FORMA 13 : Complex Systems Workshop: Sevilla 23-26 Oct 2013

I. The Workshop WHAT IS FORMA? FORMA is an initiative to promote agent-based thinking among researchers from every single discipline such as art, biology, maths, sociology, economics and urbanism. Agent-based systems are used to study complex realities, which can hardly be shaped by traditional methods. The aim of this special event is therefore, to provide participants with a solid agent-based modelling expertise that will enhance their understanding of complex processes within their field of research. The workshop is organized around a limited number of case studies, selected among proposals submitted by researchers from all over Spain. Each case study has a clear target to be addressed during the three full-days of the event, and the authors of the proposal are assigned an agent-based modelling expert plus a team of students, typically from Computer Science and Mathematics, that are in the process of developing modelling skills. *** Forma provides an interdisciplinary space where students can work together to solve real problems, while researchers from any field meet the opportunity to work hand in hand with highly skilled programmers towards their research targets. This event unfolds as a great networking tool, where minds with different assets meet to pursue common interests. After this first edition, many of the teams have continued developing their models, or established other kinds of collaboration, even sending abstracts together for international conferences etc. Apart from the "hands-on" work, the workshop includes top level speakers from the European Network for Social Intelligence, as well as daily tutorials on "soft-learning- curve" languages for agent-based modelling, such as NetLogo, and also Processing. Researchers with no previous knowledge of these tools, will develop enough understanding to prototype new case studies in the future through agent-based modelling, even if only conceptually. OPEN CALL AND SELECTION PROCESS In this first edition, a total of fourteen proposals were received, from which a total of twelve were selected, leaving out only two proposals that were too broad for the scope of the workshop. The communication strategy of the open call for proposals, being the first edition of FORMA, was mainly two-fold; On the one hand, direct contact was established with researchers from the two main universities in Seville, US and UPO, holding brief meetings to discuss potential topics within their research that could benefit from agent-based modelling. Three out of five proposals coming from US had been discussed with some members of the organization prior to their submission. On the other hand, an important effort was made to advertise both on paper and digital media. Prints were distributed throughout all main centres of US and UPO, and other cultural entities and research centres such as ICAS (Institute of Culture and Arts of Seville) and IPTS (one of the seven scientific institutes of the European Commission's Joint Research Centre), while almost every communication department from major universities in Spain, received a request via email to post the open call in their local bulletins. Some departments were also emailed directly. Other digital channels like twitter were used intensively to spread the call for proposals and participants. Finally, prior to the workshop, two Processing "jams" we organized in May and July respectively, with the purpose of introducing both the language (and its potential application for agent-based modelling), and FORMA. Both events were held at independent cultural hubs in Seville's city center (Tramallol and La Mirada Rota) and had a good number of participants from very various backgrounds. From the twelve selected proposals listed below, N3, N4, N5b, N5c and N5d were submitted by research teams from Universidad de Sevilla. Proposals P1 and P2 were submitted by independent researchers from Seville. Proposals from other Universities in Spain and abroad were: N1 (Universidad Nacional Autónoma de México), N2a (Universidad de Granada & European Centre for Soft Computing), N5a (Universidad de Alcalá) and N6 (Polytechnic University of Catalonia - Barcelona Tech). Finally, proposal N2b was submitted by an independent consultancy based in Barcelona (SustainValues) CASE STUDIES SELECTED P1. Helicoidal interrelation of prime numbers Author: Javier Sánchez Bravo N1. Birth and Death Master Equation for the Evolution of Complex Networks Author: Federico Cruz Ponce de León N2a. Modelling Consumer Behaviour and Marketing Actions Authors: Manuel Chica Serrano, Oscar Cordón García N2b. The value creation framework Author: Luisa Nenci N3. Changes in personal networks during ecological transitions Author: Isidro Maya Jariego, Carla Andreia Carvalho Gómez, Cristina Varo Martín N4. Will the frog jump out? Institution-Faculty relationships in the University of South Zembla Authors: Jose Pérez de Lama, Antonio Sáseta Velázquez, María José Lera Rodríguez N5a. Residential settlement of the population at sub regional scale Author: Carolina de Carvalho Cantergiani N5b. Mobility, social diversity and sustainability Authors: Simona Pecoraio, Pablo Pérez Ganfornina N5c. Making city in the "non-city": integration versus isolation Author: Antonio Piñero Valverde N5d. Case study for the regeneration of neighbourhoods: health patterns of barrio de Villa Coronilla, Cochabamba Authors: Salas Mendoza Muro, Carolina Ureta, Sergio Rodríguez Estévez, Luz Fernández Valderrama N6. Microbial populations within a bioreactor operating with different protocols Authors: Pablo Araujo Granda, Marta Ginovart Gisbert P2. Green Alcazar Author: Sergio Rodríguez Estévez. Salas Mendoza Muro, Francisco J. Pazos García *N=NetLogo, P=Processing SELECTION CRITERIA FOR CASE STUDIES FORMA is targeted towards studying complex processes by means of agent-based modelling. The criteria followed during the selection of case studies in this first edition have been: Contemporary subjects: The case study shall address a topic along the lines of the new information paradigm, digital societies, technological advances and other subjects in which network behaviour plays a decisive role. Complex processes: Models should aim to clarify a well-defined phenomenon, which due to its inherent complexity requires advanced computational techniques and an interdisciplinary approach. Spatial interaction: The workshop targets the study of complex processes