For a long time I’ve been fascinated by the correlation of personality scores with life outcomes, e.g. that conscientiousness is as correlated with GPA at the tertiary level of education as cognitive ability, (table 3, link). Without proper experiments, however, whether or not there may be in fact causally related, whether than simply associated cannot be confirmed. Which means there’s a data problem, specifically a lack of data problem.
Later, I discovered that one of the largest (n>2E6) data sets of self-reported (a potential problem in and of itself) personality tests was Jeff Potter’s Big5 (hosted here), which as evidenced by his Google Scholars profile is still in continuous use.
Taking the test, I realized there were two significant problems:
poor UX — it looks like a website from the 00’s and is difficult to use on mobile
and lack of context — removing the ability to save my data means that users can’t perceive changes over time or make comparisons to peers
Building on my brief experience of building Messenger bots, I felt that the Messenger UX was perfect for administering Likert style tests (5 buttons numbered 1-5 fit perfectly on mobile) and data persistence could be easily achieved via attaching user data to a Facebook ID (app) and otherwise through Login With Facebook (web).
I discontinued development in mid-2018 over the breaking Cambridge Analytica story.
Answering a question
Messenger’s Quick Response buttons lend themselves particularly well to Likert style quizzes
Seeing results in Messenger
rendered as an SVG, posted to Amazon S3, and sent through Messenger as a signed URL
The web app
redirected from the app via Login with Facebook
‘Take a Test’ redirects back to Messenger
was planning to implement comparison features
Viewing results on the web app
shows specific answers