|
Field
Abstract
|
Before
This study examines whether individuals’ responses to artificial intelligence depend on the identity dimension through which they interpret technological change. Motivated by a two-dimensional framework in which individuals can identify with an economic group or a cultural group, participants first report their expectations about the consequences of AI and are then randomly assigned to reflect on their economic identity, cultural identity, or a neutral/no identity. We examine whether inducing identity along different dimensions changes AI-related opinions and donation allocations across organizations addressing different economic or cultural consequences.
|
After
This study examines whether individuals’ responses to artificial intelligence depend on the identity dimension through which they interpret technological change. Motivated by a two-dimensional framework in which individuals can identify with an economic group or a cultural group, participants first report their expectations about the consequences of AI and are then randomly assigned to reflect on their economic identity, cultural identity, or a no identity baseline. We examine whether inducing identity along different dimensions changes AI-related opinions and donation allocations across organizations addressing different economic or cultural consequences.
|
|
Field
Last Published
|
Before
September 05, 2026 03:25 PM
|
After
September 06, 2026 07:34 AM
|
|
Field
Intervention (Public)
|
Before
Participants are randomly assigned to one of three treatments that differ in the identity dimension on which they are asked to reflect. In the economic-identity treatment, participants identify with an economic group and reflect on experiences associated with that group. In the cultural-identity treatment, participants identify with a cultural group and reflect on experiences associated with that group. In the baseline, participants either identify as a cat/dog person and complete corresponding reflection questions or do not have a reflection stage. The interventions are designed to vary the identity dimension through which participants evaluate the consequences of artificial intelligence.
|
After
Participants are randomly assigned to an economic identity reflection treatment, a cultural identity reflection treatment, or a no reflection baseline. In the economic-identity treatment, participants identify with an economic group and reflect on experiences associated with that group. In the cultural-identity treatment, participants identify with a cultural group and reflect on experiences associated with that group. In the baseline, participants do not have an identity reflection stage. The interventions are designed to vary the identity dimension through which participants evaluate the consequences of artificial intelligence.
|
|
Field
Power calculation: Minimum Detectable Effect Size for Main Outcomes
|
Before
We plan to recruit between 600 and 800 participants, with approximately equal allocation across three analysis arms: the baseline, economic identity reflection, and cultural identity reflection arms. The baseline combines the cat or dog reflection and no identity reflection conditions, with approximately equal numbers assigned to each. Assuming that participants are approximately evenly distributed between the two identity groups, a sample of 600 participants provides approximately 80 percent power at a two sided 5 percent significance level to detect treatment by identity interaction effects of approximately 0.56 standard deviations. With 800 participants, the corresponding minimum detectable interaction effect is approximately 0.49 standard deviations. For simple pairwise comparisons between analysis arms, the corresponding minimum detectable effects are approximately 0.28 and 0.24 standard deviations, respectively. These calculations do not account for gains in precision from prespecified covariates or for multiple testing adjustments.
|
After
We plan to recruit between 600 and 800 participants, with approximately equal allocation across three analysis arms: the baseline, economic identity reflection, and cultural identity reflection arms. Assuming that participants are approximately evenly distributed between the two identity groups, a sample of 600 participants provides approximately 80 percent power at a two sided 5 percent significance level to detect treatment by identity interaction effects of approximately 0.56 standard deviations. With 800 participants, the corresponding minimum detectable interaction effect is approximately 0.49 standard deviations. For simple pairwise comparisons between analysis arms, the corresponding minimum detectable effects are approximately 0.28 and 0.24 standard deviations, respectively. These calculations do not account for gains in precision from prespecified covariates or for multiple testing adjustments.
|
|
Field
Intervention (Hidden)
|
Before
Participants first report their expectations about consequences of advances in artificial intelligence over the next 10 years: economic inequality in the United States, job opportunities for people with similar skills and work experience, U.S. national security, and bias and discrimination against racial and ethnic minority groups. Next, participants are randomly assigned to one of three treatments.
In the economic-identity treatment, participants classify themselves as working/lower-middle class or upper-middle/upper class. They then complete two open-ended reflection questions about experiences commonly shared by members of their selected economic group and experiences in their own lives that make this group representative of them.
In the cultural-identity treatment, participants classify themselves as traditional or progressive. They then complete corresponding reflection questions about experiences commonly shared by members of their selected cultural group and experiences in their own lives that make this group representative of them.
In the baseline, participants classify themselves as a cat/dog person and complete corresponding reflection questions about characteristics and experiences associated with their selected group and their own connection to that group. Alternatively for half of the participants in baseline, we do not have a reflection stage.
Each reflection question requires a minimum response of 250 characters. Following the intervention, participants answer the same policy and donation questions across all three treatments. Participants later report the two identity dimensions that were not elicited as part of their assigned intervention.
|
After
Participants first report their expectations about consequences of advances in artificial intelligence over the next 10 years: economic inequality in the United States, job opportunities for people with similar skills and work experience, U.S. national security, and bias and discrimination against racial and ethnic minority groups. Next, participants are randomly assigned to one of three treatments.
In the economic-identity treatment, participants classify themselves as working/lower-middle class or upper-middle/upper class. They then complete two open-ended reflection questions about experiences commonly shared by members of their selected economic group and experiences in their own lives that make this group representative of them.
In the cultural-identity treatment, participants classify themselves as traditional or progressive. They then complete corresponding reflection questions about experiences commonly shared by members of their selected cultural group and experiences in their own lives that make this group representative of them.
In the baseline, the participants do not have an identity reflection stage.
Each reflection question requires a minimum response of 250 characters. Following the intervention, participants answer the same policy and donation questions across all three treatments. Participants later report any economic and cultural identity measures that were not elicited as part of their assigned condition.
|