The Second Agent: The Competent Actor CA

The Competent Actor is a collective agent, made by a set of Individual Competent Actors (ICAs). This agent is defined as competent because it has the skills and competences necessary to satisfy CE’s requests, as well as it is able to learn both from the environment and by collaborating with the other ICAs. Additionally, the Competent Actor has been defined as a collective agent because, according to our conceptual model, an ICA can represent different categories of actors (e.g. exploiter, explorer or catalyst—see Table 2) on the basis of the values of different parameters which are explained in details in the following sections. Each ICA is endowed with a set of Frames, that is a set of ternary strings {-1, 1,0} of length l, which represent the set of agent’s beliefs (or knowledge or capability) about the corresponding dimension of the ER. The value 0 indicates the lack of the specific competence on that dimension. Each agent is also endowed with a budget distributed among the frames. The Agent ICA makes four actions:

Action 2: The generation of the initial frame

Action 3: The generation of Individual Interpretations (exploration capability) Action 4: The generation of interactions (interaction capability)

Action 5: The generation of Collective Interpretations (exploitation capability)

Action 2: The Generation of Initial Frame of ICAs

The initial frame of each agent is generated on the basis of two parameters: the scope s and the competence c. The scope s is the probability that the agent ICA has the complete knowledge to produce the required interpretation of an ER. In other words, if the probability is 0 then any value of the frame is 0, which means that the ICA has no knowledge at all. If s = 1 each value of the string frame is -1 or 1. Finally, when s = 0.5, each agent has a probability that half of the value of the frame are randomly set equal to 0. Therefore, the scope represents a measurement of the specialization of an ICA.

The competence c is the probability that an element of the frame matches the corresponding element of the ER string. If c = 1, the agent ICA is endowed with a proper frame; vice-versa, if c = 0 the endowed frame is totally wrong. The generation of the frame of each agent depends on both the probabilities: F = f(s, c). More specifically, the lower the value of such probabilities, the lesser the probability to produce the proper output required by the CE.

Action 3: The Generation of Individual Interpretations II

The first task of each ICA is to develop an Exploration activity—that is to interpret an Innovation Opportunity on the basis of its frames (the strings of scope s < l, and competence c) in order to produce an Individual Interpretation. Each agent ICA has a probability p to modify its current Frame in order to “catch” an Innovation Opportunity IO. More specifically, each element could be randomly modified in order to learn from the CE according to the probability p. Clearly, when the element of the Frame is equal to 0, it cannot be modified. The modified Frame is the Individual Interpretation. The modified Frame is memorized by the agent. The exploration activity has a cost.

Action 4: The Generation of Interactions

Each agent ICA moves within the space of action. The agent’s movements are guided by the structure of the space of action, by behavioural rules and by randomness. In the CARIS model the space of action is unstructured—namely, we do not have any predefined network or grid. The space of action has been modelled as the surface of a torus, where agents move in a random way. If two agents are in the same time in the same place then they can decide whether to interact in agreement with the following Action 5. The interaction activity has a cost.

Action 5: The Generation of Collective Interpretations

The second task of ICAs is to develop the exploitation activity. In order to carry out this task, each ICA must choose the most suitable partners in order to combine its Individual Interpretation (II) with that of other agents and create a Collective Interpretation CI fitted with the Innovation Opportunity. This activity is guided by:

  • • The Cooperation propensity (T) of each agent
  • • The value of competence c of possible partners
  • • The Hamming Hij distance between two Individual Interpretations (II)

The three parameters are combined in the following formula:

where Lij is the probability that the agent i decides to cooperate with the agent j. If Lij is positive and also Lji is positive then agents i and j will cooperate. In other words, both agents i and j should evaluate positively the benefits of cooperation.

The cooperation model of Formula (1) is a modified version of the interaction model of Cowan and Jonard (2009). We assume that forming a partnership has a probability of success strictly related to the optimal overlap of knowledge stocks of possible partners. The probability of a successful cooperation increases and then decreases with the overlap of the Individual Interpretations. The overlap is measured by the complement to 1 of the normalized Hamming distance between the Individual Interpretations IIji of agents i and j, namely (1 — Hy). The peak occurs when the overlap is equal to 0.5. The increasing and decreasing probability of cooperation is modelled by Hy x (1 — Hy). The second factor influencing the probability of cooperation is the reputation of the potential partner, which is measured by its level of competence cy. The third factor influencing the Cooperation Propensity is Ti, which captures both the propensity to cooperation of the individual actor, and the boundary conditions (culture, incentives) that influence the cooperation. The number 4 is a scaling factor, allowing Ly to vary between 0 and 1.

The results of Action 4 are multiple collaborations, that form a set of network of individual agents which combine their Individual Interpretations (namely, knowledge or capabilities) to produce a Collective Interpretation submitted to the Competitive Environment for evaluation and reward. Of course, the Collective Interpretation is embodied in a product or service that can be delivered to market. The exploitation activity has a cost. Now we are ready to define the last actions developed by the Competitive Environment.

Action 6: Evaluation

Each Collective Interpretation CI is evaluated by the CE on the basis of an Acceptance Threshold. If the Collective Interpretation overcomes such threshold, it is accepted and rewarded.

Action 7: Rewarding

The reward obtained by a Collective Interpretation is distributed among the agents ICAs that contributed to it, according to the contribution they gave in terms of competences to the successful interpretation. At the beginning of simulation each ICA is endowed with a budget distributed among the Frames populating the individual memory of agents (in the first step of the simulation only one Frame is contained in the individual memory of each ICA). The budget associated to Frames will be decreased according to the costs sustained for exploration, interaction and for exploitation activities. The reward for a Frame of an ICA involved in a successful Collective Interpretation will increase the budget of the Competent Actor. Only Frames with a positive budget will survive and the number of surviving Frames for each agent is a proxy of the capability to learn and innovate. At each time step the budget available for each ICA is the sum of budgets associated to its Frames. When the total budget of an ICA becomes equal to 0 the ICA dies and disappears. The difference between the agents’ total budget at the end of simulation and that at the beginning is a measure of the success of RIS. It is a proxy of the capacity of RIS of creating value.

The Netlogo CARIS model

Fig. 4 The Netlogo CARIS model

This meta-model has been implemented on the Netlogo platform. Figure 4 represents the interface of Netlogo model (the model is available on request).

 
Source
< Prev   CONTENTS   Source   Next >