RCTs
RCTs force a belief to survive comparison. Random assignment separates a treatment’s effect from selection bias, confounders, observer bias, and the stories people tell after seeing an outcome.
A doctor received the 1949 Nobel Prize for developing lobotomy. By 1951, roughly 20,000 people in the United States had undergone the procedure. FutureIQ’s provocation is brutal: if lobotomy had faced a randomized comparison in time, its prestige might not have survived.
E1 R1Make the groups differ by one thing
Ordinary statistics can show that people who received a treatment did better. They cannot tell you whether the treatment caused the difference. Participants may have selected themselves, begun with different advantages, received different attention, or been judged by observers who expected improvement.
Randomization attacks that ambiguity before results appear. It assigns treatment independently of the traits that might influence the outcome, creating a comparison group that should differ systematically only in exposure to the intervention. The outcome gap can then challenge the proposed mechanism instead of merely reflecting selection, confounders, or observer expectations.
E1 R1Where it shows up
Two millennia of bloodletting
Bloodletting persisted in Europe until the late nineteenth century and was still recommended in a 1923 medical text. George Washington reportedly lost 3.75 litres of blood in ten hours while being treated for a throat infection and died. A long history of use supplied tradition and anecdotes; it did not supply a randomized comparison.
R1The wellness-program trap
Workplace programs offering exercise, weight loss, smoking cessation, tai chi, and financial wellness sound beneficial. But if volunteers outperform nonparticipants, the program may be receiving credit for the motivation or circumstances that made those employees volunteer. Random assignment is what separates participation effects from participant selection.
E1The gold standard has boundaries
Not every question can or should be randomized. A parachute trial is impractical, and a trial can become unethical when a potentially dangerous treatment is assigned to one arm or an effective treatment is withheld from a placebo group. Randomization strengthens causal comparison; it does not erase feasibility or ethics.
R1Design the comparison before collecting praise
When testing a program that has more eligible people than available places, allocate access randomly and define the outcome in advance. Compare those offered the program with those not offered it; do not treat the enthusiasm or later success of self-selected participants as proof.
E1Episodes that teach this
- The Randomness of Science - Why Randomised Controlled Trials (RCTs) Work - FutureIQ start here 4,109 views