in which spatiality, be it literal or abstract, has a strong influence over the behaviour of the system at play. Visual understanding: Proposals shall request for results that can be directly communicated through a graphic visualization of the model and be easily understood by all audiences. Clear identification of the problem: Because of the limited time to develop the models during the workshop, it is fundamental that a key question is clearly identified in the author's proposal. TEAMS & ORGANIZATION Once the case studies were selected, one month prior to the workshop, a brief documentation explaining each proposal was uploaded to the website. Then, all participants were invited to fill in a form via email indicating three of the selected case studies ordered by preference, and also, a short note on their background. With that information the organization distributed the participants in teams of 4 to 8 members per team, and assigned a modelling assistant to each one of them as well. Every participant was informed of the team they had been assigned to beforehand. The workshop opened on Wednesday 23th Oct, 5.00 pm with a short introduction. Then followed a lecture by F. J. Miguel Quesada, introducing social simulation and agent based modelling. And finally, a round of ten-minute presentations of all the proposals, carried out by their respective authors. Next day kicked off at 9.00 am with another lecture by F. J. Miguel Quesada, suggesting the ODD protocol as a means to create a common understanding between researchers and technical modelling experts. Afterwards, teams started working on their proposals until evening, first addressing the main features of the model on a conceptual level, and then outlining the structure of the code itself (teams were asked to write an ODD protocol and to design work-flow diagrams of their models before actually coding them). Modelling assistants had the task to lead and work hand in hand with all team members. Directors and guest tutors performed individual deskcritics with all teams throughout the day. During the first half of the day, there were two parallel programming sessions of two hours each, to introduce Processing and NetLogo to those participants who had no previous experience with either of the two languages. The third day, Friday 25th, opened with a morning lecture by Francisco Linares, looking into real applications of agent based modelling in the field of Sociology. The rest of the day unfolded quite like Thursday, including the two-hour programming sessions, working in teams (now focusing on code development) and desk-critics. Finally, on Saturday 26th, all teams presented the results of their models throughout the morning, and by the end of the last presentation there was a discussion session and debate which marked the end of FORMA13. CREDITS Organized by: Culturadigital http://culturadigital.cc Chair of Computer Science and Artificial Intelligence, Universidad de Sevilla http://www.cs.us.es Collaborators: Laboratori de Simulació de Dinàmiques Socio-Històriques http://sct.uab.cat/lsds Chair of Information Technology, Universidad de Huelva http://www.uhu.es/dti/ Support: SINTELNET (European Network for Social Intelligence) http://sintelnet.eu eCompleXcity Project TIC-6064 http://ecomplexcity.org Fidetia number, it should experience an increased rate of change. The analysis of the networks will be done with other software to facilitate the implementation of the model and increase the depth of the conclusions. Initialization: The other variables shall be chosen by the observer; including the total number of nodes and the type of graph to begin with, as well as the variables in the equation that have not been defined. Input: No external input needed yet. A later proposal for the upload of certain preconceived networks to the NetLogo interface may be pertinent. Submodels: The total number of nodes, for now, will remain at 100 to facilitate further analysis. The other variables in the equation have been chosen for their repeated discoveries within several realworld phenomena. Since how the network will evolve using the equations is unknown, a calibration must be made after the model is complete. Flow chart: Code: N1_code.nlogo View it on Github: https://github.com/culturadigital/forma13/blob/master/N1.nlogo Model snapshots: Figure 1: A basic outline of the program, the numbering helps to order the steps needed to run the program. Figure 2: The program running from a scale-free netw ork, it graphs the degree distribution in real time. Figure 3: It can resize nodes according to their degree and also become dynamic, moving more connected nodes tow ards the center and lesser ones to the borders. Conclusions: We were able to finish a very near final draft of the program in NetLogo thanks to everyone's understanding of the ODD protocol. The team was composed of people familiar with the interface as well as we all had a background in computer programming, which made for an easy workload. For now, the network seems to run perfectly, every tick has a newly created and newly deleted link, however the beta distribution that should form is not as expected. The implementation of the variable probability does not seem to be the problem as the method was modified from the "Lottery Example" code of the NetLogo library, so a revision of the behavior of the equation itself is needed. Without the program, this would not have been so evident. We were also able to implement different starting networks to start the program such as a scalefree network, however we were unaware of the "Network" extension of Netlogo that allows you to do exactly that with much less effort. All in all, we were very successful in achieving the goals put forth in the beginning. References: L. Luthi, et al. Cooperation and community structure in social networks, Physica A 387 (2008) 955-966. Roberto Alvarez-Martinez, et al. Birth and Death Master Equation for the Evolution of Complex Networks. Related models: http://ccl.northwestern.edu/netlogo/models/PreferentialAttachment https://github.com/shawngraham/netlogo https://github.com/donovantc/peertrustsimulation N2A. MODELLING CONSUMER BEHAVIOUR AND MARKETING ACTIONS Authors: Manuel Chica Serrano - European Centre for Soft Computing [email protected] Oscar Cordón García - Universidad de Granada [email protected] Collaborators: Juan Galán Páez Mercedes Gómez de la Rosa Lurdes Caleya Ortiz David Solís Martín Brief: The goal of the case is