and demonstration (20 min.), procedure of informed consent (translated in Bulgarian, 15 min.), individual assisted trial (5 min.) and, finally, unassisted game session to determine their style of play (15-20 min.).

Half of the volunteers (N=17) played the game with emotional adaptation switched on (forming the experimental group), unlike the others (the control group).


20

Preliminary results 1/3

The quiz reveal (five-level Likert scale from 1 to 5):

certain preference to play the game with emotion-based adaptation (question Q1 average=3.9, sd=1.1)

positive appreciation of that adaptation of shooting difficulty (Q2 average=4.0, sd=0.8)

the same for discovering difficulty (Q3 average=3.8, sd=0.7),

and brightness and contrast (Q4 average=4.1, sd=0.7).

No statistically significant improvements were found in playing time and shooting performance with emotional adaptation (because all players fulfilled the goal)

Statistically significant correlations between answers to questions Q1-Q4 and some game metrics


21


Positive statistically significant correlations (0,34÷0,55) between preferences to play with emotional adaptation and shooting/discovering efficiency, discovering performance and game session time reveal the beneficial impact of emotional adaptation inside the experimental group.

In contrary, significant negative correlations (-0,38÷-0,51) between degree of liking emotional adaptation of the difficulty and shooting/discovering efficiency and difficulty can be explained by appreciating emotional adaptation for making the game harder when found higher disgust (decreasing player efficiency and performance) or easier when found higher fear, sadness or surprise (decreasing attained difficulty).


22



23

Metrics of solving are checked only for correlations with adapted brightness and contrast, as they are not emotionally adapted.

Adapted brightness and contrast are correlated:

positively to solving performance (because of attracting player’s attention to the puzzles)

and

negatively to discovering performance (because of attracting player’s attention to other objects)

24

Rush For Gold (playing style)

9/04/2016

Generated labyrinth (maze game)

3D quiz game (assessment)

3D puzzle

Adaptation in Applied Video Games e-Society’2016

Next steps

Conclusions 1/2

Player-centric adaptive game play possesses essential advantages compared to the non-adaptive gameplay

The synergy of using player´s performance, emotions, and playing styles is very promising

All they are identified in an implicit way at run time and used for adapting game mechanics, dynamics and audio-visual content .

25


Conclusions 2/2

The results reveal a promising and positive appreciation of emotion-based adaptation and, as well, interesting statistically significant correlations between such adjusting of dynamic difficulty of tasks and player´s self-report.

More experiments with real time player´s feedback on adaptation use, additional self-reporting and qualitative analyses are needed for revealing how such emotion-based adaptations affect game playability and antecedents, experience and consequences of flow (Kiili, 2006).

26


Thank you for your attention!

Discussion

More info at: http://adaptimes.eu/

27


Engineering

Holistic player modeling for controlling adaptation in video games