The Information-Age Rumor Mill

This model is based on the original Rumor Mill model from the NetLogo Model Library.

Significant Changes

rumor target -- There is now a "target" of the rumor. This is who everybody is talking about.
interest attenuation -- An agent's "interest" in the rumor is modeled spatially. The further away from the rumor target, the less likely the agent is to be interested in retelling the rumor.
self-centeredness -- In a world with self-centered agents, they are less likely to spread rumors about others; perhaps because they are too busy talking about themselves.
the internet -- When the internet is available to the agents, they may choose to email the rumor to their friends.
spam-capacity -- Rumors are likely to spread more quickly when agents can email many of their friends in one shot instead of just a couple.

Model Analysis

The Experiment

The BehaviorSpace runs were designed to explore how interest attenuation and the availability of the internet affect the spread of the rumor. An important observation from the original model is that the rumor tends to spread in a linear fashion. How does the introduction of these new parameters affect the rumor's spread?

Behavioral Regimes Observed

Availability of the internet leads to an exponentially growing rate rumor spread once it takes hold.
- Without the internet, the rumor tends to spread at a linear rate once an initial core of agents comes into the know. The internet allows distant agents to learn the rumor--these agents proceed to spread the rumor by word of mouth and by email, possibly to previously untouched pockets of agents, who in turn do the same. Since the population of agents to whom the rumor has not yet been told is finite, the rumor's spread always trails off towards the top of the curve no matter how the rumor is being spread.
Spam capacity accelerates, but does not fundamentally change trends seen in the model.
- Although a higher spam capacity certainly spread the rumor faster, it does not change the structure of the growth pattern. The key to the exponential change in spread rate with the internet is the availability of agents who have not yet heard the rumor. These pockets of agents will still be found with a low spam capacity, but not as quickly.
Interest attenuation has a limiting effect on the exponential growth experienced with the internet.
- Because it limits the population of agents, high attenuation will return the spread of the rumor in the information age closer to the level of the traditional, pre-internet model.

Chart -- note that Blue is no internet, no attenuation; Pink is internet with attenuation, and yellow is internet with no attenuation. I'm not sure why Excel won't export the key correctly...

