The keynote discusses:
1) The role of AI in social games and different views on this topic.
2) How social actions can be modeled through operational logics in game design.
3) Examples of AI-based social game prototypes that incorporate elements like semi-autonomous avatars.
3) An experimental social game prototype called the Pataphysic Institute that uses an AI architecture called the Mind Module to model character personalities and emotions.
New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024
The role of ai in social games eladhari2011 uppsalauni
1. Keynote
The role of AI
in social games
Uppsala
August 2011
Mirjam Palosaari Eladhari
Gotland University, Sweden
2. Common views:
• ”Social games doesn’t have any AI”
• ”Social games doesn’t need any AI – they have people”
In this talk:
• What does ”AI” and ”social” mean anyway?
• Social actions in terms of operational logics
• AI based game design
• Research prototype(s)
• A recipe
3. AI
Artificial intelligence
the study and design of intelligent agent where
an intelligent agent is a system that perceives
its environment and takes actions that
maximize its chances of success.
(Russell & Norvig 2003 – a commonly used textbook)
the science and engineering of making intelligent
machines .
(McCarthy, 1956 – coining the term)
4. Social
The term Social refers to a characteristic of
living organisms (humans in particular, though
biologists also apply the term to populations
of other animals).
It always refers to the interaction of organisms
with other organisms and to their collective
co-existence, irrespective of whether they are
aware of it or not, and irrespective of whether
the interaction is voluntary or involuntary.
5. Games on social
networks
• We say “social games” – we mean “social
network games” (Siegel 2011). Usually that
means:
• Facebook
• Asynchronous play
(like old chess-by-
mail games)
• Metaphors for socializing
(like sending gifts)
6. Social game types
If Critera: Interactions between players
mediated by game rules
• LARP
• Table top & card games
• MMORPGS
• Competitive FPS
• Asynchronous games on social platforms….
• etc
Broad! Most games except single player games
would fit the critera, making it irrelevant.
7. Social actions
In games, many low level processes together
represent metaphors for actions.
We often to call games “social” if these processes are metaphors for
social actions. (ie, “greet” is social, while “shoot” isn’t seen as social
in and everyday context.)
Operational logics has been defined as follows (Wardrip-Fruin and
Mateas, 2009): “An operational logic defines an authoring
(representational) strategy, supported by abstract processes or
lower-level logics, for specifying the behaviors a system must
exhibit in order to be understood as representing a specified
domain to a specified audience.”
9. In this talk Social
games are
• (game) Systems where coexisting organisms
interact.
• The organisms are intended to appear life-
like (or believiable)
• The actions avialable to organisms in the
system are metaphors for social actions (in
the everyday sense of living beings socialising)
10. AI based Game design
AIGD is the practice of creating games where
the AI is an intrinsic part of the game design,
intertwined in the game mechanics.
When we talk about AI in games we approach it not only as a set of
techniques, such as BDI architectures or path planners, but from a
more philosophical perspective: the AI in AI based game design
refers to when computational processes are designed and
implemented to create the impression of something that acts
with intentionality. This ‘something’ can be an agent, such as a
non-player character, but it can also be a process that creates
content for the game, such as quests, or new levels in a game.
www.aigamedesign.com
11. Why AIGD
• In the space of AI based game design, established AI technologies
are used for innovative game design, and the needs emerging from
game designs push forward the innovation of AI technologies.
• A commonly held view regarding AI and games:
• it is enough to create the impression of intelligence, the “smoke
and mirrors”, by choosing the right cues that makes players use
their mental models of a representation, relying on players’
immersion and imagination.
• This is not a reason to not go beyond smoke and mirrors!
• We can do real magic! Especially if we use the intelligence of the
players. (again, remember Else’s talk on A-Life yesterday)
13. Semi Autonomy
Semi-autonomous avatars are agents whose
actions are controlled partly by users and
partly by artificial intelligence (AI) components.
14. Approaches
to semi-autnomous avatars
• Relief
• Players can be relieved of cognitive and operational load by for
example automating the animations of body-language of avatars.
• Expression
• Means of expression through body-language, types of actions
performed, and reaction tendencies can express the nature of
specific avatars to other players in the same world.
• Impression
• Character-information available only to avatars' own players and
personalised, subjective world-representations create individual
impressions of worlds and avatars' parts in them.
More info here: Semi-Autonomous Avatars in Games, Eladhari, M.P, 10th International Conference on Intelligent Virtual
Agents, Philadelphia, Pennsylvania, USA, September 2010.