to study and model the consumer behaviour when choosing and buying different products of a specific market. The behaviour of good companies are also modelled by applying dyanmic marketing actions when sales change. These actions will try to maximize their sales in the competitive market of the model. Key question: What are the sales of our product in the market and how do our marketing actions affect my sales? ODD Protocol: Goal: Simulation of Orange Aquarius sales in Sevilla with respect to the competitors Gatorade, Radical and Trina. The aim of each Enterprise is to maximise the sales of its products within agents' population. Entities: Two kind of agents are involved: consumers and products/enterprise. Interactions between consumers are modelled by means of a randomly generated small-world shaped social network. State variables: State variables for CONSUMER agent: Priority to the price of a product ([0,1]). A valuation of 1 means that consumer ignores the quality of a product and bases it choices on the price only. The opposite means that the consumer takes into account only the quality. Priority vector, to store a preference value for each of the products (real vector with [0,1] values). Influenciability by other's opinions ([0,1]). 1 means max inlfluenciailibty. Proactiveness denotes how proactive is each agent while spreading its experiences with products to its neighbourhood ([0,1]). 1 means max proactiveness. Level of expectation ([0,1]) denotes how difficult is to satisfy an agent's demand on a product. A valuation of 1 means max expectation. State variables for PRODUCT-EMTERPRISE agent: Price ([0,1]). 1 means max price possible. Quality ([0,1]). 1 means max quality possible. Sales. Number of items sold for this class of product during the whole simulation. It will be a natural number, and the goal of the enterprises during the simulation is to maximize its value. Scales: Time scale: tick = day. The time unit for the simulation is a day and the whole simulation lasts for two years. In a day (tick) consumers can buy products, and enterprises can change their marketing action plan in order to increase their products sales. Procedures: Time is discrete (one day chunks). There are always available products in stock, thus consumer can always choose between all of them. Initialization: The social network between consumers is randomly initialized following a small-world topology. During the simulation the network remains static. Most of agent's parameters are randomly initialized, for details about parameter initialization visit the code (the initial value of some parameters is bound to the value of others). The main procedure can be divided in four stages. Initialization, opinion broadcasting between consumers, buying and preferences modification stage (consumers) and finally revisit marketing action plans (enterprises): A. Buying procedure: First consumer agents choose (with 50% probability) if they have the need to buy a drink. In case the answer is positive they have to choose between one of the products. In this decision some factors are involved (price, consumer preferences, priority of the consumer for a cheap product or a high quality one, etc.), and the utility function for each product is as follows: f(product) = (priority_to_price) * price + (1 - priority_to_price) * preference_for_this_product Once the utility is computed for each product, the selection is made by means of a random weighted choice (using a weighted probability distribution), where the weights are the utility value for each product. After selecting a product, the sales for this product are updated. B. Update of product preferences based on the product purchased: After buying a product, the priority the consumer has for the product is updated based on the satisfaction obtained, the level of expectation and the price. If the consumer as a high expectation level and the product quality is low, the probability for the consumer to not being satisficed with the purchase is high. In this case, the priority of the consumer for this product will be decreased in 0.1; otherwise, it will be increased in 0.05. Expressing this in pseudocode: If quality of the purchased product is lower than the expectation level of the agent The probability for the consumer to not being satisficed is ((expectation_level - product_quality) / expectation_level); If consumer is not satisfied then priority(product) = priority(product) - 0.1 Otherwise (quality of the purchased product is greater than the expectation level of the agent) Priority (product) = priority(product) + 0.05 C. Neighbourhood influence in agent's priorities about products: We have already seen how consumer's priority about a product changes according to its satisfaction with the last product purchased. Independently from this, consumer's priorities about products change due to the opinions the consumer receives from its neighbouring agents. This process depends on other factors as the proactiveness of the agents broadcasting their opinion and the influenciability degree of the consumer accepting others' opinions. Firstly, and based on the influenciability parameter, in each day (tick) the consumer decides if today he is going to accept others' opinion. In this case, depending on the proactiveness of the neighbours, the influencing opinions are collected and an average priorities vector is computed. Once we have this vector, the consumer's priorities vectors (from the agent receiving others opinion) is modified according to the following expression: f(new_priorities) = (influenciability) * AveragePrioritiesProactiveNeighbours + (1 - influenciability ) * FormerPriorities D. Enterprises' marketing action plans in order to improve product sales: Every month, enterprises can change their products' marketing action plans in order to improve sales. In order to keep it simple, marketing action plans consist on the variation of price and quality product attributes. In a first step the enterprise decides if is necessary to act (if sales rate is high, Why am I going to change anything on my product?). Every month each enterprise checks its own and competitors' sales. If sales are above the average of competitors sales, business are going well so we do not touch anything. Otherwise, there are two possible actions (one of those is randomly chosen): Price decrease: the enterprise decides to decrease the price of the product in 0.05, and as consequence of this, the product quality also