Você está na página 1de 7

Mireille Micallef

University of Malta/James Madison University

27-Feb-2013

A Simulation of Littering on the Sliema Promenade


Mireille Micallef Abstract This project recreates a possible scenario of the events that led to the littered state of the Sliema promenade on Wednesday 25 t h October 2012. After tuning the model, variations of it were created to explore what mitigating measures can be applied in order to reduce the amount of litter and which measures are most effective. The measures considered are the introduction of bins at t ake away outlets, the imposition of fines by wardens, the use of signposts to indicate that one should not litter in the area and the presence of sweepers that remove litter and deter people from littering further. The last solution is the most effective i n reducing the amount of litter present since it not only reduces the amount of litter present but discourages the input of further litter into the environment. Introduction The coast of Sliema is a very well-known recreational attraction and commercial centre in Malta the area attracts a variety of people due to the wide range of activities that one may undertake. Some examples of such activities are exercising, swimming, strolling, relaxing, shopping and dining. Recreational and commercial activities very often overlap or lead to one another. In summary, the area is a prolific business and recreational centre and houses a promenade that is fit for recreation as well as a variety of businesses, from clothes shops to travel agents, banks and money exchange bureaus, hotels and food and beverage outlets. Unfortunately, since the area is visited by many people every day, littering can be a problem. Littering reduces the attractiveness of the area and requires resources in order to be cleaned up or prevented. A StarLogo TNG agent-based simulation that takes into account the number of people in the area at a given time, the amount of bins present and the location of take away shops was created in order to try to recreate the conditions that lead to the amount of litter present. To create the model, pedestrian counts at ten locations along the Sliema front during twenty minutes as well as the location and number of people sitting on benches, the location of litter and bins and the location of take away shops were used. Some relationships that were observed from the data, such as the presence of a large amount of litter near a take away shop, were also used to help create the model. This data was collected by me and my SERM colleagues for the purposes of another project in the SERM programme. Once a suitable relationship was established, the model was used to assess the effectiveness of certain measures to reduce the amount of litter present. Such measures include clean ups, signposts to encourage people to keep clean, increasing the amount of bins and imposing fines. The Model Terrain Creating the terrain was the first step in creating the model. This was achieved through the use of the Drawing tab and the Insert Image option in StarLogo TNGs Spaceland window. A base map of Sliema in picture format was used; it was first cropped to a square shape to correspond to the shape of Spaceland. This was done in order to avoid distortion of the map when imported into StarLogo TNG. Any unnecessary data on the base map was cleaned up by erasing and only the promenade 1

Mireille Micallef

University of Malta/James Madison University

27-Feb-2013

was left visible. The image was then imported into StarLogo TNG and stretched to cover all of Spaceland. The Edit Terrain tool was then used to raise the areas adjacent to the promenade to a height of three (by default, patches have a height of zero which means that they are level) so as to prevent the agents from walking outside of the area designated as promenade. Agents Four type of agents were then created; F & B (Food and Beverage) Outlets, Bins, Litter and Homers. Each agent was assigned a different shape so as to make it easy to distinguish between them. This was done by using the Edit Breeds tool. The only type of agent that is capable of littering is Homer. Variables and Procedures Two types of agents required specific variables and procedures to be created. For the Bins, an agent number called Bin Content was created. This number represents the content of the Bin. A procedure, called Bins Initial State, was then created in order to set the Bin Content and traits of the Bins. Initially, Bin Content is set to zero, Bin size is set to 0.15 and Bin colour is set to black for all bins. For the Homers, a shared variable; Likelihood of Littering, was created first. This variable is shown in the Spaceland window and can be adjusted by the user to represent the value at which agents are likely to litter. An agent number called Littering Potential was then created to represent the likelihood that an agent litters. Three agent Booleans were created next; Bench User, Hunger and Packaging. These variables can be set to true or false. The Bench User variable is used to distinguish between pedestrians that can move about and bench users that stay at a fixed point. The Hunger variable represents whether the agent is hungry or not and the Packaging variable indicates whether the agent is in possession of potential litter or not. Three procedures were finally created for Homers; Attributes for Moving Homers, Attributes for Sitting Homers and Movement with Heading. The first two are used in the setup phase and are very similar; they set model skin off, assign a random number from 1 to 100 to the Littering Potential, set the Bench User agent Boolean, set the Hunger and Packaging agent Booleans to false, set the colour of agents and set the size of the agent to 0.2. The difference is that in the Attributes for Moving Homers procedure, the Bench User Boolean is set to false and the colour is set to blue while in the Attributes for Sitting Homers procedure, the Bench User Boolean is set to true and the colour is set to green. These two procedures will lead to the distinction between two types of Homers; pedestrians (which use the Attributes for Moving Homers procedure and can walk about along the promenade) and bench users (which use the Attributes for Sitting Homers Procedure and remain stationary throughout the simulation). The third procedure; Movement with Heading, was created for use in the runtime phase and only by Homers that are not bench users. It commands the agent to move forward 0.5 steps if the height of the patch one step ahead of him is zero. Otherwise, it commands the agent to turn left or right by a random amount of degrees between 90 and 270 and check the height of the patch in front of him again. If the patch in front of him has a height equal to zero the agent can move ahead 0.5 steps. Procedures were created to simplify and shorten the contents of the setup and runtime phases. Setup The first commands in the setup phase were used to reset the clock, show the clock and clear all the agents in Spaceland. Agents were then created. 2

