Local Reward Research Articles

Civil Rights Success and the Politics of Racial Violence*

This investigation revises the two main explanations for the successes of the civil rights movement: the backlash thesis and business moderation theory. While both theories hinge on the political significance of severe anti-rights violence, neither approach adequately explains variation in the intensity of this contention. Introducing a political mobilization perspective, which draws attention to the competition between segregationist and moderate business organizations, I argue that the structure of local electoral rewards determined the likelihood of official instigation or toleration for anti-rights violence. Case studies of four civil rights campaigns are used to demonstrate that the severity of anti-rights contention depended upon the relative political capacities of these interests. Refining the backlash thesis, it is suggested that the civil rights movement triggered the dramatic clashes necessary for advancing national legislation only where key economic interests lacked the will or political influence to challenge successfully segregationist political mobilization. Recasting business moderation theory, this analysis indicates that victories at the state and local level prior to substantive federal legislation depended not only upon the political leverage of moderate business organizations, but on a corresponding weakness among segregationists.

Emergent synthesis of motion patterns for locomotion robots

Emergence of stable gaits in locomotion robots is studied in this paper. A classifier system, implementing an instance-based reinforcement-learning scheme, is used for the sensory-motor control of an eight-legged mobile robot and for the synthesis of the robot gaits. The robot does not have a priori knowledge of the environment and its own internal model. It is only assumed that the robot can acquire stable gaits by learning how to reach a goal area. During the learning process the control system is self-organized by reinforcement signals. Reaching the goal area defines a global reward. Forward motion gets a local reward, while stepping back and falling down get a local punishment. As learning progresses, the number of the action rules in the classifier systems is stabilized to a certain level, corresponding to the acquired gait patterns. Feasibility of the proposed self-organized system is tested under simulation and experiment. A minimal simulation model that does not require sophisticated computational schemes is constructed and used in simulations. The simulation data, evolved on the minimal model of the robot, is downloaded to the control system of the real robot. Overall, of 10 simulation data seven are successful in running the real robot.

Local reinforcement and recombination in classifier systems.

We investigate classifier systems' reward schemes by way of an example that highlights the interaction of local reward schemes and recombination. We contrast averaging schemes and maximizing schemes. Our example illustrates a sense in which certain recombination operators mesh more gracefully with averaging schemes than with maximizing schemes.

Differences in Individual Learning Between Group-Foraging and Territorial Zenaida Doves

Abstract Adaptive views of learning predict that natural selection should lead to differences in specialized learning abilities between animals that face different ecological pressures. Group-living is thought to favour social learning, but previous comparative work suggests that differences between gregarious feral pigeons (Columba livia) and territorial Zenaida doves (Zenaida aurita) exceed the specialized effect on social tasks predicted by the adaptive hypothesis. In this paper, we show that group-foraging Zenaida doves from Barbados learn an individual shaping task more quickly than territorial Zenaida doves from a site 9 km away. These results suggest that the scramble competition associated with group-foraging favours several types of leaming, both social and non-social, and that its effects are more wide-ranging than previously thought. Since genetic isolation between Zenaida dove populations is highly unlikely, the results also suggest that differences in foraging ecology may lead to different learned responses to local reward contingencies as well as natural selection for different genotypes affecting learning. In some cases, the standard comparative prediction of ecologically-correlated learning differences may therefore not distinguish between adaptive specialization and general process theories.

PRISONER'S DILEMMA AND THE PIGEON: CONTROL BY IMMEDIATE CONSEQUENCES

In three experiments pigeons played (i.e., chose between two colored keys) iterated prisoner's dilemma and other 2 x 2 games (2 participants and 2 options) against response strategies programmed on a computer. Under the prisoner's dilemma pay-off matrix, the birds generally defected (i.e., pecked the color associated with not cooperating) against both a random response (.5 probability of either alternative) and a tit-for-tat strategy (on trial n the computer "chooses" the alternative that is the same as the one chosen by the subject on trial n - 1) played by the computer. They consistently defected in the tit-for-tat condition despite the fact that as a consequence they earned about one third of the food that they could have if they had cooperated (i.e., pecked the "cooperate" color) on all the trials. Manipulation of the values of the food pay-offs demonstrated that the defection and consequent loss of food under the tit-for-tat condition were not due to a lack of sensitivity to differences in pay-off values, nor to strict avoidance of a null pay-off (no food on a trial), nor to insensitivity to the local (current trial) reward contingencies. Rather, the birds markedly discounted future outcomes and thus made their response choices based on immediate outcomes available on the present trial rather than on long-term delayed outcomes over many trials. That is, the birds were impulsive, choosing smaller but more immediate rewards, and did not demonstrate self-control. Implications for the study of cooperation and competition in both humans and nonhumans are discussed.

Learning probabilistic RAM nets using VLSI structures

Hardware-realizable learning probabilistic RAMs (pRAMs) which implement local reinforcement rules utilizing synaptic rather than threshold noise in the stochastic search procedure are described. The design allows for both global and local rewards and penalties (in this latter case implementing a modified version of backpropagation). The architecture allows for serial updating of the weights of a pRAM net according to a reward/penalty learning rule. It is possible to generate a new set of pRAM outputs at least every 100 mu s, which is faster than the response time of biological neurons. >

Cycles in extensive form perfect information games

We define and analyse a new class of perfect information games. The nodes of a directed graph G are partitioned into n player sets. Starting at a fixed node of G an infinite path is created as follows: If the current node belongs to player k, then player k chooses any successor node. A local reward n-vector is assigned to every arc. The payoff corresponding to the infinite path is the long term average of the local reward vectors. Such games are called DGA games. Negative and positive results are obtained for the existence of Nash equilibria in certain types of pure strategies (e.g., stationary and automated strategies). Applications to duopoly pricing models and “surveillance games” on graphs are given.

A local regulator's rewards for conformity in policy

This note is an empirical investigation of the benefits that accrue to local building departments for a regulatory policy that conforms to the relative strength of local interest groups. The study uses an interest group model and applies it to data for building departments in 1100 municipalities. It finds that material benefits in terms of budgets and salaries go to building agencies which set a strictness of regulation that reflects the balance of interest group strengths in the locality.

The ‘gatekeeper’ hypothesis and the international transfer of scientific knowledge

This paper deals with the question of problem selection by scientists, in particular scientists residing and working in small national communities. Do such scientists take their cue from the immediate environment and accept as their task the application of knowledge to national development? Or do they consider themselves members of an international scientific fraternity and adopt the problems defined by the leading luminaries at the major centers? The data presented here were collected in one small scientific community, that of the Republic of Ireland. We conclude that an Irish scientist's integration with foreign colleagues (and presumably his willingness to accept their influence) is a function of (1) the relevance of his/her work to the Irish economy i.e. the existence or non-existence of local rewards, and (2) the state of paradigm development in his/her field. If work on national problems is rewarded, then this alone can be expected to wean a scientist away from excessive reliance on foreign colleagues.