decreased in 0.05 (bound parameters). Quality increase: the enterprise decides to increase the quality of the product in 0.05, and as consequence of this, the product price also decreased. After updating the marketing action plans, the enterprise waits one moth to analyse the effects of the changes and reconsider the plans again if necessary. In this model, enterprises do no have access to any information about consumers preferences, thus the can base their decisions only on sales information. Input parameters (User configurable input parameters in the model): Consumers' population. Number of available products. Quality - price relationship in order to compute the quality property of the products. Some parameters are generated by means of a normal distribution with fixed standard deviation (0.1) and user provided mean. The input parameters for the means are: Mean for the consumers expectation level. Mean for consumers' priority towards price. Mean for consumers' priority. Mean for consumers' proactiveness. Code: N2a_code.nlogo View it on Github: https://github.com/culturadigital/forma13/blob/master/N2a.nlogo Model snapshots: Figures 1, 2. Product prices are randomly generated. Product quality is computed adding a random number w ithin the range [- quality/price-relationship, quality/price-relationship]. Consumer parameters expectation level, priority tow ards price, influenciability and proactiveness are computed follow ing a normal distribution w ith 0.1 as standard deviation and the follow ing input parameters as mean: mean_expectation_level, mean_price_priority, mean_influenciability and mean_proactiveness. The size of consumer turtle is proportional to its priority tow ards the price. The colour of the consumer turtle represents the purchased product (w hite for none). Conclusions: We developed the model using Netlogo and configured different parameters to run simulations (number of products and costumers, ratio price-quality for the products, agent activity in the social network, etc...). A graphic with the sales lets us observe changes in sales when companies vary their products characteristics (quality and/or price). Future extension of the model will consider marketing actions depending on the characteristics of the consumers. References: Tesfatsion, Leigh. "Agent-based computational economics: Growing economies from the bottom up." Artificial life 8.1 (2002): 55-82. Janssen, Marco A., and Wander Jager. "Simulating market dynamics: Interactions between consumer psychology and social networks." Artificial Life 9.4 (2003): 343- 356. Delre, S. A., et al. "Targeting and timing promotional activities: An agent-based model for the takeoff of new products." Journal of business research 60.8 (2007): 826- 835. Shafiei, Ehsan, et al. "An agent-based modeling approach to predict the evolution of market share of electric vehicles: A case study from Iceland." Technological Forecasting and Social Change 79.9 (2012): 1638-1653. Related models: http://ccl.northwestern.edu/netlogo/models/community/customerBehavior http://ccl.northwestern.edu/netlogo/models/community/EXPLORE http://ccl.northwestern.edu/netlogo/models/community/Neteffect http://ccl.northwestern.edu/netlogo/models/community/ModelMarket http://ccl.northwestern.edu/netlogo/models/community/Location%20Game http://ccl.northwestern.edu/netlogo/models/community/OptionsMarket http://ccl.northwestern.edu/netlogo/models/community/Facebook http://ccl.northwestern.edu/netlogo/models/community/Fall-of-banks https://sites.google.com/site/jkgeconoeng/GBLCSA.zip http://www.css.gmu.edu/node/81 https://github.com/ancadumitrache/DutchHousingMarketSim network state at the beginning of the process. Inputs: The model uses data from social networks in DL format. Code: N3_code.nlogo View it on Github: https://github.com/culturadigital/forma13/blob/master/N3.nlogo Model snapshots: Figure 1. Initial state of a netw ork (data is obtained from real w orld). Figure 2. State of the netw ork and its measures after 247 ticks of evolution applying model rules. Figure 3. Final state of the netw ork after 500 ticks (18 months), it is useful to compare measurements of this netw ork w ith the ones obtainied in the real experiments, netw orks in T2. Conclusions: We have designed a model that simulates the behaviour of students who changes their environment. After implementing the model and performing several experiments we have found that is possible to model this behaviour through few simple rules. Just taking into account frequency of losing and winning relationships between people, it can demonstrate that the change in the social environments follows simple rules although the evolution of networks can looks very complex if we analyze it like a centralized system. References: Domínguez, S. & Maya-Jariego, I. (2008). Acculturation of Host Individuals: Immigrants and Personal Networks. American Journal of Community Psychology, 42, 309-327. Maya-Jariego, I. & Armitage, N. (2007). Multiple Senses of Community in Migration and Commuting: The Interplay between Time, Space and Relations. International Sociology, 22 (6), 743-766. Maya-Jariego, I. & Holgado, D. (2005). Lazos fuertes y proveedores múltiples de apoyo: comparación de dos formas de representación gráfica de las redes personales. Empiria. Revista de metodología de ciencias sociales, 10, 107-127. Molina, J. L., Maya-Jariego, I. & McCarty, C. (2013). Giving Meaning to Social Networks: Methodology for Conducting and Analyzing Interviews based on Personal Network Visualizations. In Betina Hollstein & Silvia Domínguez (Eds.). Mixed Methods Social Network Research. Designs and Applications. Cambridge: Cambridge University Press. Related models: http://ccl.northwestern.edu/netlogo/models/TeamAssembly https://github.com/donovantc/peertrustsimulation N4. WILL THE FROG JUMP OUT? INSTITUTION-FACULTY RELATIONSHIPS IN THE UNIVERSITY OF SOUTH ZEMBLA Authors: José Pérez de Lama - Universidad de Sevilla [email protected] María José Lera Rodríguez - Universidad de Sevilla [email protected] Antonio Sáseta Velázquez - Universidad de Sevilla Collaborators: Jesús Horno Cobo Miguel Gutiérrez Villarrubia Sergio Morales Casas M. Carmen Romero Ternero Lorenzo Perea Lerate Iván Coronado Moreno Ricardo Jiménez Odero Rocío García Brief: During the great crisis of 2007-2013, the central government of the evil Charles Kinbote and his minister E. Shade, are inflicting severe policies cutting down budgets on social services, and particularly in