|Ph.D Student||Cohensius Gal|
|Subject||Games: Models, Mathematical Objects and Artificial|
|Department||Department of Industrial Engineering and Management||Supervisor||ASSOCIATE PROF. Reshef Meir|
|Full Thesis text|
I have studied games in three fields of research: game-theory (GT), combinatorial-game theory (CGT) and AI agents that play recreational games.
The games found in the GT field are models of real-world phenomena, their purpose is to predict the behavior of decision makers when they interact with one another.
The second field deals with games as mathematical objects, it produces a fertile ground for theorems. However beside the beauty of their properties, they are usually not particularly fun to play, nor capture some realistic situations (such
as common economic interactions). The third field is about AI agents that play games that people play for fun.
In the GT chapter we propose Proxy-Voting, a voting mechanism that improves the outcome of certain elections.
In two follow-up works we show that this mechanism can benefit truth discovery and crowdsourcing.
In the CGT chapter we examine a variant for the game Nim. For this variant, we found and proved the optimal strategy for two special cases of interest.
In the general case, we proved that when looking at the outcome as a function of the starting position, we get a periodic function, that is to say, the outcome repeats its values in regular intervals.
In the AI-agents chapter we wrote and analyzed an expert-level playing agents for the game of Spades, a trick-taking-card-game which resembles Bridge.
The bidding algorithm that those agents use is superior to any published Spades-bidding algorithm. The agents were implemented in an online game engine, where they play tens of thousands of games daily. Their win rate proves that they are superior to average recreational human players.