NetLogo User Community Models
by Guido Fioretti (Submitted: 06/02/2009)
WHAT IS IT?
The Garbage Can Model of Organizational Choice (Cohen, March and Olsen 1972) is the most famous model of organizational decision-making. With this code, we reproduced the original Garbage Can model in an agent-based setting (Fioretti and Lomi 2008a, 2008b).
In the Garbage Can model, decision is made when the members of an organization apply a solution to an opportunity for making a choice. Note that solutions exist before problems, and that decisions can be made even without solving any problem. Eventually, a problem may be there to be solved, but this is not necessarily the case.
In this decision process, choice opportunities take the role of "garbage cans" where solutions and problems are dumped. Hence the name of the model.
The GCM can be seen as a sort of chemical reactor where participants (decision-makers), opportunities, solutions and problems have been dumped. Through random meetings of these elements, decisions are made. We interpreted the GCM as an agent-based model where participants, opportunities, solutions and problems are four classes of agents.
Participants are denoted by yellow men. Opportunities are denoted by orange squares. Solutions are denoted by red circles. Problems are denoted by violet triangles.
An organization starts its life with a given number of participants, choice opportunities, solutions and problems. These initial values are set by means of the parameters "initial-number-of-participants", "initial-number-of-opportunities", initial-number-of-solutions" and "initial-number-of-problems", respectively. These parameters may take any integer value greater or equal than zero.
Eventually, an organization may want to attract or expel participants, opportunities, solutions and problems in the course of its life. The corresponding net input-output flows are set by means of the parameters "net-flow-of-participants", "net-flow-of-opportunities", "net-flow-of-solutions" and "net-flow-of-problems", respectively. If these parameters are greater than zero, the organization attracts participants, choice opportunities, solutions or problems. If these parameters are less than zero, the organization expels participants, choice opportunities, solutions or problems into the environment. If these parameters are zero, the flows between the organization and its environment compensate one another. Since, as we shall see, an organization may destroy participants, opportunities, solutions or problems in order to make a decision, the net flows of these three classes of agents may be required to be positive in order for an organization to be viable in the long run.
The parameters "net-flow-of-participants", "net-flow-of-opportunities", "net-flow-of-solutions" and "net-flow-of-problems" take values in the [-5, 5] interval in decimal steps. If they take integer values, they denote the number of participants, opportunities, solutions or problems that enter or exit the organization at each step. If they have a decimal component, this denotes an additional number of agents that enter or exit the organization every 10 steps. For instance, "net-flow-of-participants = 3.4" means that in general 3 participants enter the organization at every time step, but 7 participants enter the organization every ten time steps.
The switches "stop-flow-par-at", "stop-flow-opp-at", "stop-flow-sol-at" and "stop-flow-pro-at" enable the user to stop these flows after a selected number of simulation steps. These parameters are ineffective when they are set to zero.
Participants, opportunities, solutions and problems may be required to exit the organization after decision-making. This option is chosen by means of the parameters "participants-exit?", "opportunities-exit?", "solutions-exit?" and "problems-exit?", respectively.
If the switch "participants-exit?" is ON, on each patch where a decision is made, all participants exit. This reflects the idea that all of them are involved in decision-making, be it resolution or overflight.
If the switch "opportunities-exit?" is ON, on each patch where a decision is made only one opportunity exits. This is chosen at random among the opportunities on the patch. The underlying idea is that decision-making requires only one choice opportunity.
If the switch "solutions-exit?" is ON, on each patch where a decision is made only one solution exits. The underlying idea is that even when decision-making involves many participants and solves several problems, only one solution is applied. If the decision is made by resolution, the chosen solution is the one with highest efficiency. If the decision is made by oversight, a solution is selected at random among those on the patch.
Finally, the switch "problems-exit?" is only effective when decisions are made by resolution. If it is ON, all problems on patches where a decision is made, exit.
Participants, choice opportunities, solutions and problems walk freely on the model space if both "decision-structure = 0" and "access-structure = 0". If "decision-structure = 1" participants and choice opportunities are ranked in order of importance and participants are only allowed to move on to opportunities of less or equal importance. If "decision-structure = 2" participants are only allowed to move on to opportunities of equal importance. If "access-structure = 1" problems and opportunities are ranked in order of importance and problems are only allowed to move on to opportunities of less or equal importance. If "access-structure = 2" problems are only allowed to move on to opportunities of equal importance.