15. Goals
A shared aim of these approaches is to
increase the believability of elements in the
game worlds and the sense of presence and
immersion for players.
17. Game examples
• PromWeek: Social physics
• Pataphysic Institute: mental physics
• Sims Social: avatar simulations
18. Prom Week
• Prom Week: simulates characters in a
network of relationships. God-perspective
for player. Goals: play with the characters’
social actions in order to reach
relationship goals.
• AI: CiF
architecture
Comme il Faut 2: A fully realized model for socially-oriented gameplay, by Josh McCoy, Mike Treanor, Ben Samuel,
Brandon Tearse, Michael Mateas, and Noah Wardrip-Fruin. In Proceedings of the FDG Intelligent Narrative
Technologies 3 Workshop (FDG-INT3 2010), Monterey, CA. June 2010.
19.
20. Pataphysic Institute
• PI: simulates characters in networks of
relationships. 3rd person Role playing
persoective for player: Goals attune own
others mood in order to defeat common
enemies.
• AI:
The Mind Module
21.
22. Sims social
Taking the multiagent
simulation model and
applying a main-
character- perspective
on it.
Using real-world
relationships – the
character the player
controls interact with
characters of facebook
friends. (timely
illustration of the 3rd
era of gaming as
described by Garriot
2011 GDCE)
23. Social Perspectives
• PromWeek: Player manipulates many other
characters (multi-agent simulation) from a top-
down perspective.
• PI: Player control own avatar while interacting
with other players’ characters, mobs and
NPCs.
• Sims Social: Player controls a single character
while it interacts with representations of
facebook friends’ characters. (asyncronous
play, hybrid of multi/single player)
25. Goal: support Role-
Play in VGWs
• Cater for complex, interesting avatars
• support players in expressing consistent,
interesting player characters in VGWs?
• body-language etc.
• Support by rule system in order to make the
actions performed characterising for particular
avatars.
Mirjam P Eladhari, Michael Mateas University of Californa Santa Cruz
26. Characterising Action
Potential (CAP)
CAP is what characters can do at a given moment that
characterise them, both in terms of observable behaviour and
in expression of true character - a character's essential nature,
expressed by the choices a character makes.
The observable characteristics include visual appearance, what
body language characters use, what sounds they make, what
they say, and most important, what they do and how they
behave.
27. Experimental set-up
• Agent architecture: The Mind Module
• Prototype Game World: The Pataphysic
Institute
• Play-tests and interviews which are video
taped and transcribed verbatim. Allows
assessment of:
• Interaction
• Players’ attitudes
• Players’ mental models of system.
More on the method in Design for Results: Considerations for
Experimental Prototyping and Play Testing Using Iterative Game
Design, by Eladhari M. P., Ollila, E.M.I. Game Research Methods,
University of Tampere, Finland, 8-9 April 2010.
29. The Mind Module
• Spreading Activation Network of Affect nodes
• Semiautonomous agent architecture.
• Semiautonomous agents are partly controlled by
their players, and partly controlled by context-
sensitive action potential and expression possibilities,
as well as by varying degrees of autonomous
reactions to in-game situations specific to the VGW
the agents inhabits.
• Gives avatars virtual minds in addition to virtual bodies.
36. Pataphysic Institute
emotional combat system!
Personality trait based characters!
Player-created bosses!
Manifestations of own Traumas and epiphanies!
37. Pataphysic Institute
• Prototype virtual game world where the
personalities of the inhabitants are the base for
the game mechanics.
• When interacting with other characters, the action
potential depends upon the player character's
current mood and personality
38. production
PI is built in the company Pixeltamer's framework for web based
multiplayer games and is played in a web browser through a Java
applet.
PI is an application developed for conducting experimental game
design research using iterative design and guided play tests. The
MM is used as a library.
• Programming of Client and Server by Christoph Pech.
• IPIP NEO - scoring system and report routines provided by John A.
Johnson, rewritten to C++ by Christoph Pech
• Graphics: 3DS Max models by Ola Persson
• Leveldesign by Musse Dolk
• Scripting by Johan Sköld
39. Basic Game Play
• Players need to defeat physical manifestations of negative
mental states by using
• Spells
• Affective actions
• The spells available to characters depends on their
personality and current mood.
• The affective actions available to a character depend on
the current mood.