Mireille Micallef

University of Malta/James Madison University

27-Feb-2013

Different numbers of Homers were created in different locations according to the pedestrian counts and number of bench users found in the data. Two types of Homers were created by using the Attributes for Moving Homers or Attributes for Sitting Homers procedure. These attributes represent whether the agent is a pedestrian or a bench user. Although benches are found all along the Sliema front, the number of bench users was grouped and created at a single location that corresponds to the location of the pedestrian count that is closest to them. Therefore at the location of each pedestrian count, two groups of Homers were created; pedestrians and bench users. Three F & B Outlets were created next. These outlets represent take away shops along the Sliema front. The first outlet was created at the coordinates that correspond to the actual location of the take away shop. The second and third take away shops are actually found across the road from the promenade. For simplicity, these outlets were created on the promenade at locations that are across the road from where they actually are. The 47 bins that are found along the promenade were modelled next. Each bin was created at the location that corresponds to the actual location and the Bins Initial State procedure was called to set the size, colour and Bin Content of each bin. Runtime A forever block was first placed in the Runtime page of the StarLogo TNG canvas. In the section that applies to Homer agents the following actions are performed: 1. Approximately 0.5% of pedestrians are randomly selected to have their Hunger agent Boolean set to true and their colour set to yellow. 2. If the agents Hunger is set to false and the agent is a pedestrian, the program checks whether the height of the patch ahead is zero. If it is zero, the agent moves forward 0.5 steps and turns left and right by a random number of degrees between one and ten. If the height of the patch is not equal to zero, the agent is turned left and right by a random amount of degrees between 90 and 270. If he agents Hunger is set to true and the agent is a pedestrian, the program checks the location of the agent and sets its heading towards the closest take away shop. The Movement with Heading procedure is used in this case. 3. Next, if the Likelihood of Littering set by the user is greater than the agents Littering Potential, the agent hatches a randomly coloured litter agent of size 0.05 and the Littering Potential of the pedestrian is set to a random number between 1 and 100. 4. In order to represent the fact that agents arrive and leave the promenade two blocks in the forever block were created. The first block makes approximately 0.5% of the pedestrian agents create another pedestrian agent while the second block makes approximately 0.5% of the pedestrian agents die. Collisions When a pedestrian whose Hunger was set to true collides with an F & B Outlet, the pedestrians Hunger is set to false, the Packaging is set to true, the colour changes to red and the value of Littering Potential is decreased by 5 in order to make the agent more likely to litter. When a pedestrian in possession of Packaging obtained from a collision with a F & B Outlet while hungry collides with a Bin, the Bin Content increases by 1, the agents Packaging is set to false and its colour changes to blue. Results and Analysis The simulation sought to recreate the situation in which approximately 137 agents (137 is the total number of agents that were created initially) create 150 pieces of litter in a 20 minute (1200 3

Mireille Micallef

University of Malta/James Madison University

27-Feb-2013

seconds) time interval. The Likelihood of Littering was varied in the Spaceland window in order to achieve this result. In one instance of the program, an average amount of agents equal to 129 generated 147 pieces of litter when the Likelihood of Littering had been set to 2.5. These values are close to the targeted values and therefore it was decided that with a Likelihood of Littering equal to 2.5 it is possible to recreate the actual situation. However, the model will not always give these results since average number of agents and the number of litter vary from one run of the simulation to another due to the element of randomness that is present throughout the simulation. Two other runs of the simulation with the Likelihood of Littering set at 2.5 showed that an average amount of agents equal to 95 generated 82 pieces of litter and an average amount of agents equal to 235 generated 407 pieces of litter. These results indicate that the amount of litter generated is proportional to the amount of agents (Figure 1).

