Complex Agent Networks explaining the HIV epidemic among homosexual men in Amsterdam

In particular, whether such methods also apply to networks of small size, and thus to many real-world biological or epidemiological applications, is still an open question [1]. Many networks of relevance to epidemiology may be of relatively small size, among which social contact networks on sexually transmitted diseases for a small group (e.g. men who have sex with men, namely MSM, within a city or town) are representative.

Modelling the HIV epidemic is difficult because the true incidence1 of the HIV/AIDS-epidemic is uncertain since many people may be unaware of their infection. Secondly, HIV progression has a very long asymptomatic period which makes studies of the actual infection spreading a very complicated task [2]. Finally, the various routes of infection and the inhomogeneity of the involved population pose additional challenges to understanding the underlying knowledge of HIV epidemics.

Multi-agent systems (MAS) and complex networks (CN) are often used separately to model and simulate epidemics; however, whether existing models, which typically focus on large populations, can address epidemics among small groups (∼ 10 3 -10 4 ) are still questionable and seemingly need further validation. In terms of agent-based modelling, Teweldemedhin developed an agent-based bottom-up modelling approach for estimating and predicting theX spread of the HIV in a given population [3]; Xuan developed an extended Cellular Automata simulation model to study the dynamical behavior of HIV/AIDS transmission by incorporating heterogeneity into agents’ behavior [4]. In terms of complex network modelling, Bai discussed a sexual network spreading model for HIV epidemics [5]; Sloot proposed a new way to model HIV infection spreading through the use of dynamic complex networks, with the time evolution of the network vertices modelled by a Markov process [2].

In this paper we present the CAN approach for simulating epidemics in small networks to great detail. The CAN is a hybrid approach in which multiagent systems and complex networks are the basic methods of modelling epidemics on individual and population scales respectively. Using the CAN approach, we simulate a relatively detailed model of the HIV endemic among MSM in Amsterdam. This model includes a distinction between steady and casual relationships, which is regarded as an important aspect by others [e.g. 6]. We compare the results to the Amsterdam Cohort Study (ACS) historical data.

This paper is organized as follows. In Section 2 we review MASs and complex networks and introduce the CAN approach. We describe the HIV epidemic model in detail in Section 3 and simulation implementation in Section 4. In Section 5 we discuss a case study and its simulation results.

The CAN approach takes the MAS and CN as basic methods for modelling and simulating epidemics on individual and population scales. Agents contain specified personal information on an individual scale, whilst complex networks emphasize the relationship dynamics among the agents on a population scale. We review both concepts and the combination of them in this section.

A MAS is a system composed of multiple interacting intelligent agents, and can manifest self-organization and complex behavior even when the individual strategies of all the agents are simple. Heylighen defined selforganization as “the spontaneous emergence of global coherence out of the local interactions between initially independent components” [7].

The agents in a multi-agent system have several important characteristics [8]: Autonomy: the agents are at least partially autonomous. Local views: no agent has a full global view of the system, or the system is too complex for an agent to make practical use of such knowledge. Decentralization: there is no controlling agent (or the system is effectively reduced to a reductionistic system) [9].

Applied to HIV epidemics, a MAS represent a specified human community and viruses propagate along social contacts. Each individual has his own progression of infection and simple rules to choose partners. Each agent only knows the information about himself and his partners instead of epidemiological statistics. There is no dominating agent in the community to control the spread of viruses.

The study of complex networks is inspired largely by the empirical study of real-world networks such as computer networks and social networks. A network is a set of items, which we call vertices, with connections between them, called edges [10]. In the context of network theory, a

…(Full text truncated)…