The Model

created with NetLogo

view/download model file: Info-Age Rumor Mill.nlogo

WHAT IS IT? ----------- This program models the spread of a rumor about a specific person in the population (the "target" of the rumor). The rumor is spread in an Information Age environment, where email is available to all members of the population. The rumor can spread in two ways: a person who knows the rumor can tell one of their neighbors, or they can email the rumor to one or more friends. The less interested a person is in the the target of the rumor, the less likely they are to share the rumor with others. Spacial proximity is used to determine how interested the person is. In other words, the closer a person is to the target of the rumor, the more interested they are and thus the more likely they are to share the rumor. The neighbors can be defined as either the four adjacent people or the eight adjacent people. At each time step, every person who knows the rumor decides whether they are interested enough to share it. If so, they randomly choose a neighbor to tell the rumor to. They may also choose to email the rumor to a random number of friends. The simulation keeps track of who knows the rumor, how many people know the rumor, and how many "repeated tellings" of the rumor occur. It also keeps track of how many emails have been sent. HOW TO USE IT ------------- WRAP? is a switch which when set on allows the rumor to wrap top and bottom and left and right, as if the grid were on a torus. When set off, the rumor spreads as if the grid is bounded, without wrapping. EIGHT-MODE? is a switch that determines whether at each time step the rumor spreads to one of four randomly chosen neighbors, or one of eight such neighbors. WITNESSES-ALWAYS-TALK? is a switch that determines whether the rumor seeds (the "witnesses") will always decide to share the rumor, no matter how interested they are. This only applies to the members of the clique who join during setup. SELF-CENTEREDNESS is a slider that determines how self-involved members of the population are. The more self-involved they are, the less likely they are to spread rumors about others (they are too busy talking about themselves). INTEREST-ATTENUATION is a slider that determines how distance from the target of the rumor affects a person's interest in the rumor. The higher attenuation is, the more quickly interest drops off as distance increases. INTERNET is a switch that determines whether the internet (and email) is available for sharing rumors. Turning off this switch reverts the model from an Information Age model to a more traditional Rumor Mill model. DESIRE-TO-EMAIL is a slider that determines how likely a person is to email a rumor that they are interested in. The higher the value, the more likely the person is to email. SPAM-CAPACITY is a slider that determines how many emails a person is capable of sending out at a time. The sender will send a random number of emails between zero and spam-capacity when sending email. REPEAT-EMAILS? is a switch that determines whether people will send out batches of email containing a rumor more than once. When off, for example, Joe may send an email to four of his friends upon hearing the rumor, but next time he hears the rumor he will refrain from emailing. As with any rumor, it has to start somewhere, with one or more individuals. These can be thought of the "witnesses." There are three ways to control the start of the rumor: 1) Single source: Press the SETUP-ONE button. This seeds the initial witness at one point in the center of the screen and chooses a randomly-located target. 2) Random source: Press the SETUP-RANDOM button with the INIT-CLIQUE slider set greater than 0. This "seeds" the witnesses randomly by choosing a percentage of the population that knows the rumor initially. This percentage is set using the INIT-CLIQUE slider. The target of the rumor is picked randomly. 3) Choose source with mouse: Press either SETUP-ONE or SETUP-RANDOM, then press the SPREAD-RUMOR-WITH-MOUSE button. While this button is down, clicking the mouse button on a patch in the graphics window will tell the rumor to that patch. To run the model, you can either "step" through each time step using the STEP button or allow the model to simply run continuously using the GO button. The model will stop when everyone in the population knows the rumor. There are three plot windows associated with this rumor model. RUMOR SPREAD - plots the percentage of people who know the rumor at each time step. SUCCESSIVE DIFFERENCES - plots the number of new people who are hearing the rumor at each time step. EMAILS SENT - plots the number of emails that are sent in any given period The monitor CLIQUE% is the percentage of the patches that have heard the rumor. The three coloring buttons to the right of the graphics window give you topographic maps of the screen. The COLOR: NORMAL button colors the screen green for the target of the rumor, red for people who have heard the rumor via word-of-mouth, orange for people who have heard the rumor via email, and blue for people who do not yet know the rumor. The COLOR: WHEN HEARD button colors the screen different shades of YELLOW according to the first time that location heard the rumor. The COLOR: TIMES HEARD button colors the screen different shades of GREEN according to the number of times that location has heard the rumor. THINGS TO NOTICE ---------------- It is interesting to note that the speed of the rumor's spread can be heavily dependent on whether those who know the rumor are interested in the rumor's target. A slowly-spreading rumor can explode if emailed to a particularly interested party. Experiment with interest-attenuation to see how this affects the spread. As with the traditional Rumor Mill model, the rumor spread slows down once a critical mass of very interested parties know the rumor. It can be interesting to compare the manner in which rumors spread with the internet available versus when it is not available. Desire-to-email and spam-capacity significantly affect the spread of the rumor in the Information Age, as does the interest-attenuation. THINGS TO TRY ------------- Adjust INTEREST-ATTENUATION to get more or less people interested in spreading the rumor. How does a high attenuation affect the likelihood of a rumor spreading wildly via the internet? Explore how changing peoples' ability to communicate with large number of people via the internet affects the speed of the rumor's spread. Why is a rumor more likely to take root when people have a higher email capacity? EXTENDING THE MODEL ------------------- Here are some suggestions for ways to extend the model. - Introduce a model of rumor accuracy. As the rumor takes more "jumps", the rumor's accuracy may decrease. How accurate is the version of the rumor that outlying members of the clique hear? - Change the model for interest-attenuation. How would a more complex model of a person's interest in the rumor affect its spread? - Introduce physical barriers into the simulation. These spatial barriers would be obstacles around which the rumor would have to spread. One could imagine a room where there was only a one cell entry. How long would it take to reach the entire population in this case? And how would that curve (the function of the number of people who know the rumor versus time) compare to the spread of the rumor when there was no such barrier? How would the presence of the internet affect such a model? NETLOGO FEATURES ---------------- Note the use of the "neighbors" and "neighbors4" primitives to implement 8-mode and 4-mode, respectively. RELATED MODELS -------------- Virus, AIDS CREDITS AND REFERENCES ---------------------- Modified by Greg Dunham Oct 9, 2003 from the original Rumor Mill model by Uri Wilensky (see below for full reference). This model is itself an extension of a physical experiment where spatial proximity was not a factor in the spread of the rumor. Contact Helen M. Doerr at hmdoerr@syr.edu regarding papers in preparation. Thanks to Dr. Doerr for inspiration for this model. To refer to this model in academic publications, please use: Wilensky, U. (1998). NetLogo Rumor Mill model. http://ccl.northwestern.edu/netlogo/models/RumorMill. Center for Connected Learning and Computer-Based Modeling, Northwestern University, Evanston, IL.