 Research
 Open Access
Distributed medium access control with conditionally altruistic users
 Panayotis Antoniadis^{1}Email author,
 Serge Fdida^{2},
 Christopher Griffin^{3},
 Youngmi Jin^{4} and
 George Kesidis^{3}
https://doi.org/10.1186/168714992013202
© Antoniadis et al.; licensee Springer. 2013
 Received: 22 November 2012
 Accepted: 10 July 2013
 Published: 1 August 2013
Abstract
In this paper, we consider medium access control of local area networks (LANs) under limitedinformation conditions as befits a distributed system. Rather than assuming ‘by rule’ conformance to a protocol designed to regulate packetflow rates (e.g., CSMA windowing), we begin with a noncooperative game framework and build a dynamic, conditional, altruism term into the net utility. The effects of altruism are analyzed at Nash equilibrium for the carrier sense multiple access (CSMA)like randomaccess framework in the quasistationary (fictitious play) regime. We consider either power or throughputbased costs of networking, and the cases of identical or heterogeneous (independent) players. Unlike related work in this area, our objective is not to optimize the performance of the system but to formulate a realistic economic model capturing altruistic motivations. This model can then form the basis for higherlevel incentive schemes that will encourage altruistic, instead of selfish, behavior. In a numerical study we consider the cases of identical or diverse players and show interesting relationships between the main parameters of our model and comparisons with other alternatives discussed in the literature.
Keywords
 Medium access control
 Resource allocation
 Altruism
 Game theory
1 Introduction
Flow and congestion control are fundamental networking problems due to the distributed, informationlimited nature of the decisionmaking process in many popular access technologies. Various distributed mechanisms have been implemented to cooperatively desynchronize demand, e.g., Transmission Control Protocol (TCP), ALOHA, and carrier sense multiple access (CSMA). Typically, when congestion is detected, all enddevices are expected to slow down their transmission rates and then slowly increase again hoping to find a fair and efficient equilibrium.
If some users^{a} employ alternative implementations of the prescribed ‘by rule’ protocols, e.g., ones that slow down less than they should or even increase their transmission rate in the presence of congestion, the result could be an unfair allocation or even congestion collapse (see e.g., [1, 2]). The experience with TCP (e.g., [3]) has shown that developers do create versions of the protocol that depart from the standard cooperative (by rule) congestionavoidance algorithm, like Turbo TCP. Outside of a tactical context, noncooperative tactics at the medium access control (MAC) level have not been as widespread, perhaps due to the increased difficulty of modifying lowerlevel networking drivers by users or third parties, but such modifications are possible. As more network users behave selfishly and thereby more significantly reduce the performance of the rest, the other players are increasingly incentivized to adopt selfish strategies themselves, potentially leading to deadlock.
To address this threat, there is a steadily growing literature that analyzes the equilibria of different distributed network resource allocation games[1, 4–13]. Such models provide useful insights on the expected equilibria when users do have the option to choose alternative implementations of the MAC protocol and constitute a framework for devising and analyzing incentive mechanisms to encourage the behavior that would lead to the most desirable equilibria. For example, in a Markovian setting without fictitious play,^{b} Ma et al. [13] introduce a cooperation parameter (a probability to stop transmitting) and then follows a detection and punishment methodology regarding selfish behavior.
In addition, even when users do follow the prescribed protocol, game theoretical models could be used as analytical frameworks that enable more informed choices in the implementation of the corresponding flow and congestion control protocols (e.g., by associating a utility function to enddevices, which can then be the basis of actions by rationally selfish players). To this end, for a randomaccess local area network (LAN), several authors have recently considered the problem of distributed optimization of a global objective (total throughput, social welfare) subject to a fairness constraint. For example, in [12], a utility function design problem is studied considering estimation errors of the network state.
Our work falls into the former category of gametheoretic models but is different than the typical approach in addressing potentially selfish behavior at the MAC layer. Our objective is to formulate a more realistic utility model that captures altruistic motivations. As discussed in the next section, research in the field of experimental economics has demonstrated that such motivations do play an important role in a wide variety of public goods and commonpool games. Our analysis could then form the basis for advanced incentive schemes, which, instead of attempting to punish selfish behavior, would aim to encourage altruistic behavior under certain conditions. For example, a possible realistic outcome could be the design of a highlevel user interface which will allow users to set the urgency of their communication and which encourages users to assign lower priority to their traffic when there is evidence that other users generally are doing the same, cf. Section 7. If successful, such a mechanism will not only improve performance at any given moment but it will also allow certain users to increase their own throughput only when they really need it, improving this way also the efficiency of the system over time without the need for complex and unattractive pricing schemes (e.g., [4]).
