Stochasticity plays an important role in the evolutionary dynamic of cyclic dominance within a finite population. To investigate the stochastic evolution process of the behaviour of bounded rational individuals, we model the Rock-Scissors-Paper (RSP) game as a finite, state dependent Quasi Birth and Death (QBD) process. We assume that bounded rational players can adjust their strategies by imitating the successful strategy according to the payoffs of the last round of the game and then analyse the limiting distribution of the QBD process for the game stochastic evolutionary dynamic. The numerical experiments results are exhibited as pseudo colour ternary heat maps. Comparisons of these diagrams shows that the convergence property of long run equilibrium of the RSP game in populations depends on population size and the parameter of the payoff matrix and noise factor. The long run equilibrium is asymptotically stable, neutrally stable and unstable respectively according to the normalised parameters in the payoff matrix. Moreover, the results show that the distribution probability becomes more concentrated with a larger population size. This indicates that increasing the population size also increases the convergence speed of the stochastic evolution process while simultaneously reducing the influence of the noise factor.
The fact that the water resource governor has to allocate limited water resources to two competing agricultural water users under the administrative system (AS) leads to a principal-agent issue. Hence, this paper constructs a two-stage performance-based allocation mechanism to motivate two competing water users (referred to as the agents) to act in accordance with the interests of the governor (referred to as the principal). This mechanism is about the interaction between the governor and two water users. The governor aims to improve water resources allocation efficiency and balance economic and environmental development, while each agricultural user focuses on the overall amount of water resources they have to operate and would like to ignore environment protection issues. Besides, the total water resources invested into production or environment is two water users’ private information, which is unknown to the governor. In the first stage, the governor allocates water resources between two users according to their previous performances, including production and environmental aspects. Results indicate that the equilibrium state of this mechanism could encourage two water users to focus on production and environment performances simultaneously and to help the governor transfer the pollution cost to two water users under the external of pollution cost, which motivates them to compete for available water resources. This competition between two users will directly affect users’ behaviors. These results could improve water resources allocation efficiency significantly and realize the sustainability of water resources in the agricultural field under the current AS. This perspective could also provide a new insight for the management of agricultural water resources allocation and offer relative decision support to relative governors.
29 March 2024 in IEEE Transactions on Knowledge and Data Engineering
Multi-passage machine reading comprehension (MRC) aims to answer a question by multiple passages. Existing multi-passage MRC approaches have shown that employing passages with and without golden answers (i.e., labeled and unlabeled passages) for model training can improve prediction accuracy. However, when using the unlabeled passages, they either incur the wrong labeling problem or treat the labeled and unlabeled passages equally. In addition, they ignore the original passage information to verify the correctness of the answer. In this paper, we present MLDV-MRC, a novel approach for multi-passage MRC via M ulti-task L earning and D ual V erification. MLDV-MRC adopts the extract-then-select framework, where an extractor is first used to predict answer candidates, then a selector is used to choose the final answer. For the extractor, we adopt multi-task learning with generative adversarial training to train it by using both labeled and unlabeled passages. To train the extractor by backpropagation, we propose a hybrid method which combines boundary-based and content-based extracting methods to produce the answer candidate set and its representation. For the selector, we propose to leverage both the information from answer candidates and original passages to verify the final answer. In particular, we propose a global-local memory-augmented neural network to build the representations of original passages, which fuses the passage-level information and word-level information. The experimental results on three open-domain QA datasets confirm the effectiveness of our approach.
The booming e-commerce economy has incentivized upstream suppliers to encroach on online retailers’ (e-tailers’) operations to directly capture heterogeneous consumers, a phenomenon known as ‘supplier encroachment’. While prior literature suggests that sharing private demand information can mitigate double marginalization, the incumbent e-tailer may have a strong incentive to manipulate this information to maximize its own profit, especially when facing a supplier who is also a competitor in the digital market. To address this issue, this paper investigates the optimal pricing and signaling strategies when a supplier encroaches under asymmetric information. We derive the Perfect Bayesian Nash Equilibrium and obtain several counter-intuitive insights. First, the e-tailer’s strategic information manipulation leads to an upward pricing distortion. Second, the value of demand information shows a sharp decline in the distortion region, and this effect can be so severe that information acquisition yields negative utility for the e-tailer. Third, while the supplier generally benefits from possessing demand information, establishing a direct selling channel is not always a dominant strategy, particularly when channel competition is intense. Furthermore, we find that consumer preferences and channel substitutability significantly alter the intensity of competition and the signaling behavior of the e-tailer. These findings offer practical guidelines for e-commerce participants navigating information sharing and encroachment decisions.
This study examines how inherited local regulatory discretion shapes rule-based delegation and reward-based contracting in China’s Certified Emission Reduction (CCER) program. We develop a principal-agent model with endogenous allowance and offset markets, privately observed regional emissions, and divergent central–local policy preferences. Using equilibrium analysis and mechanism design, we analytically characterize governance arrangements in two institutional settings: Greenfield regions where no policy-specific local regime predates central intervention and Brownfield regions where a legacy regime preserves local discretion. The benchmark balances compliance-cost savings from offsets against reduced internal abatement. In Greenfield regions, the optimal upper-limit rule implements transfer-free interval delegation, preserving local discretion below a central cap and constraining it above the cap. In Brownfield regions, the legacy outside option introduces a type-dependent participation constraint, and policy adjustment requires rewards covering policy compensation and, where needed, information rent for truthful reporting. Information-rent costs shift policy toward local preferences relative to the symmetric-information contracting benchmark. Sufficiently effective rewards induce adjustment across all types; weaker rewards leave higher-emission types at their preferred policies without payment. The results identify the continuation value of local discretion as a source of implementation costs and explain how institutional history shifts policy coordination from rule-based restrictions toward reward-based contracting.