A simple, general and robust trust agent to help elderly select online services

In the age of the Internet, the elderly population may find it desirable to use online services. Unfortunately, many elderly users are not technically savvy and are vulnerable to malicious online service providers. Computational trust management systems may be a viable way to protect them, but the presence of collusion that corrupts the sources of information based on which many existing trust models make their recommendations poses a serious threat. In this paper, we propose a reinforcement learning based trust agent that does not rely on sophisticated computations or elaborate infrastructural support in the environment to work. Simulation results show that the proposed trust agent can significantly outperform existing models against collusion.