Similar to @Agaricus and @madprime I’m currently preparing some DIY Experimentation I want to perform over the next weeks. My goal is to look into whether listening to music influences my mood. As an experimental setup I want to alternate between days in which I’m allowed to listen to music ad libitum and days in which I’ll not listen to music at all.
My plan
Test between two conditions:
Condition A: listen to music during the day as I’d usually do
Condition B: don’t listen to music at all
Alternate every day (ABABAB), but I haven’t decided on a length of the experiment yet
Track “adherence” (sticking to plan) by looking at my Spotify listening history that I’m already importing into Open Humans (This would also allow for a non-binary analysis of data as I’ll have data about what I’m listening to and how much)
Track “outcomes” via Mood tracking through iMoodJournal (I’ve settled on this thanks to the work of exploring mood tracking apps done by @madprime).
Over the last few weeks I’ve done so preliminary mood tracking already just to get a hang of it and get some baseline data. During this I’ve discovered a couple of things related to the mood tracking:
First thing I noticed was that there seems to be a intra-day correlation between time of day and mood: Basically my mood seems to improve over the day. I think this is partially an effect of being stressed in the morning when waking up to the email deluge that came in from different time zones while I was asleep.
my initial plan was to record only 3 mood data points per day (morning, afternoon, evening) triggered by notifications on my phone. I thought this would be easier to stick to, as it avoids notification fatigue. Instead I noticed that I still miss many of the prompts. I think it’s most likely when the prompts come up at a time where I’m not having time to watch my phone. And by the time I get back to my phone the notification has virtually disappeared as it’s so far down the list of notifications that I forget about it.
To avoid this problem I’ve increased the frequency of reminders to an hourly notification between 08:00 and 22:00. This seems to work better as I get more data points per day.
Another problem I encountered over the last few days: I’m highly likely to miss data points when I’m out for meetings or being around friends etc. I think this might become problematic, as my gut feeling is that these missing data points are not of a random mood but would be biased as I’m more likely to be in a good mood when hanging out with friends etc. This might not be a problem for this experiment as long as there’s no difference in whether I’m meeting people between the two conditions though.
I was wondering how other people are approaching their mood tracking to get good data?!
For now my plan is to record some more days of base-line data with the higher sampling interval to get an idea of a ‘neutral’ distribution before starting the experiment. I’ll keep you folks updated!
I’m completely oblivious to my own mood, so have been thinking about doing the opposite, i.e. using my music listening patterns as an indirect measure of my mood
That’s an interesting idea as well! If you’re using Spotify to listen to music you might want to check out the Spotify integration for Open Humans. In addition to storing the songs you listen to it also grabs their classification metadata for songs (e.g. the ‘valence’ and ‘energy’ of songs). If you scroll down here there’s some details on the kind of data you can get through it: https://exploratory.openhumans.org/notebook/28/
Neat, didn’t know Spotify had all that additional data! Unfortunately, I use Google Play Music. This can be “scrobbled” into Last.fm, but the only additional information I get from there is the music genre for each track.
Does Spotify tell you how long a track was played? How inclined I am to listen each track to the end could be another interesting data point, also missing from Last.fm (all you get is the track length).
I am currently reading Why You Like It, by one of the people behind Pandora’s Music Genome Project. Possibly relevant
Unfortunately Spotify doesn’t tell you how long a track was played. But I think there’s a minimum duration you need to have listened to a song for it to be counted by them (so songs you skip more or less directly don’t show up in your data). I think someone had started working on getting a Last.fm integration set up for Open Humans, but if they don’t provide a lot of metadata about the songs that might be less useful.
And thanks for the book recommendation, goes right onto my to-read-list!
Looking at more data now it seems that this is a real problem and I’m not sure how to best solve it. I faithfully record my data points when working in my office, sit around and read a book, am out for a walk by myself etc. But once out in a social situation I’m super likely to miss my data points as I’m tuning out notifications and/or it would be rude to interrupt conversations for grabbing my phone to enter data points. E.g. last weekend I was out meeting folks all weekend and got at best a single data point per day.
Does anyone have personal experience on how to get a more even data set that’s less biased?
Could you get people around you to rate your mood? e.g. using the t-shirt equivalent of this:
More seriously, I suspect that while mood tracking can help increase awareness of your own mood, the data itself is unlikely to be precise enough to detect correlations beyond the blatantly obvious.
For the T-shirt idea, I would prefer a QR code to a phone number. With that said however, it seems to me that that idea might end up being more effective for diagnosing RBF (Resting Bitch Face) than assessing actual mood. YMMV
According to this research on music listening habits, we have now (February) reached the saddest (i.e. lowest “valence”) month of the year… Interestingly, that’s true for the southern hemisphere as well!
Thanks so much! It was really interesting to see this effect! It’s interesting to note though that the effect for most Southern hemisphere countries is much much smaller and in case of Argentina even inverted (February slightly happier than July).
I took this as an opportunity to look into the same effect for my own Spotify listening history!
The effect for my own data seems a lot less pronounced. And if one goes to just comparing February & July I see the same inversion of the effect for myself. The music I listen to in February is happier than the one in July!