In the long term we have to deal with an ‘evolutionary’ game, as defined in [14], that could pass from different states, as in [15]: In one state, all participants follow a basically cooperative strategy (cooperative by rule protocol), though the context here is an informationlimited distributed system. From this cooperative state, suppose that some players defect from the cooperative protocol and thereby (perhaps ‘greedily’) achieve better performance for themselves at the expense of all other players. A sufficient number of defectors (i.e., sufficiently poor performance for nondefectors) will incentivize all remaining players to defect, thus possibly leading to a ‘fully noncooperative’ state. In the fully noncooperative state, an informationlimited game may be prone to nonPareto (even deadlocked) equilibria. One approach to this problem is to employ mechanisms to overcome ‘loss aversion’ [16] and thereby explore play actions that may yield suboptimal net utilities in the short term (i.e., moves that appear locally irrational) but avoid getting trapped in nonPareto equilibria in the longer term. Alternatively again from the fully noncooperative state, the possibility for users to lower the priority of their transmission based on information about the behavior of others can result in a conditionally cooperative state that could help escape from the deadlock and possibly lead back to the fully cooperative one. Notwithstanding such measures, the noncooperative state may not reach the social welfare achieved in the original cooperative state, cf. the numerical results in Section 5.4.
As our main contribution here, we focus on the conditionally cooperative state of the system and formulate and analyze a novel CSMA control game with conditionally altruistic players. We model this situation by altering the net utilities of the players with a term which we identify with altruism. Altruistic tactics in evolutionary/meanfield games have long been considered; see [17] as a recent reference. In networking, altruism has been modeled as a user’s statically personalized weight on the utility of others in games of network formation [18], packet forwarding in delay tolerant networks [19], routing [20, 21], and medium access control by us in [22].
However, we argue in detail in the next section that such static altruistic models, although theoretically interesting, fail to capture important realistic attributes of altruistic behavior studied in the behavioral and experimental economics literature. In this paper, we formulate a fictitious play model where altruism by one user is based on the perceived mean modulated throughput of the other players^{c} (i.e., made ‘dynamic’) by factoring the estimated mean total channel idle time. Unlike Heusse et al. [23], who propose a windowupdate algorithm that tries to directly minimize the average idle time of the channel, in our model users will use less than their ‘fair’ share when they do not really need it, but under the constraint that others do the same. For example, large idle time may be a signal that competing devices are also behaving in a socially sensitive manner, expressing a cooperative ‘social norm.’ In this case, excessive altruism may result in an underused channel. We should stress that our objective is not to optimize the overall throughput of the system but to study the stable equilibria that such altruistic devices could reach. Finally, we do not assume that the users share information and act in a coordinated fashion, i.e., so as to form a player coalition.
This paper is outlined as follows. In Section 2, we give a brief background on altruistic behavior. A fictitious play model with dynamic altruism for a slotted ALOHA LAN is given in Section 3. In Section 4, some closely related variations of the LAN model are considered. Numerical studies are given in Sections 5 and 6, including for the case of player diversity. Finally, in Section 7, we conclude with a summary and discussion of future work.
2 Background on altruistic behavior
Economists are often criticized for the common assumption that humans are rational (i.e., purely selfinterested), which leads to a pessimistic view of the outcome of various formulated gametheoretic models. In reality, many people act ‘altruistically,’ defined as an ‘unselfish concern for or devotion to the welfare of others.’^{d}
In fact, despite this selfishness stereotype, certain branches of economics, such as behavioral and experimental economics, do incorporate social, cognitive, and psychological factors in their models of human behavior (see [16] for a historical overview), in a way not typically captured in cooperative gametheoretic frameworks.
