Chapter 3 How might we model energy?
3.5 Models of Work Performance
In the previous section we gave examples of models in which groups and norms emerged from processes involving social interactions. But not all influences on agents need be social. In this section we list some ways in which non-social context has been provided, relative to which a value or performance of the modelled system can be defined.
If interactions can have costs and benefits that are non-social, the trade off between these and social influences becomes worthy of study - not least because the costs and benefits may be of greater interest than the emergence of groups. As we noted in section Chapter 1, part of the motivation for organisation studies in management science is the idea of relating organisational structure or design to some kind of performance measure. In some of the models referred to in section 3.4 a non-social context was included - an interpretation of the culture as knowledge, or a fight for survival for an evolutionary strategy. In contagion models (section 3.3) there is an assumption that the spreading phenomenon has some value extrinsic to the system - negative in the case of disease spread, positive usually in the case of learning - which then motivates a response to the phenomenon in real systems. The network models in section 3.2 are of interest in part because of resemblance to the search for solutions in real social networks (Watts & Strogatz, 1998; Watts et al, 2002; Granovetter, 1973), and the robustness of real-world networks under failure or removal of their parts (Barabasi, 2002).
One approach to defining performance is to define some kind of environment in respect of which agents’ attributes have a value or fitness. For example, in March’s model of organisation learning (1991) “knowledge” is defined by the matching of
agents’ attribute values to an independently determined (and potentially changing) environment. Improvements in this agents-environment correspondence represent learning. For a more sophisticated environment some studies employ Kauffman’s NK fitness landscapes (Kauffman, 1993, chapter 2; Kauffman, 2000, chapter 8). Originally intended as a model of evolution in theoretical biology these have been reinterpreted as modelling intra-organisational design (Levinthal, 1997; Levinthal & Warglien, 1999), inter-organisational strategy, team make-up (Solow et al, 2002) and the search for technological improvements (Kauffman et al, 2000). The appeal of this framework lies in it representing interdependent components (organisations, decision makers, technologies), with simple parameters for controlling the level of interdependency which has been found to “tune” the “ruggedness” of the landscape - in effect, determine how easy the landscape is to search. To date, however, the “decisions” or “attributes” in these models have been limited to binary values (“Yes/No”; “Present/Absent”), making them hard to integrate with some of the cultural models in section 3.4.
Another approach to modelling performance is to set one’s agents a task or tasks. Schmidt (2000) demonstrates the capabilities of the PECS framework with the ADAM model - a search by one agent for food sources, and the Learning Group Model - model of social knowledge acquisition. Carley (1996; Carley & Lin, 1997) uses a classification task - the “Radar task” - to demonstrate the interplay between cognitive capabilities and organisational structure. Anderson (1993) describes how more sophisticated agent designs allow for the solving of tasks requiring decomposition into subtasks.
More complicated models include more than just agents and tasks. The pioneering “Garbage Can Model” of organisational decision making (Cohen et al, 1972) includes decision makers, problems, solutions and decision-making opportunities. In the agent- based version by Fioretti & Lomi (2008) all four are types of agent that must coincide if a problem is to be solved. Curiously, it also includes concepts called “energy”. Participants’ “expended energy” represents their ability as decision makers. Problems’ “required energy” represents their degree of difficulty. Comparisons of the two quantities determine the outcome of a decision making opportunity. Fioretti & Lomi’s revision of the original model, however, appeared too late to influence our own search for energy models, and it involves no groups, cultural capital or social interaction to link it with our energy theories from Chapter 2.
Social networks, on the other hand, do appear in Carley’s work on organisations’ qualities. She models organisations as multi-modal networks linking people, resources, tasks, and knowledge. Organisational performance is defined in network terms - including the lengths of paths linking resources to tasks (Carley & Kamneva, 2004; Carley & Remminga, 2004). The robustness of an organisation can be assessed by Monte Carlo simulations that reiteratively remove nodes and links and record the impact on the performance metrics (Carley et al, 2003).
Models involving this many components require much work to understand their behaviour under different parameter settings (though Hazy & Tivnan (2003) have extended one). To then add the elements of energy theories may require too much. Likewise, the other definitions of performance employed in these agent-based models may distract us from the main task of understanding energy and energisers. So until a
particular benchmark representation of work performance emerges in the literature - for which it would help if researchers placed more computer models and empirical datasets in the public domain - we may have to make do with a simple representation of our own.