Re:The Story Behind F.lux
by scott_s on Monday January 25, 2016 @09:52AM
(#10967441)
Unfortunately, that self-experiment has the problem of the subject (yourself) knowing which "treatment" you're getting. I'm not sure how to double-blind something like this, though. It may require a study where we record subjects' sleep patters, make them use a screen at night, but don't tell them that it's the screen colors that are under experiment.
[ Parent
| on HN ]
Empirically, I didn't notice and that almost screwed up the experiment; see the part where a driver change broke Redshift and I didn't notice for a month.
(As well, I would note that you may be overestimating placebo or expectancy effects here. In the meta-analyses I've seen, randomization is much more important than placebo controls.)
Interesting - if I understand correctly, you're saying that you didn't notice for a month that no color balance was happening. Then perhaps an even simpler experiment preliminary experiment could be done to just see how well subjects are able to guess if a screen has been color balanced at night. If most people can't tell most of the time, then that simplifies doing an efficacy experiment.
Not quite. The reddening is blatant, so if someone had asked me on any particular day if Redshift was on or not, I could always correctly answer. But it was randomized on a daily basis, so it is possible, even probable, for occasional long sequences like 100001 to happen where the shifting was on or off for multiple days in a row. What I'm saying is that I turned out to be paying so little attention to whether the shifting was on or off, certain that if it was off on a particular day that just meant the experiment was going as planned, that I didn't notice even 20+ days of a single setting.