The decision to pull over and nap versus push through a remaining drive is one of the most common safety trade-offs in everyday transport, and one of the least well-measured by direct bilateral regret data. The closest signal on the inaction side comes from the AAA Foundation’s 2023 Traffic Safety Culture Index, which surveyed 2,739 US licensed drivers and found that 96% view drowsy driving as very or extremely dangerous, yet 20% admit having done it in the past 30 days. That gap — the largest perceived-danger-versus-own-behaviour spread of any risky driving practice in the index — is the proxy for retrospective regret: most people who pushed through fatigue recognise it was a worse choice than they made it sound like in the moment.
The action side has no direct regret survey at all. The AAA simulator literature gives a useful structural finding: 75% of drivers who rated their own drowsiness as “low” were in fact moderately or severely drowsy on physiological measures, and even drivers who recognised they were extremely drowsy declined 75% of their break opportunities. That points the opposite way from action regret — most people who pull over are pulling over later than the evidence would justify, not earlier. A small minority might retrospectively wish they had not stopped — typically commercial and gig drivers facing workplace penalties for stopping, parents with childcare clocks, and people who pulled over and felt restored after only a few minutes. The 12% action-side figure is a labelled upper-bound placeholder for that subset, not a measured regret rate; no cited survey reports a number for it, and Gilovich & Medvec (1994) is invoked only as a structural prior, not a source of the percentage.
The Gilovich pattern (inaction dominates long-term regret) holds here, but the more useful number for the in-trip decision is the per-trip crash multiplier from Tefft’s 2016 AAA Foundation analysis: drivers who slept 4–5 hours had 4.3× the crash rate of well-rested drivers, and the multiplier rose to 11.5× for those who slept less than 4 hours. The per-trip choice to push through is more like driving at 0.05–0.10% BAC than like the small ordinary risks people often compare it to. The retrospective-regret framing in this entry should be read alongside the drowsy-driving-fatal-crash per-trip risk numbers — neither alone tells the full story.