Writing for a rule you oppose

Introduction: People sometimes act in ways that don't match their attitudes. Researchers tested whether writing an essay that argues against your own opinion changes that opinion, and whether it matters if you feel you chose to write it.

Participants: 120 first-year students (aged 18 to 21; 70 women, 48 men and 2 nonbinary students) at a large public university in Texas. Four weeks earlier, on an online survey sent to all first-year students, each had rated a proposal requiring every student to complete 20 hours of campus service per semester. Only students who opposed the proposal (a rating of 1 to 3 on a 9-point scale) were invited. They described their race/ethnicity as 38% White, 33% Hispanic or Latino, 14% Asian, 11% Black and 4% multiracial.

Method: An institutional review board approved the study, and participants gave informed consent. They were told the study was about "how students write persuasive arguments". Participants were randomly assigned to one of three conditions (40 per condition). In the high-choice condition, the researcher said: "We already have plenty of essays against the service requirement. It's completely up to you, but it would help us a lot if you wrote one in favor of it." All agreed. In the low-choice condition, the researcher said: "You have been assigned to write an essay in favor of the service requirement." Students in both conditions wrote for 10 minutes. In the control condition, students wrote for 10 minutes about their favorite place on campus.

All participants then rated the service requirement again on the same scale, from 1 (strongly oppose) to 9 (strongly favor). Afterward, the researcher explained the real purpose of the study, explained why it had been described differently at the start, and offered participants the chance to remove their data. No one did.

Results and Discussion: The table shows the mean ratings. After writing, the high-choice condition rated the requirement more favorably than the low-choice condition, with an effect size of 1.10. The low-choice and control conditions did not differ significantly.