Figure 7
From: Misinformation does not reduce trust in accurate search results, but warning banners may backfire
Rank (X-axis) does not affect the evaluation of trustworthiness (Y-axis, mean-centered) of accurate results. This lack of relationship is robust across experiments (columns) and for clicked results (top row, red) as well as non-clicked results (bottom row, blue). The trend lines represent the predicted change in trustworthiness ratings per unit decrease in rank fitted by the linear regression models. Each dot represents 19 to 431 trustworthiness evaluations, depending on the rank, experiment, and click category shown. For instance, the third point from the left in Experiment 1’s top panel shows that participants who clicked an accurate result in the third rank on the page gave it trustworthiness scores that were about 0.25 scale points higher than the average trustworthiness rating that participants gave the results they clicked across all ranks in this experiment. All slopes for rank had ps > 0.05. The figure separates ranking for clicked (top) and non-clicked (bottom) to illustrate that the non-effect is robust for both types of results. Error ranges represent 95% confidence intervals of the centered trust ratings within the rank and click category for each experiment.
