A difficult problems solved by social insects and many other collective systems is task allocation: what set of behavioral rules for choosing their job or function, at the level of individuals, produces an adaptive group-level outcome? Note that ‘task allocation’ here refers to the process of allocating workers to tasks, or equivalently, of allocating tasks to workers.
General insights: benefits of division of labor
This is a central question at each major evolutionary transition, and thus relevant in many systems from biology (e.g. how are embryonal cells allocated to particular specialized functions) to engineering (e.g. task allocation in a computing cluster) to human organizations. It is often assumed that more sophisticated, larger, and more complex systems employ more specialized individuals, generating a higher degree of division of labor that is beneficial. We’ve shown however that benefits of specialization in complex systems may be subtle, and depend on specific social and environmental conditions (Dornhaus 2008; Dornhaus 2012; Charbonneau and Dornhaus in prep). For example, specialists do not perform ‘their’ tasks better in an ant (Dornhaus 2008) and a bumble bee (Dornhaus in prep), and larger colonies do not always display more division of labor (Dornhaus et al. 2009; Skeate and Dornhaus in prep) and rev. in (Dornhaus et al. 2012). However, even if all individuals have equal ‘skills’, small costs of switching from one activity to another are sufficient to lead to the evolution of division of labor (Goldsby et al. 2012).
Mechanisms of task allocation
We are also interested in discovering the specific mechanisms of task allocation used by insects, and evaluating them in the broader context of what conditions each strategy is optimized for – which will lead to better algorithms in engineering/computing.
Example: Bumble bees, spatial sorting, and how to generate size polymorphism
In one example, we discovered that the size polymorphism in worker bumble bees is generated by spatial fidelity of nurse bees to the center of the nest, leading to heterogeneous feeding of developing brood and subsequent variation among workers in body size (Couvillon and Dornhaus 2009; Jandt and Dornhaus 2009). We also show that while the jobs performed by workers are predicted by body size (Jandt and Dornhaus 2009; Jandt et al. 2009); this had been known for foraging), there was no evidence that smaller workers performed any of these jobs better (Couvillon et al. 2010a; Dornhaus in prep). Instead, smaller workers appear to be more resilient to starvation, independent of their role in the colony (Couvillon and Dornhaus 2010; Couvillon et al. 2011).
This introduces novel hypotheses for the benefits of variation in groups, namely that groups face trade-offs between producing efficient or robust individuals, or that groups balance production of high-quality vs. cheap workers (Couvillon et al. 2010b; Jandt and Dornhaus 2014; Charbonneau and Dornhaus in prep; Dornhaus in prep; Duong et al. in prep; Rivera and Dornhaus in prep-a).
Example: Lazy ants
Another interesting phenomenon in social insect task allocation is the fact that many ‘workers’ don’t appear to do anything at all, even in the field (Jandt and Dornhaus 2009; Charbonneau et al. 2015; Charbonneau et al. in prep). While this may be driven in part by selfish interests, selfishness can at most explain a small fraction of observed inactivity (Jandt and Dornhaus 2011; Hillis et al. in prep). Nor do all inactive workers appear to be reserves for defense (Jandt et al. 2012) or for the case of worker loss (Pinter-Wollman et al. 2012; Dornhaus and Walton in prep). Inactivity may however be a side-effect of an imperfect mechanism for generating individual variation for the purpose of effective task allocation (Pinter-Wollman et al. 2012), which is a mathematically difficult problem (Cornejo et al. 2014).
Broadly, we believe that insights about the benefits and costs of different mechanisms for task allocation can become a successful example of generalizing insights on complex systems across domains; for example, between biology and theoretical computer science (Cornejo et al. 2014; Dornhaus and Feinerman in prep) or about allocation of defense resources across modular organisms (Powell and Dornhaus 2013; Powell et al. in prep).