TMS Behavior Modeling, Incentive and System Design against Piracy and Collusion

Teaching material sharing (TMS) websites grow rapidly in Taiwan due to one-outline-multiple-texts policy under nine-year compulsory curriculum. The prevalence of personal computers and peer-to-peer (P2P) networks provide convenient way to share teaching resources and experiences on TMS system. Teachers can enhance their teaching abilities and provide high-quality education to increase the competitiveness of our next generation. However, no effective incentives to encourage sharing activities are proposed.
This thesis focuses on TMS over TANet among elementary teachers. Lai (2005) designed a hybrid P2P TMS system where teachers are peers and bureaus of education play the role of super-peer. We consider four issues that are critical to the success of TMS but not addressed in Lai’s research: (i) Reinforcement of Lai’s TMS system against piracy and collusion; (ii) Design of a reputation system to evaluate teachers’ contribution to TMS community; (iii) Behavior modeling of membership growth and TM upload; (iv) Evaluation of different reward policies which influence content quality and quantity on TMS system.
We first design and add Digital Rights Management (DRM) to Lai’s system against piracy. By using symmetric encryption and integrating license management function of DRM solution, pirated files without acquiring legal licenses form bureau of education can not be decrypted for access. Also super-peers store all download transaction records for giving credits to content providers. We provide a piracy-free sharing environment for content creators to protect intellectual property rights and thus increase teacher’s willingness to share.
We then proposed a synthetic reputation for bureau of education to evaluate teachers’ contribution to the TMS community. The synthetic reputation of a peer is defined as the summation of average score received from other peers, TM downloaded times, review return ratio of the peer, and number of upload TM. Bureau of education can change the coefficient of reputation formula at different stages to induce desired behaviors.
The component of average score received contains credibility factor to reflect reliability of peers’ reviews. We adopted Personalized Similarity Measure (Srivatsa et al. 2005) to reduce the impact of reviews of collusive peers. Collusive peers get low credibility factor because their dishonest feedbacks are not similar to normal peers’. Thus we can alleviate review collusion to improve the accuracy of reputation system.
We model teachers’ willingness to join, leave, probability to upload and define the interactions among these behaviors. We used technology acceptance model (TAM) and finding of important factors such as TM quantity and quality or reward incentive (Lu 2003) to model the S-curve shape of join and upload behavior probability distribution. This model captures essential factors and reasonable behavior pattern which obey economic principle that we believe it is suitable as theoretical model for experiment.
On the basis of behavior model, we evaluate the impacts of two different reward policies, relative ranking and threshold, to allocate bonus to content creators. Under relative ranking policy, high-quality content creators with higher reputation earn more bonuses and share more contents which improve average content quality in system. Under threshold policy, teachers equally share reward and increase content quantity.
We implement license management and credibility factor calculation functions for bureau of education. License upload, license request, content encryption and decryption are developed for teachers. This TMS system provides an experiment platform for further research.