step,rule,n_before,n_after,n_removed,rationale
1,assess all study items in studies.json,5,5,0,all five items have a printed value in the cited extract (k_extracted=5; n_analysed=5)
2,harmonisable onto theta=ln(T_AI/T_noAI),5,4,1,S5 removed: DiD coefficient on GitHub output counts around a ban is not a task-completion-time comparison
3,poolable (SE obtainable from printed quantities),4,2,2,S2 removed (CI only in an unreproduced figure); S3 removed (significance stars only)