Abstract
In e-marketplaces where demand significantly exceeds supply, traditional reputation systems often fail. Sellers can maintain high sales despite poor service, and buyers are incentivized to provide dishonest ratings to discourage competition for limited goods. This paper introduces a dual-incentive mechanism that aligns individual incentives with market integrity. By rewarding honest sellers with price premiums and honest buyers with priority access to top-tier products, the system ensures that honesty remains the most profitable strategy for all participants.
Introduction
Standard reputation systems, such as those employed by eBay or Amazon, rely on the assumption that a poor reputation will lead to a decrease in sales. However, in markets characterized by limited inventory—such as peak-season hotel bookings or high-demand product launches—this assumption breaks down. In these scenarios, sellers can often sell out their entire stock regardless of their behavioral history, and buyers may strategically sabotage reputable sellers to reduce competition for scarce items.
The Scarcity Problem: Malicious Behaviors
The authors identify two primary forms of malicious behavior that emerge when goods are scarce:
- Seller Malice: High demand ensures that even low-quality sellers can clear their inventory, rendering traditional reputation "sticks" ineffective.
- Buyer Sabotage: Buyers may intentionally provide negative ratings for high-quality sellers to deter other potential buyers, thereby increasing their own chances of securing limited inventory in the future.
Proposed Dual-Incentive Mechanism
To address these issues, the paper proposes a multi-component system managed by a central marketplace controller:
1. Buyer Honesty via Scoring Rules
The system employs normalized proper scoring rules to evaluate buyer honesty. These mathematical functions are designed such that a buyer maximizes their expected score only by reporting their true belief about a seller's quality. Normalization ensures that scores remain comparable across different seller interactions.
2. Seller Reputation and Pricing
Seller reputation is calculated by aggregating buyer ratings, with higher weights assigned to ratings from buyers with high honesty scores. A quadratic pricing function then allows honest sellers to charge a "price premium." Conversely, sellers whose reputation falls below a critical threshold ($R_0$) face prices so low they cannot cover operational costs, effectively forcing them out of the market.
3. Strategic Allocation
This component serves as the primary incentive for buyers. The system ranks buyers by honesty scores and sellers by reputation, matching the most honest buyers with the most reputable sellers. An "exploration factor" (\eta) is included to allow new or lower-ranked buyers occasional access to top sellers, enabling them to build their honesty scores.
Experimental Results
The mechanism was evaluated through a controlled simulation involving 60 sellers and 600 buyers, with dynamic entry of new participants to test system stability.
Seller Performance
The simulation results confirmed a strong positive correlation between a seller's probability of delivering honestly and their resulting reputation and profit. As shown in Figure 1, honest sellers consistently earned higher profits due to the price premium allowed by the system.
Buyer Performance
Similarly, buyers who provided truthful ratings maintained higher honesty scores, which translated into significantly higher total utility. By being honest, these buyers secured priority access to high-reputation sellers who delivered superior products, as illustrated in Figure 2.
Conclusion
The proposed mechanism successfully addresses the unique challenges of limited-inventory markets by turning honesty into a direct driver of profit and utility. While the current model assumes a centralized authority and does not explicitly account for complex collusion (e.g., Sybil attacks), it provides a robust framework for promoting integrity in competitive e-marketplaces. Future work will focus on enhancing the reputation model's resilience against strategic manipulation and testing the mechanism with real-world human behavior.
Summary updated