How Do Artificially Intelligent Companions Influence Psychological Outcomes?

Participants will create an AI companion using Replika and will be asked to interact with their AI companion for at least one hour per week for 10 consecutive weeks.

Participants assigned to the AI condition will download Replika, an AI companion application, and create a personalized AI companion. Participants will be instructed to interact with their AI companion for at least one hour per week for 10 consecutive weeks. They may interact with their AI companion as much as they would like beyond the required one hour per week.

Participants will continue their lives as usual and will not be asked to interact with an AI companion during the 10-week study period.

Loneliness will be assessed using the 3-item UCLA Loneliness Scale, adapted to refer to the last week. Participants rate how often they felt they lacked companionship, felt left out, and felt isolated from others on a scale from 1 (hardly ever) to 3 (often). Higher scores indicate greater loneliness.

General anxiety will be assessed using the 2-item Generalized Anxiety Disorder scale (GAD-2), adapted to refer to the last week. Participants rate how often they felt nervous, anxious, or on edge and were unable to stop or control worrying from 0 (not at all) to 3 (nearly every day). Higher scores indicate greater anxiety.

Depression will be assessed using the 2-item Patient Health Questionnaire (PHQ-2), adapted to refer to the last week. Participants rate how often they experienced little interest or pleasure in doing things and felt down, depressed, or hopeless from 0 (not at all) to 3 (nearly every day). Higher scores indicate greater depressive symptoms.

Quantity of in-person social interaction will be assessed using three face-valid items developed for this study. Items assess the frequency of getting together with friends or family in person during the last week, the number of different people spoken to in person on a typical day, and overall in-person social activity during the last week. Higher responses indicate greater quantity of in-person social interaction.

Quality of in-person social interactions will be assessed using three items from the UBC State Social Connection Scale, adapted to refer to in-person interactions during the last week. Participants rate the extent to which they felt close to people, felt a sense of belonging, and felt a strong bond with others from 1 (strongly disagree) to 7 (strongly agree). Higher scores indicate greater quality of in-person social interactions.

Aggression will be assessed using an adapted 12-item Buss-Perry Aggression Questionnaire Short Form. Participants rate statements about their thoughts and behaviors over the past two weeks from 1 (extremely uncharacteristic of me) to 6 (extremely characteristic of me). Items are summed to create a composite score, with higher scores indicating greater aggression.

Narcissism will be assessed using the 8-item Narcissistic Personality Inventory Short Form. For each item, participants choose between paired statements, with the less narcissistic response scored 1 and the more narcissistic response scored 2. Items are summed to create a composite score, with higher scores indicating greater narcissism.

Traditional gender norm beliefs will be assessed using the 9-item G-NORM Scale Injunctive Norm Subscale. Participants rate statements concerning traditional gender roles from 1 (strongly disagree) to 5 (strongly agree). Items are summed to create a composite score, with higher scores indicating greater traditional gender norm beliefs.

Life satisfaction will be assessed using the 3-item Satisfaction With Life Scale (SWLS-3). Participants rate their current agreement with statements about their life being close to ideal, having excellent life conditions, and being satisfied with their life from 1 (strongly disagree) to 7 (strongly agree). Higher scores indicate greater life satisfaction.

Emotion regulation difficulties will be assessed using the Strategies Subscale of the Difficulties in Emotion Regulation Scale (DERS) Short Form, adapted to refer to the last week and to use past-tense wording. Participants rate three items concerning difficulty recovering or regulating emotions when upset from 1 (almost never) to 5 (almost always). Higher scores indicate greater difficulty regulating emotions.

Purpose in life will be assessed using the 5-item Presence of Meaning Subscale of the Meaning in Life Questionnaire. Participants rate their current sense of meaning and purpose from 1 (absolutely untrue) to 7 (absolutely true). One negatively worded item is reverse scored. Higher scores indicate greater presence of meaning and purpose in life.

Quantity of online social interaction with humans will be assessed using three face-valid items developed for this study. Items assess the frequency of talking to friends or family online during the last week, the number of different people spoken to online on a typical day, and overall online social activity during the last week. Higher responses indicate greater quantity of online social interaction.

Quality of online social interactions with humans will be assessed using three items adapted from the UBC State Social Connection Scale. Participants rate the extent to which they felt close to people, felt a sense of belonging, and felt a strong bond with others during their online interactions in the last week from 1 (strongly disagree) to 7 (strongly agree). Higher scores indicate greater quality of online social interactions.

Motivation to socialize will be assessed using four items from the State Motivation to Foster Social Connection Scale, including two items assessing motivation to socialize with new people and two assessing motivation to socialize with existing connections. Items are rated from 1 (strongly disagree) to 7 (strongly agree), with higher scores indicating greater motivation to socialize with others.

Perceived social support will be assessed using four items from the Interpersonal Support Evaluation List assessing emotional, informational, and instrumental support. Participants rate each statement from 1 (definitely false) to 4 (definitely true), with higher scores indicating greater perceived social support.