Two common scenarios in which altruistic behavior consistently appears in experiments with real users include the publicgoods provision and commonpool resource sharing games. For example, in the traditional publicgood provision game, where players determine their individual contribution toward the construction of a pure public good, experiments have challenged the assumption that free riding is always the dominant strategy (e.g., [24]). Similarly altruistic behaviors have been observed in a very simple resourcesharing game, called ultimatum, where one player decides how to share a fixed amount of money with another player who can decide whether to accept or reject sharing: here rejection leaves both with zero profit. Experiments show that people altruistically sacrifice their own profit to punish unfair decisions by others; see [25] for an overview of experiments with different variations of this game and interesting regularities observed.
An important lesson of experimental economics is that altruism does not seem to be a static and hardwired characteristic of humans but depends on many aspects of the environment. In other words, the level of altruism of an individual is dynamic and could change over time depending on the context and the behavior of the group [25, 26]. Indeed, the cooperation rate in many experiments has been proven to be much higher if subjects know that there is a possibility of meeting the same partners again in future periods [27], when their perception on the overall level of altruism in their group is high [28], or even just by a positive framing of the experiment [29].
From these and many other contextual factors that can affect the cooperation levels in a group, social norms are perhaps the most influential (see [30, 31]) but complex to incorporate in a simple economic model. To this end, Fehr and Schmidt [32] have proposed a utility function to model the altruistic behavior of people in ultimatum experiments, which incorporates a measure of fairness (or ‘inequity aversion’) in a static way, i.e., its main parameters are indifferent to the dynamics of the system. As a more realistic but less tractable alternative, H. Margolis argues in favor of a more dynamic and complex model, called ‘neither selfish nor exploited’ [33], which proposes a dual utility model which takes into account the history of one’s actions, the current overall behavior, the effect of altruistic action, and the developed norms in a society.
3 Slotted ALOHA randomaccess LAN with dynamic altruism
3.1 Altruistic framework with powerbased cost and concave utility of throughput
In our scenario, the high complexity of human nature and the surrounding social environment plays a less important role since the cooperation game that we study is limited in time, the identity of the players are hidden, the stakes are relatively low, and the decisions of users are mediated through a programmed device.
Chicken game
P2  

P1  Tx  No Tx 
Tx  (−ξ, −ξ)  (0, 1 − ξ) 
no Tx  (1 − ξ, 0)  (0, 0) 
the utility derived by one’s own throughput is modulated by a concave function [4, 7, 35] as modeled here in the form of a logarithm (for tractability); and we have assumed a powerbased cost^{e}Mq. Note that because we assume that the source of each successfully transmitted packet is evident to all nodes, each node i can easily estimate ${\overline{\gamma}}_{i}$. Again, though each player i optimizes V_{ i } in a noncooperative fashion, the game is called altruistic to reflect the second term in (1). In summary, in our model of an altruistic player i, benefit (utility) is derived from the success of others (${\overline{\gamma}}_{i}$) and channel idleness (α_{ i }), the latter indicating altruism on the part of others^{f}.
Note that in classical ALOHA, choosing very high (re)transmission parameter q results in wasted slots due to interference and wasted transmission power, and choosing very low q results in underused (empty) slots. Also note that a singleplay slotted ALOHA game between two identical players is similar to the game chicken. If ξ < 1 is the cost of transmission and the (normalized) payoff of successful transmission is 1, then the Table 1 gives net payoffs for collective action (transmit (Tx) or not) by the players (P1,P2).
The singleplay game has three Nash equilibria: two ‘pure’ strategies, (Tx, noTx) and (noTx, Tx), and one mixed strategy: Tx with probability q^{∗} (and no Tx with probability 1 − q^{∗}), where q^{∗} = 1 − ξ jointly minimizes the expected net gains, (1 − ξ)q_{ k }(1 − q_{3 − k}) − ξ q_{ k }q_{3 −k}, of players k ∈ {1, 2}.
In the following, we consider an iterated version of this game where players pursue mixed strategies based on observations of throughput γ_{ i } observed in steady state.
Note that if we further assume that nodes are aware of the C, M parameters of other nodes, then we can replace $\overline{\gamma}$ with the net utility of the other players as in [22] (particularly for throughputbased costs M γ).
Proof
i.e., equivalent to (3). Note that f(0) = − c < 0 and f(1) = 1 − c > 0, the latter by hypothesis. So, by the continuity of f and the intermediate value theorem, a root of f exists in (0,1).
