Research On The Intervention Mechanism Of Cooperative Game And Coordination Game: Theory And Experiment

The thesis studies experimentally the mechanisms in improving cooperation and coordination and the related theories.It includes three topics.Each of the first two topics includes an innovative experimental design of a pregame intervention.We focus on how the mechanisms function and how theories support the experimental results.The third part expands the mechanism proposed in the second topic in both theoretical and applicable aspects.In first topic,we study cooperation problem in a linear public goods game with multiple players,which has been recognized as a classic problem in behavioral game theories.We firstly propose a real-time auction-like mechanism featuring incremental commitment and test it in the laboratory.Our results show that with the new method introduced to the game,decreasing trend of the public goods provision,as was frequently observed in numerous previous studies,is eliminated.As the experiments go on,the public goods provision increases gradually,and the final provision level converges to about 75% of the maximal level.Compared with the benchmark group without mechanism intervention,cooperation efficiency is improved by nearly 327%.To better understand how the auction-like mechanism improves the cooperation performance,we add two additional “intermediate” treatments to fill in the setting gap between benchmark treatment with no mechanism intervention and main treatment with auction-like mechanism.Four treatments follow a progressive type in the sense that each treatment adds a new factor to the former one.In the first “intermediate”treatment,we add a “real-time signaling” to the linear public goods game.To be specific,during a 1-minute interval,subjects could signal their intended contribution as many times as they want to their group members,but their final contribution decision is not necessarily binding to their signal.In the second “intermediate”treatment,we add a“final commitment” based on the first “intermediate” treatment,where subjects could still signal their intended contribution,except that their last intended contribution will be committed to be their final contribution.Finally,auction-like mechanism not only requires the last signal to be binding,but also requires “incremental commitment” in the process of real-time signaling.The experiment results of these two “intermediate” treatments show that,“real-time signaling” has no effects in improving cooperation performance,by adding “finalcommitment” to the “real-time singnaling”,the cooperation rate improves slightly but is not significant.Combining the results of all four treatments we conclude that the factor “incremental commitment” in the auction-like mechanism is the key to improve cooperation.It worth noting that,we operate all treatments with random matching for two reasons.First,random matching set-up could avoid the influence of inter-temporal factors(reputation,reciprocity,retaliation,etc.),so that we can compare the results with theoretical prediction more precisely.Second,since it is apparent that high level cooperation is much harder to achieve in random-matching set-up than in fixed matching set-up,we might predict that auction-like should be effective as well in the fixed-matching set-up.In the theoretical aspect,we first show that the traditional concept of Nash equilibrium and technique of backward induction fails to predict our results.We then introduce the concept of ?-equilibrium theory developed by Radner(1986),which combines the ?-equilibrium theory with the inequality aversion model developed by Fehr&Schmidt(1999).We prove that the high cooperation rate in auction-like mechanism could be supported as an ?-equilibrium.In second topic,we study coordination problem in minimum-effort game(Van-huyck et al.,1990)with multiple players,another classic problem in behavioral game theory.When a group is trapped in the lowest-efficiency Nash equilibrium,individuals exert their resources(capacity,time,effort,etc.)at the lowest level,resulting in a waste of their resources.However,there has been little discussion on how to use this part of unused resources to improve coordination efficiency in the literature.We propose an extra-resource exchange mechanism to improve coordination by innovatively taking advantage of efforts which are not used in the coordination game.We allow subjects to use these efforts before the game starts,which can be considered as a channel to signaling and a way to implement forward induction.The usage of these unused efforts can facilitate their coordination in high efforts.We show that the mechanism significantly improves the coordination performance.91.7% of the groups with mechanism intervention reached coordination success,which was 50% higher than the control groups.Extra-resource exchange mechanism works mainly through two channels.First,subjects could signal their intention of coordinating at the most efficient Nash equilibrium by exchanging fewer resources or even exchanging no resources in thefirst stage.Second,the mechanism allows subjects to mutually observe the exchanged resources before the coordination game.It helps subjects identify the intentions of group members on coordination and do forward induction and reach the highest possible coordination level.With the help of two auxiliary treatments,we experimentally confirm the effect of two channels mentioned above.In the first auxiliary treatment,we keep the pregame exchanging stage,while shut down the window of information disclosure about subjects' exchanging decisions.The experimental results on this treatment show that when information of others' exchanging decisions are not available,subject exchange much more resources in the pregame stage,and groups' coordination level decreases to the lowest level.In the second auxiliary treatment,pregame stage is removed,subjects are exogenously endowed with efforts to play the coordination game,their efforts are set to be exactly the same to those in the treatment with mechanism intervention.By control the volatility of strategy spaces in the mechanism,the process of how subjects did forward induction were clearly observed.The results show that forward induction is the key for mechanism to boost the level of coordination.Due to the difficulties of theoretically analysis on peoples' forward induction behavior in multi-players games,current theories on this issue still mainly focus on two-players games.The results in the second topic suggests that people could correctly do forward induction even in multi-players games and reach desired equilibrium,which provide some positive evidences on the future establishment of relative theoretical work.In third topic,we explore the mechanism proposed in second topic in mainly two aspects.First,we simplify the original form of minimum-effort game to a two-player-two-strategy version.Second,we examine the mechanism in random matching set-up instead of the fixed matching set-up in last topic.The motivations of this topic are as follow:First,previous studies discussed little about coordination games in the random matching set-up and it has been found that successful coordination is more difficult in the environment of random matching(Chaudhuri,Paichayontvijit and So,2015).However,coordination problem often takes place in a random matching manner.For example,in some volunteer activities,volunteers from places all over the world are often unfamiliar with each other,but the tasks they need to complete often feature a coordination property,which could be modeled as a random matching coordinationgame.Therefore,it is of practical significance to be able to find a mechanism that can help people perform efficient coordination even in this kind of situation.Second,since the aim of the second topic is to test the effect of the mechanism on coordination improvement,our design strictly follow the classic design of fixed matching,seven-strategies minimum effort game as used in previous researches.However,both multiple strategies and fixed matching set-up put unnecessary difficulties to analyze the mechanism theoretically.Multiple strategies increase the subjects' difficulties to do forward induction,and the fixed matching set-up brings effect of inter-temporal factors(credit,guidance,retaliation,etc.).Hence,for the purpose of testing the effect of the extra-exchange mechanism in more general environment,and also of providing more experimental evidence for the theory of forward induction,it is necessary to perform experiments in a random matching set-up.Considering the difficulty of coordination under random matching environment,we think that it is a reasonable start to study from the simplest two-players version of minimum-effort game: on the one hand,if the mechanism does not work well in the case of two players,we can naturally infer that there is no need to test it in a multiple players situation;on the other hand,if the effect of the mechanism is significant in two-players situation and poor in multiple players situation,then we can analyze the difference between these two situations and try to consider the new mechanism to improve coordination with multiple players.We conducted experiments for the simplified minimum-effort game stated above,both with and without the extra-exchange mechanism.The experimental results show that the extra-exchange mechanism is effective in improving coordination,even when subjects match each other randomly.Near 80% of total cases result in efficient Nash equilibrium,which is significantly higher than the rate of 25% in the control treatment.However,there is also a new finding that worth noting for further researches.We find that some subjects display competitive preference when face random matching games.That is,rather than the win-win situation of efficient coordination,these subjects prefer to beat their counterparts in the outcome.And this type of behavior can only emerges in the environment of random matching because this is the place where subjects should not worry about the inter-temporal factors such as retaliation or reputation building.Considering that such kind of subjects may cause huge damage to the whole group when doing coordination,we think it's an important factor that is worth noting in the future researches.