Most studies on real-world multi-robot systems have been performed in controlled laboratory environments, whereas the real world is unpredictable and sometimes hazardous. I have recently suggested that the natural phenomenon of phenotypic plasticity provides a useful bioinspiration framework for making such systems more resilient in field conditions. Phenotypic plasticity occurs when a single genotype produces a range of phenotypes (observable traits) in response to different environ- mental conditions. Consistent individual behavioural differences can result from such plasticity, and have been described as ‘personalities’. At the same time, in social animals, individual heterogeneity is increasingly recognised as functional for the group. We can exploit this functional heterogeneity as engineers trying to design field robot systems, and phenotypic plasticity can provide meaningful diversity ‘for free’, based on the local experience of agents. Personality axes such as bold–shy or social–asocial can be represented as single variables, with the advantage of being transparent and intuitive for human users, and predictable in their effects. For example, in a dangerous environment, robots may become more ‘shy’ and ‘social’ to stay closer together and out of harm’s way.