All such solutions ${q}^{\ast}\underset{\xaf}{1}$ correspond to Nash equilibria because ${\partial}^{2}{V}_{i}\left(\underset{\xaf}{q}\right)/\partial {q}_{i}^{2}={C}_{i}/{q}_{i}^{2}<0$ for all $i,\underset{\xaf}{q}$. □
The following corollary is immediate.
Corollary 3.1. There is a unique symmetric Nash equilibrium point (NEP) if minq ∈ (0, 1)f^{′}(q) > 0 (i.e., f is strictly increasing), a condition on parameters N and a.
Note that there may be nonsymmetric Nash equilibria in these games, even for the case of homogeneous users, e.g., [36]. Also, it is well known that Nash equilibria of iterative games are not necessarily asymptotically stable, e.g., [37–39]. In [4, 35] for a slotted ALOHA game with throughputbased costs M γ, using a Lyapunov function for arbitrary N ≥ 2 players, a noncooperative twoplayer ALOHA was shown to have two different interior^{g} Nash equilibria, only one of which was locally asymptotically stable (see also [40]).
which is a kind of distributed gradient ascent. The Jacobi iteration is also motivated by the desire to take small steps to avoid regions of attraction of undesirable boundary NEPs, particularly those corresponding to the capture strategy (q_{ i } = 1 for some i). Note that when more than one player selects this strategy, the result is a bad outcome for the game chicken or a deadlocked ‘tragedy of the commons.’ Additionally, the players avoid the optout strategy (q_{ i } = 0 for some i). In summary, (6) represents a repeated game in which players adjust their transmission parameters q_{ i } to (locally) maximize their net utility V_{ i }.
To find conditions on the parameters of net utilities (1) for local stability of the equilibria, we can apply the HartmanGrobman theorem [42] to (5), i.e., to check that the Jacobian is negative definite. The following proposition uses the conditions of [43] for stability (and uniqueness) for a special case.
Proof
So, by the Gershgorin circle (disc) theorem (see p. 344 of [44]), all of $H\left(\underset{\xaf}{q}\right)$’s eigenvalues are less than −C + (N − 1)2A. So, if (7) holds, then all the eigenvalues of $H\left(\underset{\xaf}{q}\right)$ are negative, and so $H\left(\underset{\xaf}{q}\right)$ is negative definite. □
3.2 The marginal effect of altruism
In this section, we will write q^{∗} (of the symmetric NEP ${q}^{\ast}\underset{\xaf}{1}$ in symmetric users case) as a function of the normalized altruism parameter a := A/M, q^{∗}(a). Note that q^{∗}(0) = c := C/M.
Recall that the total throughput for slotted ALOHA, N c(1 − c)^{N−1}, is maximal when c = 1/N. The maximum total throughput is (1 − 1/N)^{N−1}≈e^{−1} for large N, i.e., the maximum throughput per player is 1 / (N e) in this cooperative setting without networking costs.
So, if c > 1/N, i.e., total throughput is less than e^{−1} because of excessive demand (overloaded system), then a marginal increase in altruism from 0 (0 < a ≪ 1) will cause a marginal decrease in q^{∗}↓ 1 / N, resulting in an increase in throughput per user γ ↑ 1 / (N e). Also, if c < 1/N, i.e., total throughput is less than e^{−1} because of a lack of demand (an underloaded system), then a marginal increase in altruism from 0 will again cause a marginal decrease in q^{∗}, but here resulting in a decrease in throughput γ (further away from the optimum e^{−1}). See Section 5.4.
4 Closely related model variations
5 Numerical studies for identical players at Nash equilibrium
5.1 Powerbased costs
5.2 Throughputbased costs
5.3 Throughput proportional utilities and costs
5.4 An example comparing altruism and noncooperation
Numerical comparison
Scenario  q ^{∗}  γ ^{∗}  V^{∗}/M 

N = 4  
Dynamic altruism  0.22  0.1044  0.36 
Cooperative  0.18  0.0972  1.34 
Static altruism  0.16  0.0935  0.53 
Noncooperative  0.50  0.0625  1.89 
N = 8  
Cooperative  0.10  0.0478  1.62 
Static altruism  0.28  0.0277  1.52 
Dynamic altruism  0.50  0.0039  3.27 
Noncooperative  0.50  0.0039  3.27 
If there is little demand, the idle time will be high and the throughput low, so the altruism term may or may not be high. If there is congestion, the idle time will be low and the throughput will also be low (the latter due to interference), so the altruism term will be low. Then the players tend to be noncooperative in the congested state (when N is high). Considering the static altruism model, for which channel idle time is not a factor, one expects that the players tend to be noncooperative in the congested state too (throughput is low due to interference), but the altruism term will be higher since the alpha (idle time) term is not involved, so the tendency to be noncooperative will be for a higher N under static altruism than under our dynamic altruism  as is the case for N = 8.
For N = 4, dynamic altruism does better owing to both higher channel idle time and higher total throughput. In this lowcontention case, the dynamic altruism term is smaller than in the static case (owing to the idletime component α) so users will value their own throughput utility more (and they can achieve more throughput owing to less contention), which is collectively reenforced by the throughput component of the dynamic altruism term. Furthermore, for this lowcontention case, the cooperative strategy has simply ‘met its demands’ (maximized utility minus cost), and under powerbased costs, increasing transmission probability has a great reduction on cost components of the net utility terms and again boosts the altruism components of the utility of other players.
6 Numerical studies with player diversity
6.1 Players with different altruism parameters
The role of a
a _{1}  ${\mathit{q}}_{\mathbf{1}}^{\mathbf{\ast}}$,${\mathit{q}}_{\mathbf{2}}^{\mathbf{\ast}}\mathbf{=}{\mathit{q}}_{\mathbf{3}}^{\mathbf{\ast}}$  ${\mathit{\gamma}}_{\mathbf{1}}^{\mathbf{\ast}}$,${\mathit{\gamma}}_{\mathbf{2}}^{\mathbf{\ast}}\mathbf{=}{\mathit{\gamma}}_{\mathbf{3}}^{\mathbf{\ast}}$  i${\mathit{V}}_{\mathbf{1}}^{\ast}$,${\mathit{V}}_{\mathbf{2}}^{\mathbf{\ast}}\mathbf{=}{\mathit{V}}_{\mathbf{3}}^{\mathbf{\ast}}$ 

30  0.15,0.10  0.13,0.074  0.754,2.37 
40  0.12,0.10  0.10,0.080  1.40,2.24 
50  0.10,0.10  0.083,0.083  2.10,2.10 
60  0.091,0.11  0.073,0.087  2.79,1.83 
70  0.079,0.11  0.063,0.090  3.56,1.82 
Following intuition, increased altruism (a_{1} > 50) by player 1 resulted in lower throughput for him and higher throughput for the other two players. Similarly, decreased altruism by player 1 (a_{1} < 50) resulted in higher throughput for him and lower throughput for the other players.
6.2 Sizes of regions of attractions under different play rates
The role of r
Volume  NEP = (0.1)1  NEP = (0.75)1 

r = 0.1  0.502  0.498 
r = 0.25  0.507  0.493 
r = 1  0.556  0.444 
r = 4  0.839  0.161 
r = 10  0.841  0.159 
γ:  0.081  0.047 
$\stackrel{~}{V}$:  1.94  1.43 
Again following intuition, a lower r effectively corresponds to a reluctance to be altruistic and thereby results in a smaller domain of attraction for the more altruistic Pareto (higher $\stackrel{~}{V}$) equilibrium $\left(0.1\right)\underset{\xaf}{1}$.
6.3 Players with different transmission priorities
Now consider the case where the factors C_{ i } in (1) are different. Equivalently, assume the factors r_{ i } affect only the throughput utility rather than the whole net utility as they do in (13). Taking the cases of such factors as in the previous Subsection 6.2 again leads to intuitively consistent results as above: users with higher priority will achieve more throughput at Nash equilibrium.
7 Conclusions
In this paper, we extended a noncooperative game framework for informationlimited MAC of a LAN by adding an altruism term that depended on both the mean throughput of the other players and the mean channel idle time. The cases of heterogeneous or homogeneous users, and of power or throughputbased costs, were considered for a quasistationary model of the game. A numerical study compares the peruser throughput under dynamic and static altruism with that of purely noncooperative dynamics, and demonstrates the advantage of altruism under moderate levels of congestion (number of players) in the homogeneous player setting and for a heterogeneous user scenario. Our numerical studies produced intuitive results which means that our model is selfconsistent and could form the basis for more sophisticated extensions. However, our dynamic altruism term is sensitive to the use of the mean idle time as a measure of the current level of altruism in the system, which could lead to wrong interpretations in certain scenarios (e.g., when demand is low or when congestion is due to a high number of users in the channel). In more advanced versions of our model, we will include the number of competing users, N, in the term expressing the current level of altruism in the system in order to avoid such misinterpretations.
