Risking It All for Love? Resetting Beliefs about HIV Risk among Low-Income South African Teens

..."'gamifying' information acquisition may be a far more effective way to deliver information about matters - such as HIV risk - which is important for individuals to understand but which many do not, with adverse consequences for their lives."

Prior research suggests that the much higher HIV prevalence among young women in sub-Saharan Africa than among males of their age cohort is linked to the high prevalence of age-disparate sexual partnerships, and that incorrect beliefs about the relationship between age and HIV risk are partly responsible. Driven by an interest in the formation and updating of risk perceptions about age-disparate sex, the researchers of the present paper conducted an experiment with a sample of low-income adolescents in Cape Town, South Africa to test the effectiveness of an information-based intervention at correcting the prior beliefs of South African teenagers about the relative risks of being exposed to HIV depending on the age of their sexual partner. This paper reports on the results.

Following an introduction - which, among other things, explores the mechanisms through which age-disparate partnerships increase HIV risk for women - Section 2 reviews evidence on interventions to tackle age-disparate sex and related behaviours among teens. Section 3 looks at the behavioural economics literature that helps explain why teens might have incorrect assumptions about the HIV risk posed by potential sexual partners of different ages (e.g., drastically under-weighting the past sexual behaviour of older men), and it briefly describes findings from a series of focus group discussions carried out with low-income teens in Cape Town as a precursor to the experiment.

Section 4 describes and provides the rationale for the intervention, and Section 5 describes how the experimental subjects were recruited and how the experiment was conducted. In brief, subjects were randomly assigned to a treatment or control group. Control subjects read through a brief essay about HIV and sexual risk, which included a brief discussion of relative risks by age. In contrast, treatment subjects played 10 rounds of a computer-based "HIV risk game". In each round, subjects were asked 2 questions, the first of which asked them to identify which of a 20-year-old man and a 30-year-old man was more likely to have HIV. The second asked the identical question but changed the gender of the people in the question. (The pattern of HIV prevalence by age in South Africa means that the correct response in either case would be to identify the older man/woman as being more likely to have HIV.) Treatment subjects received immediate feedback as to whether or not they had guessed correctly.

As explained in Section 6, the sample consists of 162 young people, most of whom were 15-19 year olds and 58.6% of whom were male. Section 7 describes key results. First, treatment subjects are significantly more likely to correctly identify which of 2 individuals is more likely to have HIV than control subjects are. The effect size is largest when female subjects answer the question about which of 2 women are more likely to have HIV. Second, treatment subjects answer more questions about HIV risk and age correctly than control subjects do. Only 7% of those in the treatment group get both questions asked wrong, compared with 35% of those in the control group; meanwhile, 63% of those in the treatment group get both questions right, compared with only 28% of controls. Regression estimates indicate that the treatment increases the mean number of questions a subject answers correctly by around 0.6. A follow-up survey 3 months later with no significant difference in attrition rates between those in the treatment and control groups shows substantially higher information retention among treatment subjects than among control subjects.

Section 8 concludes with a discussion of these results and implications for further research. One of the key takeaways is that information is more readily comprehended and retained when it is generated by individuals themselves (as in the computerised game), rather than provided ready-made (as in the traditional brochure). The researchers suggest that the prototype game they tested is potentially scalable at very low cost, since it is a simple computer-based game that can be played with little or no supervision. They contend that further research should focus on measuring effects on longer-term learning and retention as well as behavioural change in the domain of risky sexual behaviour, care-seeking, teen pregnancy, and other related outcomes.

In conclusion, this study found that playing the "HIV risk game" with repetition and instant feedback leads to substantially more accurate perceptions of the relationship between HIV risk and age than equivalent information being provided through a more traditional "brochure approach", especially when initial beliefs are relatively less accurate.