the areas of basic and university education. In the old and honorable University of South Zembla, located in the southern capital of the country, the Board and its collaborators hold an ambiguous position by supporting rebellious discourses against the followers of Kinbote-Shade, while at the same time making effective those resolutions coming from the central government among their faculty. A minority yet substantial part of the faculty, the Pink-Bloc, promotes resistance against the Board and higher authorities, is making efforts to call on a faculty strike as a proof of power, and as a means to balance political forces. However, most faculty, despite feeling under attack, remain passive, perhaps because they are skeptical towards any act of mobilization, or simply because the best way to keep their jobs is to adapt to the emergent conditions the best they can. The Pink-Bloc will take the metaphore of the frog in the cooking pot, which bacame popular after american politic Al Gore, to interpretate the political techniques of the University of South Zembla, which are being developed by the Board thanks to the unestimable collaboration of the historically honorable labor unions, which after so many years of dedication to fight capitalist oppression, have finally been abducted by the local spheres of power... Key question: Will the frog jump out? Or what are the conditions required in order for a faculty strike to take place in the University of South Zembla? And what level of success could such strike achieve upon diferent configurations of these conditions? ODD Protocol: Purpose: What we have is an attempt of a model for an hypothetic case of cutbacks that could result in a strike. All of this in an academic/university enviroment. Our intention is to develop a first approach for a more elaborated model in the future. State variables and scales: There are two kind of agents in the model. The first one is called "machine" ("aparato") and represents the government/administration. It's an observer-like agent so it's not represented explicitly in the model but his actions have a deep impact in the evolution of the situation and the mood of our second group of agents. The so-called "frogs" that represent the whole bunch of Teaching and Research Staff in the university. The "machine" is meant to be a burocratic system that have de ability of changing the aim of the "frogs". Its role is supposed to be played by the observer by pressing buttons that will trigger actions such as "confusion", "increment the dedication time", "salary reduction" and "dismissal" at any time. The "frogs" are literally represented in our model by turtles, as well as the links among them that represent the relation of influence existing in real human networks. They have four characteristics: Category: Represents the status of the frog in the university, the sector they belong to. VPS: Represents the psycological status of the agent. The natural inclination for protesting against their situation at work. Pression: Represents the pressure they are subject to. Jump: It's calculated as vps times presion and it represents the current predisposition for going to strike or protest. The color of the frogs is related to their "jump". Process overview and scheduling: The process is very simple, as time goes by, two things happen to the frogs. The first is passive infection of the feelings around them accordingly to the people in their inmediate environment, "neighbours" throughout time. The second one is conformity, that represents the natural cooling of the spirit of an annoyed person. The rest of actions are executed by the observer whenever he wants. And represent the actions of the machine. Design concepts: As a first approach, the model try to collect the basic feelings and reponses from university personnel to the actions of the machine. At this moment, the agents don't have the hability of learning from their enviroment, but it could be a nice case study for a more developed model. The links among frogs are created as a clustered graph thar could represent research, friends, or teaching groups in the university. With some people very interconnected between them and less conections with other groups. While the model is running, a graphic of the number of frogs divided in three groups (grey, green, pink) is draw in real time so you can see how does the global mood is in the model and how do machine actions affect people from a global viewpoint. The colour of a frog is a sign of his mood. It gives us visual information of the situation. There are three main groups: Grey: Passive people that are satisfied with their situation. Green: People that are unsatisfied with their situation or the situation of people near them. Pink: Active people that have a big enthusiastic attitude and are determined to change things. Initialization: One must set the number of frogs that will be in the model, the average node degree of the link graph and the percentage of fired people every time you apply dismiss. The initial percentage of different frog category follows a 40-30-30 rule. The other data are taken randomly. Flow chart: Code: N4_code.nlogo View it on Github: https://github.com/culturadigital/forma13/blob/master/N4.nlogo Model snapshots: Figure 1. A random initial configuration of the simulation Figure 2. Step show ing many active (pink) frogs due to some machine actions Figure 3. Stable situation after some time w ithout "machine" actions. Everybody is calm Figure 4. Graphic detail of salary reduction Conclusions: Our participation in Forma13 developing a Netlogo sketch of Zembla was the team's first experience working with Netlogo to digitally simulate an emergent social system. Ease of use and immediate visual and numeric outputs are excellent features of the platform. The opportunity to work in multidisciplinary teams offered by Forma13 was highly appreciated. Our short modelling experience shows that fine tuning of parameters is a key issue for the achievement of a significant simulation. Concerning the Zembla model, we think that it is an interesting concept that we have sketched, but that, of course, it would need further work to make it a relevant simulation model. In the state of development achieved at Forma13 we consider it might already have some use in triggering some debate on social dynamics among activists. In order to have real use, we think that one or several alternative models - conceptual and algorithmic - on political-and-activist motivation, networking and action should be developed, testing them against data retrieved in different real situations. References: Emmanuel RODRíGUEZ, 2003, El gobierno imposible. Trabajo y migraciones en la metrópolis de la abundancia, Traficantes de Sueños, Madrid Michael HARDT, Antonio NEGRI, 2004, Multitude: War and Democracy in the Age of Empire, Hamish Hamilton - Penguin, Nueva York John REED, 1977, Ten Days that Shook the World, Penguin Classics, London María José LERA, 2012, Israel o Palestina ¿conflicto o violencia?, en: C. Cruz, C. Ibáñez, S. Moreno, 2012 (coords.), El traje del emperador. 