Nudging Resisters Toward Change: Self-Persuasion Interventions for Reducing Attitude Certainty

Abstract

Purpose: To identify effective self-persuasion protocols that could easily be adapted to face-to-face clinical sessions or health-related computer applications as a first step in breaking patient resistance. Design: Two self-persuasion interventions were tested against 2 controls in a between-subject randomized control experiment. Setting: GuidedTrack—a web-based platform for social science experiments. Participants: Six hundred seventeen adult participants recruited via Mechanical Turk. Intervention: The experimental interventions prompted participants for self-referenced pro- and counterattitudinal arguments to elicit attitude-related thought (ART) and subsequent doubt about the attitude. The hypothesis was that the self-persuasion interventions would elicit larger and more frequent attitude certainty decreases than the controls. In the experimental groups, we also predicted a correlation between the amount of ART and attitude certainty decreases. Measures: Changes in attitude certainty were measured by participants’ pre- and post-ratio scale ratings; ART was measured by the number of words participants used to respond to the interventions. Analysis: Analysis of variance (ANOVA), χ2, and correlation. Results: A goodness-of-fit χ2 showed that the number of participants who decreased their attitude certainty was not equally distributed between the combined experimental groups (n = 104) and the combined control groups (n = 39), χ2(1, n = 143) = 28.64, P < .001. Within each intervention, goodness-of-fit χ2 with a Bonferroni correction ( P = .01 or .05/4) indicated there were significantly more “decreasers” than “increasers” in intervention 1, χ2(1, n = 86) = 6.16, P = .01, but not intervention 2, χ2(1, n = 84) = 2.02, P = .16, the nonsense control, χ2(1, n = 42) = .22, P = .64), or the distraction control, χ2(1, n = 34) = .02, P = .89. A 1-way ANOVA revealed a significant main effect for intervention on mean certainty change ( F3,613 = 4.62, P = .003). Five post hoc comparisons using Tukey’s honest significant difference (HSD) test indicated that the mean decrease in attitude certainty resulting from intervention 1 (M = −3.29) was significantly larger than the mean decrease in attitude certainty resulting from the nonsense control (M = −0.62, t = −2.72, P = .03), the distraction control (M = 0.11, t = 3.48, P = .003), but not intervention 2 (M = −0.87, t = −2.54, P = .06). Attitude-related thought was significantly correlated with attitude certainty change in intervention 1, r(158) = −.17, t = −4.28, P = .02, but not intervention 2, r(161) = −.002, t = −.03, P = .98. Conclusion: The implication for clinical practitioners and designers of health applications is that it may be worthwhile to let patients elaborate on their personal reasons for initially forming an unhealthy attitude to increase doubt about the strongly held attitude.