Thanks to everyone for your interest!
The paper is still in the process of being reviewed, but I am very happy to post a link to once it’s up.
As a quick summary:
The study compared 2 web applications for tracking mood & activities. The 1st web application allows users to submit records of mood, activities and text summaries. Users are then presented with visualizations of mood over time (past mood records, today’s records, 2 days of predicted mood).
Users can additionally plan activities into their schedule and see the anticipated impact of those activities on their mood. Participants were presented with 10 activity recommendations to possibly plan. Half of these were algorithm-generated recommendations that were individualized for the specific user, based on their logfiles. Half of the activity recommendations were based on an emotional well-being theory (ryan and deci) and aimed to improve perceived autonomy, competence, sociality. I split activity recommendations into these 2 versions so we could have some benchmark to compare an algorithm against a relatively best-effort human-generated version.
We compared the mondo version of the application with one in which participants only track mood and activities, but did not see the visualization or activity recommendations.
The main findings were -
-For the mondo version of the application - improvements in logfile mood over time, though no changes in pre-post wellbeing scales.
- For the activity recommendations, we found that participants who more often selected activities based on the well-being theory had better outcomes to those who picked ones generated by an algorithm.
The paper we submitted includes another study about well-being consequences for logging only negative mood records or logging only positive mood records for 1 month. Basically, we got the very unsurprising result that logging only negative events corresponds to decreased scores on well-being scales pre-post intervention. Positive recording only showed the opposite direction.
For the first application I described, we want to make some improvements to the algorithms and then ideally have it publically available for anyone to use. Very happy to share a link later if you want to try it out.