Earlier quoted context omitted.
That's true. However, most A/B testing doesn't result in an academic paper, and there are strict ethical rules about academic work based on human subjects. Second, obviously not all A/B tests are the same. If you're testing two different fonts and comparing bounce rates, there are far fewer ethical concerns than if you're intentionally spreading an "emotional contagion" (that's the term Facebook used). And note that…
There is nuance to this. The point of IRB is ethics. 1. If the experiment is minimal risk and expected to benefit participants (as the notorious fb experiment was—it was emotional contagion of positive emotions), then informed consent may be waived. 2. When there is a greater risk from the act of gathering informed consent (e.g., because you need to store identity for that purpose), then informed consent may be waive…
No, they tried both positive and negative.
> When positive expressions were reduced, people produced fewer positive posts and more negative posts; when negative expressions were reduced, the opposite pattern occurred.
> If affective states are contagious via verbal expressions on Facebook (our operationalization of emotional contagion), people in the positivity-reduced condition should be less positive compared with their control, and people in the negativity-reduced condition should be less negative.
- https://www.pnas.org/content/111/24/8788.full
The experiment was expected to make half of the subjects less positive.
Edit: This discussion of whether consent could be waived is sort of irrelevant when Facebook didn't even have an IRB at the time, and Cornell's IRB inexplicably declined to review the study even though two of the authors were part of Cornell.