2.5 Implementación del sistema de visión artificial
2.5.2.2 Preprocesamiento
From a modeller’s point of view we need an energy concept that fits into an explanatory framework of social interactions. Collins’s concept has its place in a whole micro-sociological theory of agents and interaction situations. Whether a form
of emotional charge or the feeling of group solidarity it is generated by interactions, and determines what interactions may occur in future, via our expectations for those interactions. Ryan and Deci’s work in self-determination theory relates to future activities the outcome of social interaction, via the notion of intrinsic motivation levels. But these activities are not themselves social interactions. In Collins the circle is complete – or the chains of interaction rituals formed.
Quinn has presented one theoretical model that integrates both sources. Using terms derived from Ryan & Deci, “autonomy”, “belongingness” and “competence” are the mechanisms by which energy in agents is generated from “social connections” (Quinn, 2007, p.84). But his model requires an understanding of what have been termed “high quality connections” (Dutton & Heaphy, 2003). If HQCs are “connections made between two people that are marked by vitality, mutuality and positive regard”, it would seem this model is restricted to positive interactions. We find, therefore, Collins’s theory to be wider in scope and have chosen the vocabulary of interaction rituals in this thesis.
There seems support for relating autonomy, belongingness and competence to Collins’s emotional energy, however.
The closest correlate of autonomy would be what happens in “power rituals” (Collins, 1990, p.35). The feeling of being controlled by others in self-determination theory becomes the experience of being dominated. Order-givers gain EE. Order-takers lose it. (Collins, 1990, p.39)
In “status rituals”, on the other hand, the “fundamental feature… is belonging or not belonging” (Collins, 1990, p.37, our italics). “Successful enactment of group membership raises EE, experiencing marginality or exclusion lowers it.” (Collins, 1990, p.39) “High degrees of emotional arousal are created especially by IRs that include an element of conflict against outsiders.” (Collins, 1981, p.1002)
Competence may be interpreted as the effective execution of both types of ritual, and perhaps also as an awareness of the material resources that have been presented and expended in order to participate in the interaction. Ritual performances that cost more than was expected suggest poor competence on the part of the participant.
Thus we find the various forms of payoff from interaction rituals provide suitable candidates for feelings of autonomy, belongingness and competence – the ABC of intrinsic motivation. This micro-sociological theory of agents is what we try to represent in our models.
As already suggested by our discussion of using the ACM, a connection can be drawn between the matching of cultural traits in the ACM and the concepts of attention focus and group solidarity in IR theory. An agent who can focus on the same objects as another agent feels a sense of solidarity with that other, and the objects become symbols of their common group membership. The awareness of belonging to something common is the first component of energy we model. Belongingness scores are based upon the proportion of cultural features for which two agents matched in traits. As a proportion this ranges from 0 to 1. In line with the ACM, however, a
failure to match in the first feature compared results in both agents exiting the IR event with Belongingness payoff = 0.
Our survey of models of “work performance” (section 3.5) failed to identify a convenient representation of tasks for agents to be “competent” at. As Kennedy’s use of the ACM has indicated (Kennedy, 1998) cultural traits can easily be mapped to fitness values. This has the virtue of being easy to program and debug, and quick to calculate during a simulation run. In view of the relative simplicity of this, and in the absence of a specific scenario to model, we represent the sense of competence using a linear mapping of trait ID numbers to fitness values. The traits an agent presents in an IR event are turned into fitness numbers and averaged across features to produce a single number between 0 and 1. More details of this process are covered in Chapter 8. IR events where agents exit after the first feature comparison has failed are taken to generate no competence score. By adopting a crude, abstract fitness function, we free ourselves up to focus on other aspects of the modelling. Should a more convincing, benchmark representation of work that is easy to calculate emerge in the literature, we will at least be ready with models to insert them into. (We shall return to the question of more demanding benchmark fitness functions for competence in 12.5.3 and F.7.)
To capture the idea of some agents dominating an IR more than others we use their expected gain values to decide who has presented a trait first for each feature. Agents with more autonomy will not wait for the other to lead but will take the initiative. Agents, who are being beaten in the race to make each contribution to the ritual, and feel themselves becoming passive spectators in someone else’s performance, will not experience the IR as energising. We calculate an Autonomy score for each participant
as the proportion of features for which that agent was first to supply a trait. Again this varies from 0 to 1, and IR events that finish prematurely score 0.
Table 2 Energy Payoff calculations explained Concept Definition
Failed IR event A failure to match traits in the first feature compared results in a failed IR event. All participants exit with payoffs of 0 and neither cultural capital nor energy levels can be updated.
Otherwise, payoffs are based on:
A. Autonomy Proportion of cultural features for which agent was first to supply the trait.
B. Belongingness Proportion of cultural features for which participating agents matched trait.
C. Competence Mean for all features of trait-based fitness values. In the simplest case, trait fitness is scaled linearly, with trait “A” scoring 1 and the qth trait scoring 0.
Our three components of payoff are listed in Table 2. Each component is a number between 0 and 1. We multiply them together to form a single figure – the energy payoff the agent receives from that IR event. Thus we simplify the programmer’s task to dealing with payoffs only in this [0, 1] range, rather than utility values of unbounded range. Standardised, bounded payoffs may also help the model user and experimenter through telling them what not to expect. In addition, in the next section we will apply exponents to each payoff component. This provides a means of weighting the components, so that, for example, Belongingness could be treated as
being more important than Competence. We make no claims to these being the definitive or best representations of the autonomy, belongingness and competence concepts found in the literature. We offer them now as first attempts, and welcome any alternative suggestions that come with arguments in their favour.
The three claims we wish to test are given in terms of the effects had by “energisers” and “de-energisers”. Having established a concept of energy we must now define what these two types of agent are.