13 propuestas para desnudar el poder, Atrapasueños Editorial, Sevilla; pp: 167-181 Richard E. Boyatzis, 2006, An Overview of Intentional Change from a Complexity Perspective, en: Journal of Management Development, Vol. 5 No. 7, 2006, pp: 607- 623; available at: www.emeraldinsight.com/0262-1711.htm Related models: http://ccl.northwestern.edu/netlogo/models/Ethnocentrism http://ccl.northwestern.edu/netlogo/models/Rebellion http://ccl.northwestern.edu/netlogo/models/community/Facebook-share... http://ccl.northwestern.edu/netlogo/models/community/Occupational-Stress... http://ccl.northwestern.edu/netlogo/models/community/Smoking-motivation... N5A. RESIDENTIAL SETTLEMENT OF THE POPULATION AT SUB REGIONAL SCALE Author: Carolina de Carvalho Cantergiani - Universidad de Alcalá [email protected] Collaborators: César Vergara Martín Antonia Chávez González Javier Fernández López José Manuel Calahorro Gómez Luis Pardillo Caballero Brief: The Community of Madrid (CM) is strategically placed in Spain and functionally important since it concentrates the transport network main connections, important industrial activities and central administrative services, among others. It results to be very active demographically, producing important territorial changes, confirmed by urban and potential urban land growth statistics, for example. The study area is an important urban-industrial corridor that connects the megalopolis of Madrid and Barcelona, named Corredor del Henares within 18 municipalities. This area has been historically characterized as population and productive activities attractor during periods of decentralization of Madrid and other neighbor municipalities. It can be considered a potential area for urban growth analysis and simulation due to its important demographic growth and land use dynamics. And moreover, those results are also important for being useful for planning and territorial management of its territory. Analyzing the dynamics of the study area, the three selected agents we consider influence its urban growth are: population, real estate agents and urban planners. They simultaneously induce this phenomenon through their own behavior, stimulating the creation of new urban settlements. The three agents' behaviors do not compete for the same goal; instead they are re-feed according to each other action. It means that the planners assign new urban areas (having into account the still not-urban areas); the real estate agents look for available places to build new residence on previously classified urban areas; and finally the population considers the offer of new and empty housing to choose new residences. The challenge in the model presented in this document is to simulate the residence election of the population that finally decides to move. Three groups of population are considered according to their behavior, defined according to their economic status (high, medium and low income), that is supposed to be the most important factor that induces residential changes. Key question: How urban growth at Corredor del Henares (Madrid) can be understood and represented through the behavior of groups of population with different economic status in respect to their residential choices? ODD Protocol: The main goal of this model is to understand how is the urban behavior at Corredor del Henares through the simulation of the residential choices made by different groups of population considering their income levels. Purpose: Simulate residential changes of high, medium and low income population. The model product is, by one side, the amount of people that change residence together with the class they belong to, and on the other side, the type of housing they choose and what their satisfaction level is. State variables and scales Entities (agents): In our case, we consider "agents" the ones who present certain behavior, do something and take a decision, while "environment" is all the components of the system that present any property that affects agents' behavior, but with static attributes (at least at this stage of the model). In our case, "agents" are high, medium and low income population. It won't be considered at individual scale (person), but as a population unit (group of people). As "environment", we consider the housing with its attribute, according to the following: State Variables (Property): At the beginning of the model, there will be a parameter called PERCENTAGE (%) that indicates the amount of random changes on initial income type of the population. Scale (Spatial and Temporal Resolution): The study area is represented by a matrix of 800 x 800, each unit with 50m x 50m. Temporally, the project aims to analyze residential changes each month, even though the temporal unit (tick) is not yet resolved (since a tick will be probably lower than that), so there will be no fixed temporal unit at this testing stage. Process overview and scheduling: The first and only action is the movement of the unsatisfied population to available residential buildings (that comply with some characteristics). In this case the order is not important, since all population act simultaneously. But as the model must define an order to be able to run, we will let it random. No external updating is necessary, as it is a simple model that will feed itself with the new information generated. Design Concepts Basic principles: There are social, economic and territorial aspects that affect the behavior of the population regarding changing residence. Although in the present proposal they are defined in a very superficial way through a limited set of actions. After the running of the model, we will be able to reanalyze those same aspects on the new distribution, and find some relation and conclusions about their relations. Emergence: The choice of residential change is unforeseen, since it varies according to a