Original Model - Number of Agents or Litter vs. Time


Number of Homers or Number of Litter 400 350 300 250 200 150 100 50 0 0 200 400 600 Time (seconds) 800 1000 1200 Homers (Avg. Pop. 95) Litter (Avg. Pop. 95) Homers (Avg. Pop. 129) Litter (Avg. Pop. 129) Homers (Avg. Pop. 235) Litter (Avg. Pop. 235)

Figure 1 Graph showing the number of agents or litter vs. time. for the original model

Variations of the Model Inspired by Literature Cialdini, et al. (Cialdini, Reno, & Kallgren, 1990), carried out a series of studies to assess the effect of norms on littering in public places. They distinguish between descriptive norms and injunctive norms. Descriptive norms represent what is typical or normal or what most people do while injunctive norms represent rules or beliefs as to what constitutes morally approved and disapproved conduct (Cialdini, Reno, & Kallgren, 1990). They postulate that littering is more likely to occur in environments that are already littered since this may indicate that the descriptive norm there is to litter and also due to the perception that littering in these environments will cause less harm. Cialdini, et al., confirmed their assumptions and further discovered that people are less likely to litter in an environment which is clean except for one single piece of conspicuous litter. In another study, Andrighetto and Villatoro (Andrighetto & Villatoro, 2011) experiment with the effects of punishment and sanction by building an agent-based model. For them, punishment implies an economic cost being imposed on the offender while sanction implies the violation of a norm. 4

Mireille Micallef

University of Malta/James Madison University

27-Feb-2013

Andrighetto and Villatoro found that sanctions have superior advantages to punishment since people are disposed to obey the norm even when there is little possibility of instrumental gain, future reciprocation, and when the surveillance rate is very small (Andrighetto & Villatoro, 2011). Thus sanctions result in higher compliance and lower costs since enforcing action is needed less often. To try to replicate the effects of descriptive and injunctive norms and punishment and sanctions, the original model was altered to produce four other models. In one of the models, three new bins were introduced at the same location as the take away outlets in the hope that more agents would collide with the bin and thus more litter would end up in bins as opposed to ending up on the ground. This could somewhat represent an injunctive norm that suggests to agents that it is not accepted to litter on the ground in this area. When one sees a bin which is relatively close, they are more likely to use it than when there are no bins in the area. In another model, two Warden agents that impose fines were introduced. The Wardens move like any pedestrian however, if there is an agent that litters within two units of distance from them, the littering potential of all the Homer agents within two units of distance becomes 100. This mechanism mimics the effect of seeing a fellow pedestrian littering and being fined for it; when one sees another person being punished they are less likely to commit the same action. This model represents the imposition of punishment. The third model includes signposts at various locations along the promenade. These signposts would show a phrase or a symbol that indicates that people should not discard their litter on the ground. Upon collision with a signpost, the Littering Potential of Homers gets reassigned a number from 2.6 to 100.6 to make sure it is above the 2.5 limit set by the Likelihood of Littering shared variable. This attempts to replicate the situation wherein anyone who notices the sign is deterred from littering on the ground; however it still allows Homers to have their Littering Potential reduced to a point at which they can litter when they collide with a F & B Outlet. The effects of such signposts will be larger when they are first installed since as time goes by, people get used to them and pay less attention to them. This model can represent both an injunctive norm and a sanction since the message conveyed by the signpost is more explicit than placing additional bins. In the next model, two agents that represent sweepers are introduced. The sweepers move like normal pedestrians. However, upon collision with litter, the litter dies and therefore it gets removed from Spaceland. Also, when Homers are within a radius of two units of distance from a sweeper, their Littering potential gets reassigned a number from 2.6 to 100.6 to make sure it is above the 2.5 limit set by the Likelihood of Littering shared variable. This alteration in the model tries to represent the situation when one sees a sweeper collecting litter from the ground in general, one would be less inclined to disposing of litter on the ground in the presence of a sweeper, however, this does not mean that one is less likely to dispose of litter on the ground at a later stage when the sweeper is not close by. This model also represents an injunctive norm or a sanction. A final model was attempted wherein if an agent saw someone else litter and the environment was littered, their likelihood of littering was increased while if they did not see another agent litter but their surroundings were littered their likelihood of littering was increased but by a lower amount. On the other hand, if agents saw someone litter and their environment was not littered, their likelihood of littering was decreased while if they did not see someone litter but their environment was clean, their likelihood of littering was decreased but by a smaller amount. This was done in order to attempt to replicate the results obtained by Cialdini, et al. (Cialdini, Reno, & Kallgren, 1990). This model was not considered further since the results showed that initially, no litter was produced but as soon as the first piece of litter appeared, the amount of litter rose and grew out of control.

