Suchergebnisse

Abstract
We study long run implications of reinforcement learning when two players repeatedly interact with one another over multiple rounds to play a finite action game. Within each round, the players play the game many successive times with a fixed set of aspirations used to evaluate payoff experiences as successes or failures. The probability weight on successful actions is increased, while failures result in players trying alternative actions in subsequent rounds. The learning rule is supplemented by small amounts of inertia and random perturbations to the states of players. Aspirations are adjusted across successive rounds on the basis of the discrepancy between the average payoff and aspirations in the most recently concluded round. We define and characterize pure steady states of this model, and establish convergence to these under appropriate conditions. Pure steady states are shown to be individually rational, and are either Pareto-efficient or a protected Nash equilibrium of the stage game. Conversely, any Pareto-efficient and strictly individually rational action pair, or any strict protected Nash equilibrium, constitutes a pure steady state, to which the process converges from non-negligible sets of initial aspirations. Applications to games of coordination, cooperation, oligopoly, and electoral competition are discussed.

We develop a model of adaptive learning with social comparisons. Actors are more likely to choose actions that recently yielded satisfactory payoffs; satisfaction is evaluated relative to an aspiration level that reflects previous payoffs and possibly other players' payoffs. This captures the phenomenon of social comparison viareference groups. We show that if agents compare themselves to those who are receiving higher payoffs then in stable outcomes all payoffs must be equal. If, however, agents' aspirations are driven by less ambitious social comparisons then very unequal distributions can be stable. We apply our general results to collective action problems in socio-political hierarchies and derive conditions for stable exploitation. Finally, we develop a computational model, which shows that increases in payoff inequality make outcomes less stable.

One of the best known ideas in the study of bounded rationality is Simon's satisficing; yet we still lack a standard formalization of the heuristic and its implications. We propose a mathematical model of satisficing which explicitly represents agents' aspirations and which explores both single-person and multi-player contexts. The model shows that satisficing has a signature performance-profile in both contexts: (1) it can induce optimal long-run behavior in one class of problems but not in the complementary class; and (2) in the latter, it generates behavior that is sensible but not optimal. The model also yields empirically testable predictions: in certain bandit-problems it pins down the limiting probabilities of each arm's use, and it provides an ordering of the arms' dynamical use-probabilities as well.

In the last decade, there has been a resurgence of interest in problems of cooperation, stimulated largely by Axelrod's work. Using an innovative tournament approach, Axelrod found that a simple strategy, tit-for-tat (TFT), was most successful in playing the repeated prisoner's dilemma (PD) in a noiseless environment. However, recent analytical work has shown that monitoring problems caused by noise significantly impair TFT's effectiveness. The primary purpose of the present research is to discover whether there exist alternative strategies that perform well in noisy PDs. To investigate this question, the authors conducted a computer tournament. The results of the tournament demonstrated that, consistent with analytical work, TFT performed rather poorly. In contrast, strategies that were generous (i.e., cooperated more than their partners did) were quite effective.

Empirical studies suggest that mission-oriented bureaucrats bias their design of program alternatives to increase the odds that a superior will choose the kind of program the officials want. However, political executives may anticipate this manipulation and try to reassert control. These struggles are examined in three models. In Model 1 a senior bureaucrat is interested only in missions; the bureaucrat's political superior controls him or her by rejecting inferior proposals and entertaining new options from other policy specialists. Model 2 is a principal-agent analysis. Here the official is interested only in budgets; the official's superior reduces search bias by creating an ex ante incentive scheme. In Model 3 the bureaucrat cares about both budgets and programs; the superior uses both his or her final review authority and ex ante incentives to reduce agenda manipulation. The models' contrasting implications for the political control of bureaucracy are examined.

The empirical literature on the control of bureaus notes that politicians have difficulty observing bureaucratic output, but this insight is rarely represented informal models. To analyze how bureaus use this uncertainty strategically, we develop a model of expertise-based agenda control, building on the Niskanen (1971) and Miller and Moe (1983) tradition. We show that under some plausible conditions, bureaus will underestimate the benefits, and overestimate the costs, of their programs. In the model, politicians are neither passive nor omniscient: they anticipate the bureau's strategic behavior and establish a monitoring system to counteract it. This possibility of detection changes the bureau's behavior: even imperfect monitoring reduces the bureau's deception of the legislature, whether or not the legislature's demand for the bureau's services is concealed. Moreover, uncertainty by itself matters: if the legislature makes it harder for a risk-averse bureau chief to predict demand or penalty, the bureau will restrain its deception.

The garbage can theory of organizational choice is one of the best-known innovations in modern organization theory. It also has significantly shaped a major branch of the new institutionalism. Yet, the theory has not received the systematic assessment that it both deserves and needs. We evaluate the early verbal theory and argue that it fails to create an adequate foundation for scientific progress. We then analyze and rerun Cohen, March, and Olsen's computer model and discover that its agents move in lockstep patterns that are strikingly different from the spirit of the theory. Indeed, the simulation and the theory are incompatible. Next, we examine how the authors have built upon these incompatible formulations in developing the theory further. We assess this larger program, which includes the March-Olsen version of the new institutionalism, and find that many of the problems that attended the original article have intensified over time. We conclude that a fundamental overhaul is required if the theory is to realize its early promise.

The garbage can theory of organizational choice is one of the best-known innovations in modern organization theory. It also has significantly shaped a major branch of the new institutionalism. Yet, the theory has not received the systematic assessment that it both deserves & needs. We evaluate the early verbal theory & argue that it fails to create an adequate foundation for scientific progress. We then analyze & rerun Cohen, March, & Olsen's computer model & discover that its agents move in lockstep patterns that are strikingly different from the spirit of the theory. Indeed, the simulation & the theory are incompatible. Next, we examine how the authors have built upon these incompatible formulations in developing the theory further. We assess this larger program, which includes the March-Olsen version of the new institutionalism, & find that many of the problems that attended the original article have intensified over time. We conclude that a fundamental overhaul is required if the theory is to realize its early promise. 4 Figures, 52 References. Adapted from the source document.