41. ’pataphysics
Jarry (the early 1900s, Paris)
"the science of imaginary solutions, which
symbolically attributes the properties of
objects, described by their virtuality, to their
lineaments“
Raymond Queneau has described 'pataphysics as
resting "on the truth of contradictions and
exceptions."
42. ’pataphysics
‘Pataphysics is an inner posture, a discipline, a
science and an art; it allows every human to live
her life as an exception, which doesn't prove any
other law than its own.
Above my favourite definition. My translation from a book in Swedish by Claes Hylinger
43. Back-story
• Players are introduced to the back story of PI before they
log on, by reading the diary of Katherine, an investigator
who was sent in to PI to investigate the consequences of a
mysterious event called the Outbreak. In PI, reality has
been replaced by the inhabitants interpretation of
reality, and their mental states are manifested physically in
the environment.
• The head of human resources at PI has taken upon
himself the task of understanding the new and unknown
world by applying personality theories. He forces everyone
in PI to take personality tests, and studies what types of
abilities these persons get: Mind Magic Spells.
• Teresa focuses on the finding that social interactions
between people suddenly in acutely concrete emotional
reactions. She calls these Affective Actions (AAs), and
tries to understand her changed environment by studying
the patterns of these.
48. Node Weights
Karl: what I can do and can’t do is tied to my mood! The mood seems to be connected to my personality and my emotions. I
think its like that for all of us. But we seem to be able to gain more resistance and energy as we learn to act in this world.
Personality Trait -> Emotion Emotion -> Mood
Personality traits affect how strongly characters feel emotions. Emotions affect characters’ mood.
49. Mind energy and
resistance
Karl: what I can do and can’t do is tied to my mood! The mood seems to be connected to my personality and my emotions. I
think its like that for all of us. But we seem to be able to gain more resistance and energy as we learn to act in this world.
52. Mood Area restrictions
for affective actions
Teresa: Affective actions are quite remarkable. They force an emotional reaction!
If you are in a receptive mood that is.
Players can perform affective actions towards
other characters in order to change their mental
state in both positive and negative ways. By
affecting others mood's the selection of their
available spells and AAs is changed.
Example:
The AA Comfort can be used successfully on
targets that have an active emotion node of
Sadness, but only if the player's own avatar is not
in the area of Furious on the mood co-ordinate
system
54. Manifestations
Single Sentiment Manifestations (SSMs) Compound Manifestations (CMs)
- origin in a single emotion
- unique, that is, there is only one
- limited amount of mental resistance and - CMs are stronger than SSMs in terms
energy. of larger maximum amounts of MR
and ME
- The spells an SSM can cast increase the
value of 'their' emotion in targeted entity. - more versatile in their behaviour, they
can cast both spells and AAs.
Instantiation: scripted or created - In order to vanquish a CM generally
automatically when avatars experience several players need to co-operate.
strong emotions. For example, if a PC 'feels'
a Joy intensely the SSM Joy Jumbo is Instantiation: CMs are scripted or
instantiated in proximity to the PC. authored by players.
56. Curses and Blessings
• Avatars can be affected by the spells
Sentiment Curse and Sentiment Blessing.
• Sentiment Curse gives an avatar a strong
negative sentiment that has a zero decay rate.
• Example: curse of Guilt.
The way to get rid of this sentiment is to
create a manifestation of the sentiment, a
compound manifestation (CM). If the CM is
vanquished, the sentiment disappears.
63. Play testing
Film
Camera
Player 1
1
(Object of
Game
Desire)
Master
Player 2 Player 3
In Clients:
-Log files of play (one file for
each avatar) Film
-Camtasia recordings of
Camera
:-voice (microphone in
headset),
-Face (webcam),
- Actions (taping of screen)
64. Play testing
Player 1
(Object of
Game
Desire)
Master
Player 2 Player 3
65. Plat test scale and
scope
• Multi-player tests – 3 players + 1 game master
• 3 scenarios:
1. use AA’s according to roles in a group
2. learn to use spells towards manifestations
3. Player’s authoring of manifestation.
Cooperate in group to overcome them.
66. Scenario 1-AAs
Scenario: Two avatars competing for the affections of the third
GM asked players to picture a situation where two of them (Player 1
and 2) had played together for a while, and that the third (Player 1)
was new to them,
67. 2. Learning spells Defeating
manifestations
Scenario 2: GM takes group to area with negative single
sentiment manifestations.
Goal: cooperate to overcome them, and learn own special
personality spells.
68. 3 player authored foes
GM ask player to curse each other. Cursed player manifests the emotion
they cursed with. Players’ goals: deafeat the foe using grouped indivuals’
spells. Use AAs to affect each others moods needed to defeat foe.
70. uses of Affective
actions in scenario 3:
1. as authorial building blocks when creating
their compound manifestations,
2. to manipulate the mood of each other’s
avatars when they attempted to neutralize the
manifestations, and
3. to diminish values of the emotion nodes of
the opposing manifestations.
71. Vertical slice test
outcomes
• In the test: Using vertical slice in order to narrow down
• Successful designs
• Results of general
interest
• Design-wise for prototype: Players’ creation of
manifestations, and the strategies used to defeat them: a
combination of rule-set, and ”real” past experiences.
• General regarding avatar & agent behaviors: levels of
autonomy & focus on secondary behavior (read more in
digra’11 paper)
Game Mechanics and Dynamics of Social Actions in a Prototype Multiplayer Game World, by Eladhari M.P, Proceedings
of DiGRA 2011 Conference: Think Design Play, Utrecht School of the Arts, the Netherlands, 2011.
73. AI + ’social’ games =
= New AI tech!
Design challenges in social games can drive
new develelompent in AI technicues – new
problems, new solutions
= New types of games!
Games with metaphores for social actions –is
in the domain of beleviable agents and
expessive AI – new game experiences built
using knowledge about social behaviors.
www.aigamedesign.com
74. Why it works
• Leverage the power of relationships in the
“third era” (Garriott 2011) – both between
individuals, and formations of groups.
• Notion of what is ”real” (Bartle 2003). Player-
created ’real’ content in frameworks of fiction:
compelling. (If the balance suits players’
golden mean for real-connection vs escapism.)
75. Final words.
I suspect that:
Player created/generated “realness”
+
mental/social physics
+
rule-systems giving appropriate affordances for action
+
support for avatar expression, impression & relief
=
meaningful fun.
76. Reading
• Comme il Faut 2: A fully realized model for socially-oriented gameplay, by Josh
McCoy, Mike Treanor, Ben Samuel, Brandon Tearse, Michael Mateas, and
Noah Wardrip-Fruin. In Proceedings of the FDG Intelligent Narrative
Technologies 3 Workshop (FDG-INT3 2010), Monterey, CA. June 2010.
• Game Mechanics and Dynamics of Social Actions in a Prototype Multiplayer Game
World, by Eladhari M.P, Proceedings of DiGRA 2011 Conference: Think
Design Play, Utrecht School of the Arts, the Netherlands, 2011.
• Design for Results: Considerations for Experimental Prototyping and Play Testing
Using Iterative Game Design, by Eladhari M. P., Ollila, E.M.I. Game Research
Methods, University of Tampere, Finland, 8-9 April 2010. The paper was
selected for publication in a special issue of the journal Simulation &
Gaming
78. Recent Play tests
Play-tests of the PI prototype focussing on
- Authored relationships within the group (triangular tensions
given by sentiments)
- Players’ authoring of CMs – their contributions to the game
world
- patterns of play when the group deals with the CM’s they
created. Role-takings and strategies.
79. FUTuRE TESTING:
- players’ attitudes towards CAP in PI for
role-play
- Sentiments for story construction (more)
- Semi-autonomy – balance of player-
control versus autonomous behaviours of
avatars
- Autonomous agents (NPCs) with MMs,
use of reactive planning for behaviour
selection
80. Current Work
Analysis of play test data
• Video’s of game play in the client, along
with player’s face and voice
• Surveys
• Log files
• counts of Affective Actions towards group.memebers,
depending on sentiments between avatars
• Counts of Affective Actions and Spells between entities
when conquering the CMs players authored
81. Future Directions
Prototyping and play-testing methods for game
design research.
Utilisation of bio-feedback from players, both for
design and for play-testing.
85. Stance
How affordances and action potential of avatars
in multiplayer settings can affect narrative
potential.
86. Outline
• Semi Autonomy
• Relief, impression, expression
• PI and MM
• Agent Architecture: The Mind Module
• Prototype: The Pataphysic Institute
• CAP of semi-autonomous avatars and
narrative potential.
87. Impression
Character-information available only to avatars'
own players shows the action potential of the
avatars - what they can do at a given moment
and how. The properties of avatars can be
used to represent the game world subjectively,
helping players to identify with their
representations and immerse themselves into
story worlds.
Ex: in Michael’s Subjective Avatars users receive
text descriptions of environments which react
avatars emotional states.