Mireille Micallef

University of Malta/James Madison University

27-Feb-2013

Results and Analysis of Varied Models The Likelihood of littering was kept constant at 2.5 throughout the simulation of all models. Each model was run several times until comparable average populations were obtained. This was done so that the results could be compared across different models. The results are graphed in Figure 2.

Varied Models - Number of Agents or Litter vs. Time


200 180 Number of Homers or Number of Litter 160 140 120 100 80 60 40 20 0 0 200 400 600 800 1000 1200 Time (seconds) Homers (Bins at Take Aways, Avg. Pop. 116) Litter (Bins at Take Aways, Avg. Pop. 116) Homers (Wardens, Avg. Pop., 116) Litter (Wardens, Avg. Pop., 116) Homers (Signposts, Avg. Pop. 119) Litter (Signposts, Avg. Pop. 119) Homers (Avg. Pop. 129) Litter (Avg. Pop. 129)

Figure 2 Graph showing the number of agents or litter vs. time. for the original model and varied models Table 1 Summary of Results

Model Bins at Take Aways Wardens Signposts Sweepers

Average Population 116 116 119 129

Litter 120 96 101 81

From Figure 2 and Table 1 it can be seen that the most effective measure is to introduce sweepers in the area. The introduction of wardens and signposts can be said to be equally effective since the results from the two simulations are rather similar. The least effective measure is the addition of bins at take away outlets. Sweepers were the most effective measure since apart from reducing the amount of litter; they discourage pedestrians from littering further. They are more effective than wardens and signposts because the latter do not remove any of the litter present but only deter pedestrians from littering. Increasing the number of bins at take away outlets does not aid much in reducing the amount of litter because it is not a prominent enough reminder of the injunctive norm that implies that littering

Mireille Micallef

University of Malta/James Madison University

27-Feb-2013

is not approved of. The results obtained show that sanctions and injunctive norms are more effective than punishment. Difficulties and Sources of Error The information available about the Sliema front is limited and only represents a short time frame. It was collected in one day and no repeated readings were taken. There is also an element of human bias in the data collected since it was collected by different people. This might impact the validity of the model. Due to the coarseness of the terrain patches in the StarLogo TNG software, the area where agents are allowed to move is slightly larger than the actual area of the promenade. Some inaccuracies in the terrain might also have been introduced during the cleaning up of the base map. Coordinates throughout the project are therefore only approximates. Conclusion This study has shown the possible effects of some common measures that are taken to mitigate the problem of littering in public places. While the models indicate that the best solution would be introducing sweepers to the area, it does not mean that this is the most cost-effective measure. In order for the measure to work as modelled, two sweepers would have to be present all the time on the coast of Sliema this might prove to be costly. The solution involving the introduction of wardens suffers from the same problem. Probably, the most cost-effective solution would be to install signposts in the area; however, people might become used to them and not take notice anymore in which case they would lose their effectiveness. In order to test this hypothesis, the model would have to be fashioned in such a way as to simulate learning over time. References Andrighetto, G., & Villatoro, D. (2011). Beyond the Carrot and Stick Approach to Enforcement: An Agent-Based Model. In B. Kokinov, A. Karmiloff-Smith, & N. J. Nersessian (Ed.), European Perspectives on Cognitive Science. Sofia: New Bulgarian University Press. Cialdini, R. B., Reno, R. R., & Kallgren, C. A. (1990). A Focus Theory of Normative Conduct: Recycling the Concept of Norms to Reduce Littering in Public Places. Journal of Personality and Social Psychology, 58(6), 1015-1026.

Você também pode gostar