User:OrenBochman/WGT/Methodology

Motivation

The original motivations was to investigate factors which can increase cooperative behavior in a wiki. However it soon became clear that questions are of general interest could be answered by perusing this investigation. These include:

• how to increase user engagement?
• how to increase article quality?
• what (policy) causes a wiki to close/open up as an information society?
• how to make smaller wikis more successful?
• can one evaluate the impact of a proposed policy in a lab settings?
• can one model/simulate/predict/define experiments to test impact of policy on editor loss/gender gap?

This work may be of interest to mainstream CSCW researchers since it will also consider behavior of non cooperative nature:

• conflict
• combat - wars of attrition
• competition
• control
• coercion
• vandalism
• elites

Prior Art

Game theory has been used previously in a couple of papers discussion Wikipedia.

The Method

• Wiki Game Theory, is the use of mathematical modeling of behaviour patterns existing within the community of a peer produced

CSCW such as Wikipedia.

This type of modeling can reflect four levels of reality:

1. reality - against which other models should be tested.
2. evolutionary models - models in which more variables are allowed to run wild. (changing policy, technology).
3. dynamic models - stability of static solutions, repeated games. (time dependent, fixed policy) changing parametric incentives.
4. static models - used in this.

As a theoretical framework it can be used to resolve paradoxes that been observed by community members and researchers.

• These models can be useful in four use cases.
  • classify behavior - in a finer form using aspects of model. (players, strategies, information, etc).
  • qualify - different behaviors - these can be strategies, pure equilibrium, mixed equilibrium and evolutionary stable strategies.
  • quantify - to estimate the outcomes of rational behavior and detect obstacles to rationality.

In more practical sense this theory should make testable predictions:

• How agents will behave in a sufficiently large sample of modeled situations modeled by the game.
• How a game scenarios would be expected to evolve over time due to invasion of different strategies.
• Systemic fragility arising from different risk factors - such as cheating.
• meta-policy prescription that could increase the success of the wiki.
• it can be used to compare academic and Wikipedia and print models of knowledge development. (Wikipedia, research Journal and Print Encyclopedia respectively)

Criticism

The game theoretic approach assumes rationality on the part of players. In realty players are known to make mistakes. However most of these models have been developed based on reports of both qualitative and quantitative research. And while motivation and rationality are difficult to encode numerically, the advantages of the models is that they simplify reality but are still able to capture phenomena at a level which allows making testable predictions.

Mathematical Methods

Solving problems in the game theoretic paradigm is done according to a number of established solution concepts. For example

• MiniMax.
• Deletion of dominated strategies.
• Backward induction. (assume forward play will be rational).
• Equilibrium.
  • ESS.
  • Proper.
  • Trembling Hand.
  • Sub-game Perfect.
  • Pure Strategy.
  • Nash.
  • Correlated.
• Forward iinduction. (assume past play was rational)
• Core.
• Shapely Value.
• Extended Core.
• QRE - to test models with players that are not always rational.

Experimental Methods

From its inception the theory of games has been viewed as belonging to applied mathematics and accordingly its models compatible with experimental testing.

• some famous experimental games include
• w:Ultimatum
• w:Dictator
• Consensus

see also [1]