Multi-agent projective simulation: A starting point