Bainbridge’s November 1983 analysis of automated industrial processes distinguishes detailed checking from timely response. If people must inspect computer decisions, the methods, decision rules and pace must be followable, even at some loss of technical efficiency. Otherwise they may be unable to check each decision while the process runs. For failures within seconds without advance warning, she argues for reliable automatic response; if unavailable and failure consequences are unacceptable, the process should not be built. Slower failures or a safely stopped process allow different roles. This historical argument counters blanket manual-approval requirements, not all human involvement. Applying it to research requires specifying the real deadline and what can safely pause; the paper did not test scientific workflows.