Experimental Design
In the first survey, a sample of 400 U.S. adults will be recruited via Prolific to take a 5-minute survey. In the survey, participants will be given information about the study and then asked if they consent to participate in the study. If they consent to participate, they will proceed with the rest of the survey. Participants will be shown brief background information about recovery housing and SUD. They will then be told that they are being asked to predict how people in another study will respond to four images of people pictured below.
They will then be shown four images of people that were AI generated. Images include one black man, one black woman, one white man, and one white woman. After seeing the images, participants will be asked to predict how many participants in the second study will support the establishment of recovery housing at the state and local level, will agree that the individuals pictured are likely to stay in recovery, and will support a licensing policy for recovery housing. Then participants will be asked a series of demographic questions including their age, gender, race, and experience with SUD. They will be told that they may earn an additional $100 from participating in the study. Specifically, they will be told that the study researcher will randomly select one of the prediction questions below, and the participant whose prediction is closest to the actual outcome of the second survey will get $100.
In the second survey, a separate sample of 4,000 U.S adults (3,500 main sample + 500 rural sample) will be recruited via Prolific to take a 12-minute survey. In the survey, participants will be given information about the study and then asked if they consent to participate in the study. If they consent to participate, they will proceed with the rest of the survey. Participants will be asked a series of demographic questions including their age, what state they live in, they zip code and residential county, and if they live in a rural or non-rural area. Participants will then be provided background information about SUD and recovery housing.
Participants will then be randomized into one of nine treatment conditions. If participants are randomized into the control condition, they receive no additional information and proceed to the next steps of the survey. If participants are randomized into one of the four picture conditions, they are shown an AI generated image of a person (either a white woman, black woman, white man, or black man) and asked a series of questions about how they feel about the person portrayed in the image. If the participants are randomized into one of the four story conditions, they are shown an AI generated image of a person (either a white woman, black woman, white man, or black man) and asked to read a personal story of SUD recovery. They are told that the personal story of recovery is real and that the person in the story looks similar to the AI generated picture provided. They are then asked to rate how they feel about the person portrayed in the image and story.
After being randomized into one of the treatment conditions, participants will be asked if they support the establishment of a recovery housing in different locations, if they support different policies related to SUD and recovery housing, and how they feel towards those with SUDs. Participants are then asked a 20-item adapted version of the Community Attitudes towards Mental Illness scale to assess levels of social SUD stigma. Finally, participants are asked a series of demographics questions including their gender, race, education, income, political leaning, employment profession, and religion.