Why Stop Now-Mao

This note last modified September 1, 2024

#notesFromPaper Year : Tags : citizen science - engagement Authors: Mao Kamar Horvitz

a session is contiguous work, an aggregated session is contiguous work w/ breaks

people usually work, then come back after 24 hours. If they don’t, they are far less likely to come back.

seeing interesting tasks initially instead of average tasks makes people more engaged

if a player is spending longer or varying their votes less, it indicates that they are less engaged

how do you characterize workers?

  • time spent on tasks
  • # of past sessions
  • average time spent on session
  • time of day when starting a session
  • # of short breaks taken
  • length of contiguous sessions
  • gap between sessions? How does this compare to other gaps
  • rolling average of time taken per task? How does this compare to previous rolling averages for the same worker?
  • stats for current session? How does this compare to stats for the same worker in past sessions?
    • # of tasks completed
    • time spent

The most predictive thing was worker features combined with task features. Contiguous session features and aggregate session features weren’t as important

talks about how much data you would need to build a classifier like they did. I saw the number “one month” thrown out

also talks about generalizing their model