high number of combinations of non-deterministic criteria. The action will depend on the satisfaction, that in the present simulation is defined by the coincidence of the income level and housing typology (which change could be random). Adaptation: The population occupies housing of their own economic level, and among them, they prefer the ones with better quality respect to territorial position. For Diagram 2. Initialization ("Go" function) Diagram 3. Creation of new inhabitants / land ow ners in the settlement Diagram 4. Actions executed by the Administration Diagram 5. Actions executed by land ow ners Diagram 6. Updating impact levels Code: N5c_code.nlogo View it on Github: https://github.com/culturadigital/forma13/blob/master/N5c.nlogo Model snapshots: Figure 1. System status at simulation startup. Green patches are protected land. White - rural land. Red - built area over protected land. Brow n - built area over rural land. Figure 2. Unexpected turning point of impact trend at tick 200 (50 years) Conclusions: After an initial model has been developed, we have realised that assesing impact levels and especially satisfaction degree among the population, require a much deeper analysis prior to the modelling phase. Nevertheless, we have discovered an unexpected phenomenon that might lead us to some fundemantal findings in our research. Despite the simplicity of the model, the simulation shows a turning point in the impact level of the settlements after aproximately 50 years of its initialization. Of course, the time lapse is not very reliable, but if this turning point does happen at some point it could be of very useful value to understand and reduce the processes of degradation inherent to self-organizing communities. Moreover, this initial attempt has proved very useful to define a long term agenda in our study of irregular settlements through agent based modelling. References: Decreto 2/2012, de 10 de Enero, de regulación de edificaciones y asentamientos en el SNU (BOJA núm. 19, de 30-enero-2012); y, de modo específico, las Normativas Directoras que lo desarrollan (Orden de 5 de marzo de 2012) Related models: https://github.com/dgolicher/netlogo https://github.com/adriancbjie/IBG http://www.gisagents.org/2010/07/netlogo-gis-examples-from-backspacesnet.html http://fs.urba.arch.unifi.it/en/ric/netLogo.php.html/ N6. MICROBIAL POPULATIONS WITHIN A BIOREACTOR OPERATING WITH DIFFERENT PROTOCOLS Authors: Pablo Araujo Granda - Polytechnic University of Catalonia (Barcelona Tech) [email protected] Marta Ginovart Gisbert - Polytechnic University of Catalonia (Barcelona Tech) [email protected] Collaborators: Francisco Romero Campero Yago Fernando Rodríguez Raúl García Bravo Carolina Caballero López Javier Atoche Ortega Francisco M. Romero Romero Brief: Biotechnology experimentation in the laboratory with microorganisms is not always viable, involves a series of constraints such as: cost, time, safety, security, equipment, sterility and others. These tasks could be managed with the virtual experimentation. The use of a virtual bioreactor would facilitate the understanding of the real system favoring final handling. People in different training courses seek to respond a series of questions, such as: what is a bioreactor? Could I handle a bioreactor? What happens within a bioreactor? Which behavior has a population of microorganisms within a bioreactor? The answers could be obtained using a simulated complex system (virtual bioreactor), which facilitates the analysis of the behavior and response of different populations of microorganisms in different conditions of culture medium that can be controlled by the observer. The choice of a basic approach to study microbial system, either at the level of population (usually continuous with differential equations) or based on the individual (discrete and computational model) is an important decision that should be taken depending on the project, the system characteristics, the problem to study and assumptions when the continuum is not applicable. Individuals based Models (IbMs) deal with individuals (live actors) as discrete autonomous entities. These individuals have particular characteristics that change according to the biological actions of its own behavior, interactions with others and the environment. Population-level approaches are mainly used for prediction purposes, due to its simplicity and efficiency computational, and they are widely used. The IbMs power is the features offered to interpret, understand and investigate the dynamics of living systems. Allow us to treat the intrinsic variability of all living being, with the emergence of the population behavior from the actions and interactions at the individual level. The main objective is to develop a microbial bioreactor simulator, which is easy to understand, and that it could be adapted to specific applications and useful for people on the life sciences field. Parameters: the system will be a virtual bioreactor operating with different protocols (batch or continuous culture), which will contain initial amounts of nutrients, a number of viable microorganisms, who will adapt to the environment that they are, because they consume nutrients and generate end products, maintaining they viability over the time. Agent (micro-organism) variables: a. Position in space (XY coordinates) b. Biomass c. Maintenance requirements d. Reproduction biomass e. Reproduction time The environment variables (culture medium): a. Nutrient concentrations b. Product concentrations System variables: a. Nutrient input flow. b. Culture medium output flow c. Dilution ratio d. Operation time Key question: Is it possible to simulate the behavior of a microbial population in different conditions within a bioreactor? ODD Protocol: 1. Overview Purpose: The purpose of this model is to provide a virtual bacterial bioreactor with an operating protocol to manipulate the entrance of nutrient, stirring and output of medium in order to produce biomass. Only a qualitative study of some mechanisms is acceptable to explore or investigate the behaviour of this system. In addition, the idea to implement this model into the 3D version in the near future has motivated some of the approaches used. Entities, state variables, and scales: Basic entities are bacteria and spatial cells of the medium where these microbes evolve. The bacterial attributes are: an identity number, a location (i.e., the spatial xy-coordinates of the grid lattice), a mass (as a source of energy that can be used for the reproduction and movement) the mass to start reproduction and the mass for the entrance to the latency state. The grid cells attributes are: the x and y integer coordinates for each site, the nutrient content and end product content. It can simulate a population of up to hundreds bacterial cells in a 40x40 spatial grid cells. This is a non-parameterized version so there is not a real units correspondence. Process overview and scheduling: At each time step of the simulation, a set of actions a set of actions are performed. The sets of rules governing the behaviour of each bacterium are: Uptake of nutrient particles in order to increase the bacterial biomass, with excretion of the end product to the spatial cell. Reproduction by bipartition with the generation of two "new" bacterial cells when the mass is greater than a threshold. Viability that is assumed till the mass is under certain threshold, translating the cell to a latency state. (i) Active motion only when the bacterium has not achieved nutrient (randomly to a neighbour patch) The sets of rules regarding spatial cells include: The entrance of fresh medium. The culture stirring or agitation to avoid adhesion of bacteria in the input flow mechanism or in order to achieve homogenization. Output of medium (with nutrient and product) and bacteria. The global scheduling of the simulation model is made up of various elements: (1) initialization of the system with the input data chosen by the user, where initial configuration of the bacterial population and the spatial environment are set up, as well as the parameters for the operating protocol; (2) the main loop (time step), in which all the rules for each bacterium and the medium are implemented and repeated, and the external actions on the system are applied, until reaching the end of the simulation; (3) the output of results at the end of each time step. 2. Design concepts Basic principles: The general concepts and basic principles of this model are taken from the knowledge of the general biology of the bacteria and the experimental setup with bioreactors. Emergence: The operating protocol or external manipulation of the bioreactor is decisive for the dynamics of the biomass production and behaviour of the bacterial population. These features arise from the growth, reproduction and viability of the individuals which are conditioned by local nutrient availability (affected by the entrance or not of fresh medium, and/or the output of the growth medium). Therefore, the individual activity is highly affected by the operating protocol of the culture. Adaptation: Individual adaptation of the individuals to the availability of the nutrient conditions found to their environment. Sensing: The bacteria are sensitive to the presence or absence of nutrient in its neighbour. Interaction: Interactions between bacteria are indirect; they are conducted by the interactions with surrounding medium. There may be competence for nutrient when availability is not sufficient. Stochasticity: It is introduced into the model when setting some individual actions: To fix the initial position of the individuals in the medium In the control of the minimum mass for reproduction and for the entrance of the latency state, using Gaussian distributions around the expected mean values. In the position of a new bacterium during the reproduction process. To choose the patches to perform the output of the medium. To perform the shake of the medium in order to change the position of the bacteria. Observation: Number of individuals as a counter, and temporal evolutions of the total biomass, the amount of nutrient and the amount of end product. 3. Details Initialization: The input data builds the system in its initial state, generating the "world" with spatial characteristics, a distribution of nutrient (each patch has an amount of nutrient obtain from a normal distribution with a specific mean value and standard deviation fixed by the user) and a population with initial individual characteristics for each bacterium: mass, a mean biomass for reproduction and a minimum mass to entry in the latency phase, that can be fixed by the user. The inputs that have to be set up are initial number of bacteria, initial amount of nutrient. Possible operating protocols are: nutrient entrance (there is the option to open or close the entrance of the nutrient in each time step, flow-on or flow-of), stirring or shake the system, and output of the medium (nutrient, product and biomass), the strength of the shaking of the bioreactor and the diffusion coefficient for the nutrient and product can be fixed. Input data: Not considered. Submodels: At each time step the model considers the following individual actions: ii) Uptake and growth. The quantity of nutrient that a bacterium may absorb is assumed constant and it is limited by the available nutrient of its own spatial cell, this nutrient increase its own biomass. Otherwise, if there is not nutrient the bacterium entries in the latency or hibernate state. iii) Reproduction is done by bipartition. Each individual cell has a particular mass to initiate this process, which is given by the mean value chosen by the user and a Gaussian around it. Once this mass is achieved, the bacterium splits, and two new bacteria appear. One remains in the same location and the other chose one of the neighbour patch, each one having half of the progenitor biomass. iv) Cell viability is controlled. If the individual does not find nutrient it change to the latency state. As soon as the bacterium can achieved nutrient, its leaves this latency state and became viable again. i) Movement. A bacterium can move randomly from its spatial cell to a one of the 8 adjacent spatial cells only if it has no nutrient in its own patch, and it decreases of own mass with this action. Diffusion of nutrient and end product is performed in the medium. Stirring or shaking the culture can be decided by the user. The entrance of fresh medium identified as the entrance of nutrient is carried out according to the handling instructions given by the user at the beginning of the simulation or during th extracting process. When a specific amount of nutrient has to be introduced, there is a shaking process, and an amount of the medium is taken out of the system (this is done with the extraction of an amount of nutrient and end product, as well as a set of bacterial cells chosen randomly). Flow chart: