Why Your Algorithm Settles for the Wrong Answer

Fixed Points and Strike Mandates

In program analysis, fixed-point computations often converge to the least or greatest fixed point depending on initial assumptions. The author notes that humans consistently develop algorithms that converge to one extreme, as if we have a common blind spot. For example, dead value elimination typically starts with all values live and prunes, yielding the greatest fixed point, but we actually want the least. The same pattern appears in student union strike mandates in Québec: starting with unions already on strike leads to deadlock, while starting with all mandated unions and removing those not meeting conditions finds the greatest fixed point. The author argues that choosing the right initial value is crucial, and that settling for suboptimal solutions should be a deliberate choice, not an oversight.