1: \begin{abstract}

2: A random-batch method for multi-species interacting particle systems is proposed,

3: extending the method of S.~Jin, L.~Li, and J.-G.~Liu

4: [{\em J. Comput. Phys.} 400 (2020), 108877]. The idea

5: of the algorithmus is to randomly divide, at each time step, 

6: the ensemble of particles into small batches

7: and then to evolve the interaction of each particle within the batches until

8: the next time step. This

9: reduces the computational cost by one order of magnitude, while keeping a certain

10: accuracy. It is proved that the $L^2$ error of the error process behaves 

11: like the square root

12: of the time step size, uniformly in time, thus providing the convergence of the scheme.

13: The numerical efficiency is tested for some examples, and

14: numerical simulations of the opinion dynamics in a hierarchical company, 

15: consisting of workers, managers, and CEOs, are presented.

16: \end{abstract}

17: