-
Notifications
You must be signed in to change notification settings - Fork 2
XX 01 Model Description
In general, agents will change their mask behaviors based on their situation. Also, different professionals will have different proportion of wearing masks at the model initialization.
When self-infection happens, the agent is willing to change their masks to a better one.
Proportion of agents with mask for each profession at initialization. The sum of the proportion of each profession should be 1.
Infection rate will vary depending to different masks the agent is wearing.
This is a comprehensive introduction about the emotion model including details and design.
The emotion model adopts a four-stage emotional progression: indifference, worried, afraid, and numb, which named as the IWAN model. Each agent is assigned an emotion state which affects the behavioral activity of the agent and these four emotion states are also changed according to certain probabilities. Not only that, these transition probabilities also change according to external situations, and we compute a value called the concern level, which is used to change these probabilities. For example, if the agent feels more anxious about contagion and the concern level goes up, then the probability of going from the worried to the afraid state goes up.
The concern level acts as a dynamic scalar, adjusting transfer rates to reflect varying intensities of emotional transitions, thereby imparting a degree of flexibility to the model. It is worth noting that such modulations in concern level are not arbitrary but are influenced by a combination of external stimuli, both positive and negative. An in-depth exploration of these external factors and their quantifiable impact on the concern level will be detailed below:
Community Factor: the community factor reveals the infection increase or decrease rate, an increasing infection rate within the local community amplifies fear and concern among the population.
Family Infection: when a family member get infected, the whole family intend to raise the concern level.
Self-infection: when agents get self-infected, they intend to raise the concern level.
Crowding Factor: the crowding factor is used only an evacuation event is started and the agent has arrived to the evacuation center. The crowding factor is introduced because the agent will be afraid of get infected due to the high-density population. In this case, the crowding factor will increase the concern level. The crowdedness is calculated as current occupancy / max capacity.
Constant Decay: the constant decay is used for balancing the concern level.
As mentioned above, the transition probabilities will be controlled by concern level. To be more specific, if the concern level increases, the transition probabilities of indifferent->worried, worried->afraid, afraid->numb will increase, the transition probabilities of worried->indifferent and afraid->worried will decrease, vice versa.
Mask Forget Rate: the forget rate indicates the probability of forget wearing masks at the beginning of the day.
Mask Change Intention: the intention indicates the probability of changing their mask wearing state when they get infected.
In the previous model (ALIFE 2023 Version), the mask model is designed as a basic version: the agents will be assigned a fix mask change intention, which is a probability of increasing mask-wearing behavior when get self-infection, and change their mask-wearing behavior based on it.
In this version, we introduce more dynamics to the mask model. First, we connect the behavior of mask wearing and changing with the emotion states, so emotions are reflected in the action of wearing the mask. The mask will be taken off due to certain factors, which reflects the true situation more accurately.
In reality, agents can't wear masks all the time or wear the same masks all the time. Therefore, we need to design a mask-wearing behavioral intention to change agents' behavior of wearing masks. First, for mask forget rate, if an agent wears a mask on day t-1, it will choose whether to take off the mask or keep it in its original state on day t according to the set probability. For the choice of mask forget rate value, it will be determined by the emotional state of the agents, which have been listed in emotion.json file.
For the Mask change intention rate setting, the agent will only tend to change the behavior of wearing the mask if the agent itself is infected. For example, for agents in the WORRIED state, if they are infected or their infection deepens and they are not wearing an N95 mask, then there is a 30% chance that their mask behavior will be upgraded. Conversely, we can also set the intention rate of taking off the mask when agents are recovered, but in this case, the rate should be lower than upgrading the mask.
The above is the current logic, and more customized mask change intention can be set, such as
- when the agent's family has elderly people
- when entering densely populated or high-risk places: hospitals, shopping malls, evacuation centers, etc.
In fact, some of the factors here have already been indirectly taken into account by the calculation of the concern level, so further thinking is needed when expanding this part.
The current design of this part is simple, the evacuation events are only introduced to influence concern level, please refer to Crowding Factor in the part one.