Maybe the same as with drugs in sports: Self-experimentation can be an expedient shortcut, and scientists are often very competitive people. If results obtained through self-experimentation are rewarded, many scientists would be tempted to do it. Contrary to doping in sport, however, in science you need to at least do something different each time for it to be publication-worthy. That institutes a big skill floor and considerable risk, so I think a self-experimentation epidemic is unlikely. Generally I still think self-experimentation is good, precisely because it's such a shortcut.