A Special Report
Computer Simulations: New-Media Tools for Online Journalism

The Simulator
Test how different behavior patterns affect the spread of the HIV virus, with the Java-based HIV Simulator.

HIV Simulator FAQ
What about condoms? What's been left out? Is more sex really safer sex? Explore the nuances of the simulation in the Simulator FAQ.

The Code
Delve into the Java code that runs the simulator.

How the simulator works

A well-written newspaper story is easy to skim. You can read the first several paragraphs and gather the main point. A simulator, though, is so open ended that even running it several times can take too much time. This short summary is intended to report one set of data that backs up the essential point of the article: changing small details can have large effects.

Here are some graphs tracking the spread of HIV in six different runs of the simulator. The first shows the percentages of men and women infected when the rules of engagement in the model are the "relentless male." In this case, every promiscuous man searches for a second willing female until he finds one. This setting is most similar to Kremer's model, and as he predicts, the spread of HIV slows as more women in the model become promiscuous.

The blue line in these graphs shows the spread of HIV when 5 percent of the women are promiscuous; the light orange line reports the spread when 10 percent of the women are promsicuous; the dark orange line shows what happens when 20 percent of the women are promisuous. In each case, 100 percent of the men are promiscuous. Each simulation runs for 200 time units. As you can see, more promiscuity slows the spread of HIV, but in the end almost the same percentage of men and women are infected.

The second pair of graphs shows what happens when the men in the model approach 10 potential partners during each time unit and mate with everyone who accepts their propositions. In this situation, more promiscuity leads to a faster spread of HIV. As before, the blue, light orange and dark orange lines report what happens when 5 percent,10 percent and 20 percent of the women, respectively, agree to the men's advances.

Your results may vary. In each case, the male-to-female transfer rate of HIV was set to be 20 out of 1,000 and the female-to-male rate was set to be 10 out of 1,000. The simulation ran for 200 time units, and HIV-infected people died after 150 time units. No new people were born. The promiscuity factors were all set to be 100 percent, so neither men nor women turned down a potential liaison.

Time units in the simulation
This controls how long a simulation will last. A longer simulation takes more time, but may be necessary to examine the long-term effects of death on the spread of HIV in a closed population.

Number of couples
This simulation is strictly heterosexual. The programming is easier if you deal with homosexual populations, but the interesting questions emerge in the heterosexual world because of the different rates of transmission from men to women than from women to men.

A bigger number takes much longer to run, but it is potentially more accurate.

Male-to-female transfer rate:
In 1,000 encounters, the number of times an infected man will pass the disease to a woman. Setting this variable to 500 is the equivalent of flipping a coin. Some studies indicate that this rate can differ from the female-to-male transfer rate.

Female-to-male transfer rate:
In 1,000 encounters, the number of times an infected woman will pass the disease to a man.

Infection lifespan:
How long the infected people live. If they die off, then they stop spreading the disease.

Male promiscuity rate:
The percentage of men who are "promiscuous" -- that is, who spend their evenings approaching women to see if they are promiscuous. This is either true or false for each man in a couple. If the man is promiscuous, he will seek extramarital liaisons.

Female promiscuity rate
The percentage of women who are "promiscuous." The current version of the program assumes that all sexual overtures originate with men. This model was chosen to simplify programming, not to reinforce stereotypes.

Male promiscuity factor
The likelyhood that a man who has found a willing partner during any time unit will actually have sex with her. If this variable is set to 50 percent, you can pretend that the man flips a coin each time he meets a potential conquest and only procedes half of the time.

Female promiscuity factor
The likelyhood that a woman will accept an man's advances. If it is set to 50 percent, you can pretend she flips a coin. If she is not promiscuous, she automatically rejects the offer.

Relentless male versus ten-timing male
This is the most interesting switch. If you choose "Relentless Men," they will spend each evening looking for a liaison until they find it. If you choose "ten-timing," the men approach 10 women and sleep with each one who is willing.

This makes a big difference. If the men are relentless, then the spread of AIDS slows as the rate of promiscuity increases. But it only slows mildly. If the men are limited to approaching 10 women a night, then increasing the number of promiscuous women increases the spread of AIDS.