Agent-based model in Python

I am currently writing a paper in which I'm trying to develop an agent-based model that describes the process of language acquisition throughout the EU. I have a database from a survey from which I sample to get agents with the following properties: 1) country of residence; 2) nationality(ies); 3) age; 4) age when they finished education; 5) mother tongue(s); 6) profession (one of eight categories); 7) living condition (countryside, small city or big city); 8) first, second and third foreign language (if any) and related level of fluency (from 1, basic, to 3, very good). Agents have an additional property, whose value depends on their mother tongue(s). Based on their native language, each agent takes on a vector of values that defines the linguistic proximity between their native language and all official languages of simulated environment. At the beginning, no agents is learning a foreign language. Their status is "not learning". After the first iteration, agents start learning a language with a certain probability p and switch their status from "not learning" to "learning". At every time step, an agent learns the language she picks, that is, her level of fluency in that language increases. The speed at which she acquires competences in that language (that is, the speed at which her level in that language increases from 0 to 3) depends directly on the proximity between the language that she is trying to learn and the closest language among her native language(s) and the foreign language(s) in which she is fully proficient. The shorter the distance between the two languages, the less the number of time steps required by the agent to reach proficiency in the new language. When the agent has reached proficiency, she switches her status back to "not learning" and the process starts again. Every agent is programmed to be able to learn up to three foreign languages in total, including the ones that she already knows from the get-go. When asked to pick a new language to learn, agents can use one of three strategies: 1) they can pick the language that is closer to their native language, that is, the language they will learn in the shortest amount of time; 2) they can pick the language that has the highest number of native (L1) and non-native (L2) speakers; 3) a combination of strategies 1 and 2, that is, they can pick a language close to their own and having a relatively high number of L1 and L2 speakers. During the simulation, the model updates and keeps track of a number of values. It keeps track of the total number of L1 and L2 speakers of every language. This information is crucial and affects the model in two ways. At the micro level, it affects individual agents' decision about the language they should learn, if they are using strategies 2 or 3 in the list above. At the macro level, this information is necessary for the system to establish the OWLs of the language regimes. Indeed, the three regimes consider, respectively, the one, three or six most spoken languages. The model keeps track of the level of disenfranchisement, which is a direct consequence of the language regime. Finally, the model keeps track of the individual properties of the agents, with a view to identifying the socio-demographic categories most likely to be disenfranchised.