Background: Historically, the field of suicide prevention has relied on broad, one-size-fits-all messages for outreach that do not account for the limitations of human cognition or the effectiveness of personalization techniques. Recent evidence suggests that using nudges - small features that attract our attention and influence our behaviors by targeting predictable cognitive biases - are more effective for increasing engagement with suicide prevention materials than standard messaging practices. Although nudges are effective, research clearly indicates that certain nudge strategies work better than others for different people, that people prefer different resources, and that tailoring messages to a particular group substantially increases behavioral engagement. Further, the one-size-fits-all approach inherently selects the messaging and resource preferences of majority populations and neglects underrepresented communities. Thus, studying multiple nudge and resource preferences simultaneously in both majority and underrepresented communities is crucial for determining best messaging practices. The purpose of this study is to merge recent advances in data segmentation and psychometric profiling with nudges to develop personalized messages that maximize engagement with crisis resources and equally incorporate the preferences of underrepresented communities.
Aims: (1) To increase behavioral engagement with crisis resources (i.e., safety plan, crisis line numbers) through personalized nudges and (2) Use computational modeling to compare the projected estimates for potential lives/costs saved through standard messaging, one-size-fits-all nudges, and personalized nudges.
Sample: Participants will be recruited during two study phases. In Phase 1, the sample will consist of 3,000 English-speaking adults recruited from Amazon's Mechanical Turk (MTurk). In Phase 2, participants will be recruited using a Facebook advertising campaign. Facebook estimates that 386 to 1,100 participants will be reached per day on a daily budget of $5.00. Based on our proposed budget outline we expect to deliver 1.2 to 3.4 million Facebook messages.
Procedures: (1) Participants recruited using MTurk will provide information on demographics, personality, nudge/resource preferences, and history of suicidal thoughts and behaviors to inform data segmentation and psychometric profiles. These profiles will be used for personalized nudges in Phase 2. (2) A Facebook advertising campaign using three conditions (personalized nudges, one-size-fits-all nudge, control message) will be created. Advertisements will be directed to users according to their psychometric profiles. Participants who engage with advertisements will be redirected to web pages that include crisis resources.
Outcomes: Advertisement engagement counts, crisis line entry counts, and safety plan completion counts will be collected to measure the performance of each condition. Estimated reductions in suicide deaths, suicide attempts, and medical and work loss costs will be the outcomes for the computational model.
Potential impact: This project tests the effectiveness of a cost- and time-effective intervention that is highly scalable, theoretically informed, and equally incorporates the preferences of underrepresented communities. Testing this strategy's feasibility on the world's largest social media platform provides suicide prevention organizations around the world with a simple framework to increase engagement with prevention materials. On a large scale, even incremental gains in crisis resource engagement could significantly reduce suicide deaths, attempts, and related financial costs.