Curious about the ‘declining hazard rate’. That would also look like ‘the apparent hazard rate is higher on shorter tasks’. Could be explained by the couple of reasons I gave in a footnote for longer tasks being incommensurably slightly ‘easier’: more tendency to include repetitive elements, more opportunity to take alternative sequential paths. Or could be explained by some constant ‘difficulty’ term that doesn’t scale with human time-to-complete—perhaps ‘understanding the spec’. Or if the shortest tasks systematically have their human time-to-complete underestimated (but you’d need a reason to suppose it’d go this way rather than the other).
Curious about the ‘declining hazard rate’. That would also look like ‘the apparent hazard rate is higher on shorter tasks’. Could be explained by the couple of reasons I gave in a footnote for longer tasks being incommensurably slightly ‘easier’: more tendency to include repetitive elements, more opportunity to take alternative sequential paths. Or could be explained by some constant ‘difficulty’ term that doesn’t scale with human time-to-complete—perhaps ‘understanding the spec’. Or if the shortest tasks systematically have their human time-to-complete underestimated (but you’d need a reason to suppose it’d go this way rather than the other).