All projects

LSA authoring studio

The internal tool where we built the skill graph, wrote exercises with a live preview, and read the statistics.

The adaptive test is only as good as its content, so most of the engineering went into the tool used to make it.

  • Skill tree editor. An interactive ECharts tree with a context menu to add, move and delete skills and to link prerequisites across branches. Whole trees can be cloned to build a variant for another audience.
  • Competence graph. A force-directed view of the same data that makes cross-links and orphaned skills visible at a glance.
  • Exercise editor. One editor per answer type with a live LaTeX preview. For image labelling you click on the picture to place the drop targets. A regex tester generates example strings that a pattern would accept, so authors can see what a free-text answer would let through.
  • AI review. Every exercise could be sent to GPT-4 with a prompt asking it to review the exercise the way an experienced math teacher would: wording, ambiguity, difficulty.
  • Statistics. Solve-time histograms, pass rates with median and percentiles per exercise, and a sunburst chart of root deficits across a whole class.