This semester, our fellowship did some research - in order to measure which presentations of charities attract the most interest to effective altruism. More specifically, we held giving games in Annenberg dining hall, where we gave students a dollar to donate to one of three charities. After they had donated, we asked them to sign up for our email list and that of The Life You Can Save. We presented different types of charities and presented charities with and without effectiveness information, and we measured the effect on email signups. Our main finding was that global poverty charities yielded a greater number of email signups than did more speculative charities.
We set up next to the exit of Annenberg dining hall and asked exiting students (mostly Harvard freshmen) to donate a dollar, which we provided, to charity. They did this by filling out an online form on one of our laptops, which presented them with three charities and prompted them to choose one.
We had three different sets of charities, of which we randomly selected one for each subject. The control set - Against Malaria Foundation, Schistosomiasis Control Initiative, and GiveDirectly - contains interventions in global health and poverty, described by a few bullet points that did not mention cost-effectiveness. The “effectiveness” set consists of the same charities, with descriptions extended by one bullet point to comment on the charity’s cost-effectiveness; for example, “According to GiveWell, a third-party evaluator, distributing deworming medicine to one child costs $0.73-0.99, with SCI paying about 70% of these costs.” Finally, the “speculative” set - Against Malaria Foundation, Machine Intelligence Research Institute, and 80,000 Hours - contains one option in global poverty, one in in existential risk, and one in meta-EA; described without mentioning cost-effectiveness
After choosing a charity, the subject returned the laptop and was asked by the experimenter (an HEA member) why they chose what they did. The experimenter then briefly described HEA and asked whether the subject would like to sign up for our email list or that of The Life You Can Save. We measured the number of signups for each list.
|HEA email signups||36||28||26|
|TLYCS email signups||21||16||13|
In informal terms, we found that 62.1% of subjects signed up for HEA’s email list in the control group, compared to 47.5% in the cost-effectiveness group, and only 41.9% in the group with more speculative charities. The difference between the control and speculative groups was strong enough that the result is statistically significant. The results were qualitatively similarly (but with fewer overall signups) for TLYCS’s email list.
We used two-tailed Fisher’s exact tests, which give the following p-values and relative risks. (A relative risk of 0.5 means that the intervention group is half as likely to show the effect as controls.)
|p (Fisher’s exact)||Relative risk (95% CI)|
|Speculative, HEA||0.03||0.68 (0.47-0.96)|
|Effectiveness, HEA||0.14||0.75 (0.52-1.07)|
|Speculative, TLYCS||0.07||0.67 (0.42-1.06)|
|Effectiveness, TLYCS||0.32||0.80 (0.53-1.23)|
Presenting Harvard students with more traditional, global poverty charities appears to generate more interest in effective altruism than does presenting them with charities in existential risk and meta-EA. In our case, we received significantly more email signups when presenting global poverty charities.
It is hard to draw conclusions with respect to the inclusion of cost-effectiveness information. We don’t know whether the decrease in signups resulted from having cost-effectiveness information or from the blurbs just taking longer to read; this was a flaw in the design of our experiment.
Next time we do giving games to generate interest in EA, we’ll keep these results in mind! Thanks for reading!