What Engineering True Cost Is

The first paper defined engineering. It identified True Cost as the qualifier that distinguishes engineering from academic work — the property that failure imposes harm on parties outside the engineering activity itself. The treatment of True Cost in that paper was binary. Either the activity has it, in which case the structural criterion can be met, or it doesn't, in which case the activity is academic regardless of how rigorous it is. The binary treatment was sufficient to draw the engineering boundary. It is not sufficient to answer a question the software industry has been failing to resolve for fifty years: should software engineers be credentialed, and if so, how? Attempts have been made. The IEEE Computer Society offers software engineering certifications. ABET accredits software engineering degree programs. Some jurisdictions have offered software engineering PE exams. Various national and international bodies have proposed licensure schemes. None of them have produced credentials the industry takes seriously, none have produced the kind of professional accountability structure that civil or electrical engineering has, and none have answered the basic question of who should certify what when software fails. Meanwhile, software increasingly controls systems whose failure kills people. Autonomous vehicles must decide between collision outcomes. Medical devices deliver treatment in ways where bugs end lives. Industrial control systems manage chemical processes whose failure can level neighborhoods. Avionics software flies aircraft. Surgical robots cut into bodies. Grid management software dispatches power to hospitals on backup. The regulatory regimes around these domains have evolved without ever clearly explaining why they certify what they certify. The software industry remains confused about whether its own engineers should be credentialed, by whom, and for what. This paper resolves the confusion. The resolution comes from unfolding True Cost into its domains. Once the domains are visible, the credentialing question answers itself.