In the future, we will also consider a mixed scenario of of three types of players:

Those that follow an original protocol that enacts distributed/informationlimited cooperation (by rule) for flow and congestion control

Altruistic but pragmatic (second defectors) who will defect to avoid starvation, and

Selfish (first defectors), who will cooperate only to avoid starvation.
Note that both types of defectors can engage in an evolutionary cycle^{j} of tactical transitions as they assess the social communication norms in the LAN among users who are active presently and in the recent past, classifying the active users into the above three categories in particular. One can ask what distributed congestion and flow control protocol can best deal with defectors of both types^{k}? Given that players will be intermittently active or may be active with communication of differing degrees of priority, can an altruistic framework (possibly with an evolutionary ‘wrapper’) be designed to effectively conduct priority scheduling in this LAN context? A challenge here is dealing with the greedy user who declares all of their communication as high priority. Again, our ultimate aim is to achieve fair and efficient throughputs for by rule cooperators and altruistic defectors alike, while not starvingout/shunning the deemed selfish defectors.
Endnotes
^{a} In this paper, we use the terms user, player, participant and node interchangeably.
^{b} I.e., without steadystate estimates of certain quantities.
^{c} Such estimates are feasible in our application context of a CSMA local area (broadcast) network with a relatively small population of active participants, but is not possible for many other networking contexts, e.g., the example of TCP congestion control mentioned above. So, it would be difficult to obtain estimates of ‘social norms’ to form the basis of ‘rational’ altruism for TCP.
^{d} See http://dictionary.reference.com/browse/altruism
^{e} Powerbased costs are borne whether or not the transmission is successful.
^{f} Obviously, we could have combined α_{ i } and ${\overline{\gamma}}_{i}$ in different ways, instead of a product form, to form an altruism modifier for the net utility of player i, cf. the next section for other model variations.
^{g} I.e., not including the stable boundary deadlock equilibrium at $\underset{\xaf}{q}=\underset{\xaf}{1}$.
^{h}cf. Section 4.4 for a discussion of asynchronous play.
^{i} This is similar to the classical slotted ALOHA example where all (identical) players choose a common q=1/N to maximize total (and individual) throughputs(without considering costs).
^{j} Possibly at a slow time scale of human response.
^{k} At a (faster) time scale of machine response.
Declarations
Acknowledgements
GK was supported by NSF CISE grants 0524202 and 0915928 by a Cisco Systems URP gift; PA was supported by EINS, the Network of Excellence in Internet Science EC’s FP7 grant 288021. YJ was supported by NRF2013R1A1A3A04007104. Elements of this paper were presented at the IEEE CDC, Atlanta, 2010 and AdHocNets, Paris, 2012 conferences.