Participants are characterized by their ability in solving
Energy and efficiency take values within a range specified by the sliders "min-energy-par" and "max-energy-par", "min-efficiency" and "max-efficiency", "min-energy-pro" and "max-energy-pro", respectively. Energy and efficiency can be assigned according to one of the three following criteria that can be chosen by means of the sliders "dist-energy-par", "dist-efficiency" and "dist-energy-pro", respectively:
A problem is solved when a participant has enough personal energy and a sufficiently efficient solution such that their product is greater or equal to the energy of the problem. Intuitively, this process is analogous to that of a machine that transforms potential energy into kinetic energy. For instance, a car extracts potential energy from gasoline (the "energy" of the decision-maker) wasting a certain fraction of it (the "efficiency" of the solution) in order to cover a distance (the "energy" of the problem). When problems are solved, decision is made "by resolution".
However, a key feature of the Garbage Can model is that a lot of decision-making may not solve any problem at all. For instance, choice opportunities may be used as showrooms in corporate politics rather than as occasions to solve problems. In these cases, decision is made "by oversight". In the model, decision-making takes place when participants, choice opportunities, solutions and eventually problems happen to be on the same patch.
Decision-making by resolution takes place when at least one participant, at least one choice opportunity, at least one solution, at least one problem are on the same patch and the sum of the energies of the participants on the patch, multiplied by the efficiency of the most efficient solution on the patch, is greater or equal to the sum of the energies of the problems on the patch. Most often, decision-making by resolution occurs when just one participant, one choice opportunity, one solution and one problem happen to be on the same patch and the energy of the participant, multiplied by the efficiency of the solution that she is using, is greater or equal to the energy of the problem. Decision-making by resolution is marked by a green square.
Decision-making by oversight takes place when at least one participant, at least one choice opportunity and at least one solution are on the same patch. No problem must be there to be solved, and no energy balance is required. Thus, decision-making by oversight occurs most often when a participant, a solution and a choice opportunity happen to be on the same patch. Decision-making by oversight is marked by a sky-blue square.
If problems are on a patch with participants, solutions and choice opportunities, but the energy of the participants and the efficiency of the solutions is not sufficient to solve the problems, all agents on the patch are blocked and no decision is made. Blocked decision processes are marked by a white square.
However, if there is more than one choice opportunity on the patch, or if an additional opportunity just walks on a patch where decision is blocked, then the least important opportunity takes away the problem with the highest energy and walks together with it until they find a participant and a solution that are able to solve the problem. This is called a "flight" from a problem. During a flight, the color of the square where a decision process is blocked is set back to black. The opportunity and its problem move away together, and they are marked by a brown square.
Once the most difficult problem left, the blocked participants may be able to make a decision, either by resolution or by oversight. Thus, flight is very important in organizational decision-making. It represents all situations where a difficult problem is postponed in order to face other issues.
The four monitors "participants", "opportunities", "solutions" and "problems" report the cumulative number of the corresponding agents. They reflect both net flows and deaths.
The monitor "Decisions by Resolution" reports the cumulative number of decisions that are made by resolution. The monitor "Decisions by Oversight" reports the cumulative number of decisions that are made by oversight. The graph reports the number of decisions by resolution and by oversight that are made at each tick.
The monitor "Total Flights" reports the cumulative number of times that a problem flied away with another opportunity. Sometimes, a flight enables decision-making for the remaining agents. The monitor "Flights that cause Resolutions" reports the number of flights that enabled decision-making by resolution. The monitor "Flights that cause Oversights" reports the number of flights that enabled decision-making by oversight. Thus, the sum of "Flights that cause Resolutions" and "Flights that cause Oversights" is less or equal to "Total Flights". Note that, since decision is made one tick after a flight, the decisions caused by a flight may not reflect in the monitors "Decisions by Resolution" and "Decisions by Oversight" if the simulation is stopped just after a flight. Also note that, if "decision-structure" and "access-structure" are not set to 0, it may happen that an opportunity hinders the way of an opportunity that wants to fly. In this (very unlikely) case, the monitors "Flights that cause Resolutions" and "Flights that cause oversights" do not actually cause flights or oversights, respectively.
