LEADERSHIP PHENOMENON IN THE AGENT-BASED HERDING MODEL

Aleksejus Kononovicius1, Vygintas Gontis1

1 Institute of Theoretical Physics and Astronomy, Vilnius University, Lithuania

It is often thought that endogenous interactions are behind many big spontaneous events observed in the complex socio-economic systems. Social cooperation, spontaneous emergence, formation of financial bubbles, financial flash-crashes and even mass panic may be caused by the endogenous dynamics and certain general similarities of human psychology [1, 2]. The tendency to rely on others, importance of social interactions, tight coupling suggests that it might be possible to influence the collective behavior of the complex socio-economic systems.

One of the suitable frameworks to test this idea is known as agent-based modeling [3, 4]. In this framework a generalized entity, known as agent, is used in place of real-life interacting parts of the modeled system. While the real-life interactions might be very complex, agents are frequently assumed to interact by following very simple rules. Despite the underlying simplification the complex collective behavior emerges as a result of the interactions between these agents.

In this contribution we consider agent-based herding model proposed by Alan Kirman [5], which explains herding behavior in ant colonies and is also successfully applied to other socio-economic scenarios [6]. Leadership can be introduced into this model rather simply by assuming that some of the agents occupy fixed states and may interact with every other agent in the system. In this case one agent transition probability (from state i to state j) takes the following form:

$$\mu(i \to j) = \left[\sigma_j + \frac{h}{N}X_j + hM_j\right] \Delta t,$$

here $\sigma_j$ is the individual transition rate to state j, h is responsible for herding behavior transitions, $M_j$ is a number of leaders in the j state. Note that $\Delta t$ in the above should be small enough for only one transition to be probable.

Figure 1
Fig. 1. Influence of the leaders on the stationary opinion distribution of the system (squares - M1 = M2 = 0, circles - M1 = 20, M2 = 0, triangles - M1 = 0, M2 = 20). There are 104 agents in the system. Other model parameters: σ1 = σ2 = 2, h = 1.

As you can see in the figure above the leaders, despite being only in a small number, are able to significantly impact the output of a large system. This result suggests some of the possible applications - one may facilitate spread of a new technology or product by donating it to influential individuals (e.g., government, journalists), central banks and market regulators could stop financial bubbles and crashes by initiating trades based on the economic fundamentals rather than plain speculation.


[1] G. Akerlof, J. Shiller, Animal Spirits: How Human Psychology Drives the Economy, and Why It Matters for Global Capitalism (Princeton University Press, 2009).

[2] J. P. Bouchaud, Crises and collective socio-economic phenomena: Simple models and challenges, Journal of Statistical Physics 151, 567-606 (2013).

[3] R. Conte et al., Manifesto of computational social science, European Physics Journal Special Topics 214, 325-346 (2012).

[4] M. Cristelli, L. Pietronero, A. Zaccaria, Critical Overview of Agent-Based Models for Economics, Proceedings of the School of Physics "E. Fermi" CLXXVI, 235 - 282 (2012).

[5] A. P. Kirman. Ants, rationality and recruitment, Quarterly Journal of Economics 108, 137-156 (1993).

[6] A. Kononovicius, V. Daniunas, Agent-based and macroscopic modeling of the complex socio-economic systems, Social Technologies 3, 85-103 (2013).