Authors’ Affiliations
References
 Cagalj M, Ganeriwal S, Aad I, Hubaux J: On selfish behavior in CSMA/CA networks. In Proceedings of the IEEE INFOCOM. Miami; 13–17 March 2005.Google Scholar
 Raya M, Aad I, Hubaux JP, Fawal AE: DOMINO: detecting MAC layer greedy behavior in IEEE 802.11 hotspots. IEEE Trans. Mobile Comput 2006, 5(12):16911705.View ArticleGoogle Scholar
 Akella A, Seshan S, Karp R, Shenker S, Papadimitriou C: Selfish behavior and stability of the Internet: A gametheoretic analysis of TCP. In Proceedings of ACM SIGCOMM. Pittsburgh; 1923 August 2002.Google Scholar
 Jin Y, Kesidis G: A pricing strategy for an ALOHA network of heterogeneous users with inelastic bandwidth requirements. In Proceedings of the CISS. (Princeton; 20–22 March 2002.Google Scholar
 Alpcan T, Basar T, Srikant R, Altman E: CDMA uplink power control as a noncooperative game. Wireless Netw 2002, 8: 659670. 10.1023/A:1020375225649View ArticleGoogle Scholar
 Wicker S, MacKenzie A: Stability of multipacket slotted Aloha with selfish users and perfect information. In Proceedings of IEEE INFOCOM. San Francisco, CA; 30 March–3 April 2003.Google Scholar
 Jin Y, Kesidis G: Dynamics of usagepriced communication networks: the case of a single bottleneck resource. IEEE/ACM Trans. Network 2005, 13: 10411053.View ArticleGoogle Scholar
 Meshkati F, Chiang M, Poor H, Schwartz S: A gametheoretic approach to energyefficient power control in Multicarrier CDMA systems. IEEE JSAC 2006, 24(6):11151129.Google Scholar
 Lee J, Chiang M, Calderbank R: Utilityoptimal randomaccess protocol. IEEE Trans. Wireless Commun 2007, 6(7):27412751.View ArticleGoogle Scholar
 Long C, Zhang Q, Li B, Yang H, Guan X: Noncooperative power control for wireless ad hoc networks with repeated games. IEEE J. Selected Areas Commun 2007, 25(6):11011112.View ArticleGoogle Scholar
 Jin Y, Kesidis G: Distributed contention window control for selfish users in IEEE 802.11 wireless LANs. IEEE JSAC Spec. Issue. Noncooperative Netw 2007, 25: 11131123.Google Scholar
 Cui T, Chen L, Low S: A gametheoretic framework for medium access control. IEEE J. Selected Areas Commun 2008, 26(7):11161127.View ArticleGoogle Scholar
 Ma R, Misra V, Rubenstein D: An analysis of generalized slottedaloha protocols. IEEE/ACM Trans. Netw 2009, 17(3):936949.View ArticleGoogle Scholar
 Weibull J: Evolutionary Game Theory. Cambridge, MA: MIT Press; 1995.Google Scholar
 Nowak N: Five rules for the evolution of cooperation. Science 2006, 314: 15601563. 10.1126/science.1133755View ArticleGoogle Scholar
 Camerer C, Loewenstein G: Behavioral economics: past, present, future. In Advances in Behavioral Economics. Edited by: Camerer C, Loewenstein G, Rabin M. Princeton: Princeton Univ. Press; 2003.Google Scholar
 Huang M, Caines P, Malhame R: Social dynamics in mean field LQG control: egoistic and altruistic agents. In IEEE Conference on Decision and Control (CDC). Atlanta, Georgia: IEEE; 15–17 December 2010.Google Scholar
 Avrachenkov K, Elias J, Martignon F, Neglia G, Petrosyan L: A Nash bargaining solution for cooperative network formation games, Networking 2011 Lecture Notes in Computer Science. 6640: 307318.Google Scholar
 Hui P, Xu K, Li V, Crowcroft J, Latora V, Lio P: Selfishness, altruism and message spreading in mobile social networks. In Proceedings of the First IEEE International Workshop on Network Science For Communication Networks (NetSciCom). Rio de Janeiro; 24 April 2009.Google Scholar
 Azad A, Altman E, Elazouzi R: From altruism to noncooperation in routing Games. In Proceedings of Networking and Electronic Commerce Research Conference (NAEC). Dallas: ATSMA; 8–11 October 2009.Google Scholar
 Sharma Y, Williamson D: Stackelberg thresholds in network routing games or the value of altruism. Games Econ. Behav 2009, 67: 174190. 10.1016/j.geb.2009.06.006MathSciNetView ArticleGoogle Scholar
 Kesidis G, Jin Y, Amar A, Altman E: Stable Nash equilibria of LAN medium access games under symmetric, socially altruistic behavior. Proceedings of IEEE CDC 2010 [Technical report number arXiv:1003.5324 available at http://arxiv.org] [Technical report number arXiv:1003.5324 available at ]Google Scholar