The Garbage Can model comes with the following indicators:
Furthermore, the fraction of decisions made by oversight and resolution is disaggregated with respect to the importance of choice opportunities. If the decision structure is either hierarchical or specialized participants have a degree of importance, meaning that only certain participants are allowed to make use of certain opportunities. Likewise, if the access structure is either hierarchical or specialized problems have a degree of importance, meaning that only certain problems gain access to certain choice opportunities. In both cases, choice opportunities need to be ordered by their importance as well. Thus, the importance of choice opportunities can be used to rank decision-making both when the decision structure is hierarchical or specialized and when the access structure is hierarchical or specialized.
The degree of importance of opportunities, just like the degree of importance of participants and problems, is indicated by their identification number. The lowest numbered opportunities are the most important ones. Since the number of opportunities may vary during a simulation, classes of importance are defined in percent terms with respect to the current population of opportunities. Four classes are defined, entailing one fourth of all opportunities each.
The indicators that appear at the bottom right of the interface have the following meaning:
These indicators are not available if opportunities do not exit after decision-making or both the decision structure and the access structure are non segmented. Even if these conditions are satisfied, they remain unavailable until at least one decision is made.
Additional information on decision-making can be obtained by switching on "show-details?". If "show-details?" is ON, each time a decision is made a line is printed on the Command Center. Each line says whether the decision was made by oversight or resolution, which agents it involved according to current numbering, which was their identification number at the time they were created as well as the time step when they were created. If the simulation stops because the step specified by "stops-at" is reached, information on the current and original identification numbers of surviving agents is written on the Command Center. If this option is chosen it is advisable to enlarge the Command Center.
Finally, the following graphs are available:
HOW TO USE IT
The population of participants, opportunities, solutions and problems should collectively not exceed 1000 units, unless the simulation space is enlarged. Conversely, at least a few tens of each breed are needed in order to obtain interesting outcomes.
The energy of problems should not be too different to the energy of participants multiplied by the efficiency of solutions. Depending on energy configuration, maximum differences of one or two units are recommended.
With hierarchical or specialized decision structures or access structures, the number of decisions and particularly the number of decisions by resolution tends to be small. Thus, it is advisable to run the model with many opportunities, many problems, sufficiently high values of participant energy, sufficiently high values of solution efficiency and sufficiently low values of problem energy.
THINGS TO NOTICE
The most important and most robust conclusion of the Garbage Can model is that most decisions are made by oversight, few problems are solved and flight is widespread. A large number of flights do not cause decision-making. Most, but not all configurations are such that flights that cause decisions by oversight are many more than the flights causing decisions by resolution.
The energy of problems affects all the above indicators, all of which measure some aspect of the difficulty of decision-making. Energy distribution influences the outcome of decision-making as well.
Another interesting conclusion is that the most important opportunities are less likely to solve problems than the least important opportunities. This can be seen by experimenting with hierarchical decision and access structures.
THINGS TO TRY
The default values of parameters have been chosen in order to resemble the original model by Cohen, March and Olsen as closely as possible. The user is invited to move away fron default values exploring alternative combinations.
The user can try different combinations of:
EXTENDING THE MODEL
This is an agent-based version of the original Garbage Can model. A common criticism of this model is that organizational structures (decision structure, access structure) are not endogenous, which is weird for a model of organizational decision-making. In subsequent variations of the original model, we shall overcome this shortcoming.
This version runs on NetLogo 4.0.4. The previous version, also available under the rubric "User Community Models", was developed on NetLogo 2.1.0.
With respect to the previous version, the following changes have been made:
The following bugs have been fixed:
CREDITS AND REFERENCES
Cohen M.D., March J.G. and Olsen J.P. (1972) A Garbage Can Model of Organizational Choice. Administrative Science Quarterly, 17 (1): 1-25.
Fioretti G. and Lomi A. (2008a) The Garbage Can Model of Organizational Choice: An agent-based reconstruction. Simulation Modelling Practice and Theory, 16 (): 192-217.
Fioretti G. and Lomi A. (2008b) An Agent-Based Representation of the Garbage Can Model of Organizational Choice. Journal of Artificial Societies and Social Simulation, 11.
This model was built by:
(back to the NetLogo User Community Models)