Artificial Intelligence-Assisted Intervention in Supporting Pepople Experiencing Homelessness: A Systematic Review

Background and Purpose: Homelessness is a significant social issue worldwide. Individuals experiencing homelessness face a variety of challenges, such as a lack of affordable housing, social isolation, and limited access to healthcare and social services. Interventions that used Artificial Intelligence (AI) have emerged as potential tools in improving well-being and quality of life for targeted population. However, there is a lack of systematic review in summarizing and synthesizing the evidence of these AI-assisted interventions. Gaps remain in understanding the specific challenges AI aims to address, the types of AI technologies used, their effects, and the concerns associated with applying AI to homeless populations. This study aims to systematically identify AI-assisted interventions for homeless populations, evaluate existing findings, and generate practical implications for researchers, practitioners, and developers.

Methods: We followed the Preferred Reporting Items for Systematic review and Meta-Analysis (PRISMA) to search for English articles published in MEDLINE, Web of Science, PubMed, Embase, and PsycINFO. We used search terms related to artificial intelligence (e.g., machine learning, robot) and homelessness (e.g., unstably housed, unsheltered). A forward and backward search was conducted to include additional articles. Articles were included if a study used AI technology to assist intervention targeting individuals who experience homeless. Two authors screened the articles independently and extracted data from each study. A quality review was conducted using a customized quality assessment tool based on the features of reviewed article and relevant items from National Heart, Lung, and Blood Institute (https://www.nhlbi.nih.gov/health-topics/study-quality-assessment-tools).

Results: Twelve studies published between 2017 and 2024 were included in the final review. These studies were conducted in Bangladesh, Korea, and the United States. The interventions mainly used machine learning algorithms, robotics, human-computer interaction, and AI chatbots to facilitate service provision, resource allocation, alcohol and HIV prevention, and optimal placement of service facilities. Specifically, three studies testing the effectiveness of interventions for HIV prevention reported favorable outcomes. One study piloted an AI-powered system and yielded positive results in supporting the adoption and rehabilitation of homeless children in developing countries. Eight studies yielded overall positive outcomes. The eight AI-assisted approach were still in the development and validation stages, aiming to assess preliminary feasibility or gather feedback from potential users. Ethical concerns discussed in the included articles such as fairness, bias, reliability, trust, and transparency. Practical issues involved feasibility, sustainability, meaningful impact, and resource constraints.

Conclusions and Implications: This study identified and evaluated AI-assisted interventions for homeless populations. To our knowledge, this is the first systematic review on AI-assisted interventions for people experience homelessness. While the number of implemented studies remains limited and most interventions are still in development phase, the included studies showed promise in supporting homeless populations in meeting their needs. The findings suggest that AI technologies could significantly improve service delivery and resource allocation for homeless populations. Careful consideration must be given to the ethical and practical challenges. Further research is needed to implement AI-assisted interventions in real-world settings, test their feasibility and meaningful impact, and ensure that these technologies are both effective and equitable.