 Heusse M, Rousseau F, Guillier R, Dula A: Idle sense: an optimal access method for high throughput and fairness in rate diverse wireless LANs. In Proceedings of ACM SIGCOMM.. New York: ACM; 2005:121132.Google Scholar
 Ledyard J, Roth A: Public goods: a survey of experimental research. In Handbook of Experimental Economics. Edited by: Kagel J. Princeton: Princeton University Press,; 1995.Google Scholar
 Camerer C, Fehr E: Measuring social norms and preferences using experimental games: a guide for social scientists. In Foundations of Human Sociality: Economic Experiments and Ethnographic Evidence from Fifteen SmallScale Societies. Edited by: Henrich J, Boyd R, Bowles S, Camerer C, Fehr E, Gintis H. Oxford: Oxford University Press,; 2004.Google Scholar
 Levitt SD, List JA: What do laboratory experiments measuring social preferences reveal about the real world? J. Econ Perspect 2007, 21(2):153174. 10.1257/jep.21.2.153View ArticleGoogle Scholar
 Fehr E, Fischbacher U: The nature of human altruism. Nature 2003, 425: 785791. 10.1038/nature02043View ArticleGoogle Scholar
 Bicchieri C, Xiao E: Do the right thing: but only if others do so. J. Behav. Decis. Mak 2009, 22: 191208. 10.1002/bdm.621View ArticleGoogle Scholar
 Dufwenberg M, Gächter S, HennigSchmidt H: The framing of games and the psychology of play. Games Econ. Behav 2011, 73(2):459478. 10.1016/j.geb.2011.02.003View ArticleGoogle Scholar
 Ostrom E: Collective action and the evolution of social norms. J. Econ. Perspect 2000, 14(3):137158. 10.1257/jep.14.3.137View ArticleGoogle Scholar
 Bicchieri C: The Grammar of Society: the Nature and Dynamics of Social Norms. New York: Cambridge University Press; 2006.Google Scholar
 Fehr E, Schmidt K: A theory of fairness, competition, and cooperation. Q. J. Econ 1999, 114(3):817868. 10.1162/003355399556151View ArticleGoogle Scholar
 Margolis H: Cognition and Extended Rational Choice. New York: Routledge; 2007.View ArticleGoogle Scholar
 Brown G: Iterative solutions of games with fictitious play. In Activity Analysis of Production and Allocation. Edited by: Koopmans T. New York: Wiley:; 1951.Google Scholar
 Jin Y, Kesidis G: Equilibria of a noncooperative game for heterogeneous users of an ALOHA network. IEEE Commun. Lett 2002, 6(7):282284.View ArticleGoogle Scholar
 Jin Y, Kesidis G: A channelaware MAC protocol in an ALOHA network with selfish users. IEEE JSAC Spec. Issue on Game Theory Wireless Commun 2012, 30: 128137.Google Scholar
 Seade J: The stability of Cournot revisited. J. Econ. Theory 1980, 23: 1527. 10.1016/00220531(80)900289MathSciNetView ArticleGoogle Scholar
 AlNowaihi A, Levine P: The stability of the Cournot oligopoly model: a reassessment. J. Econ. Theory 1985, 35: 307321. 10.1016/00220531(85)900468MathSciNetView ArticleGoogle Scholar
 Zhang A, Zhang Y: Stability of Nash equilibrium: the multiproduct case. J. Math. Econ 1996, 26(4):441462. 10.1016/03044068(95)007601View ArticleGoogle Scholar
 Menache I, Shimkin N: Fixedrate equilibrium in wireless collision channels. In Proceedings of the Network Control and Optimization (NETCOOP). Avignon; 5–7 June 2007.Google Scholar
 Shamma J, Arslan G: Dynamic fictitious play, dynamic gradient play, and distributed convergence to Nash equilibria. IEEE Trans. Auto. Contr 2005, 50(3):312327.MathSciNetView ArticleGoogle Scholar
 Perko L: Differential Equations and Dynamical Systems. New York: SpringerVerlag; 2011.Google Scholar
 Rosen J: Existence and uniqueness of equilibrium points for concave Nperson games. Econometrica 1965, 33(3):520534. 10.2307/1911749MathSciNetView ArticleGoogle Scholar
 Horn R, Johnson C: Matrix Analysis. Cambridge: Cambridge Univ. Press; 1988.Google Scholar
 Basar T, Olsder G: Dynamic Noncooperative, Game Theory. Academic Press. Philadelphia: Society for Industrial and Applied Mathematics; 1999.Google Scholar
 Bertsekas D, Tsitsiklis J: Convergence rate and termination of asynchronous iterative algorithms. In Proceedings of the 3rd International Conference on Supercomputing. Crete; 5–9 June 1989.Google Scholar
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.