In modern society, appearance is a key means of self-expression, and healthy hair and scalp are essential elements of beautiful hairstyles. With the advent of an aging society, the importance of preventing hair loss and maintaining healthy scalp and h...
In modern society, appearance is a key means of self-expression, and healthy hair and scalp are essential elements of beautiful hairstyles. With the advent of an aging society, the importance of preventing hair loss and maintaining healthy scalp and hair has been increasingly emphasized, and consumers prefer natural plant extract-based products with fewer side effects and excellent efficacy. However, research on scalp and hair care using plant extracts in Korea is scattered across individual studies, lacking systematic evidence. Additionally, while AI (LLM) technology has been rapidly advancing, there is a need for clear guidelines for its ethical application in research methodology. This study has two purposes. First, to develop and present a systematic process for utilizing AI (LLM) models within ethical parameters in meta-analysis research. Second, to provide quantitative evidence for effective intervention methods, product types, and plant extract types by conducting an integrated analysis of effect sizes from domestic human application studies on scalp and hair care using plant extracts.
This study employed meta-analysis methodology, and although publication years were not restricted when searching major domestic databases, the initial search yielded 12,894 articles, but after applying PICOS selection criteria, 23 articles were finally selected. Hedges' g, which corrects for small sample bias, was used as the effect size measure, and a random-effects model was applied. A key methodological feature of this study was the use of AI (LLM) as an auxiliary tool for meta-analysis. AI (LLM) was utilized for meta-analysis information retrieval, literature search, duplicate article identification, article selection, data extraction, data coding, and R script generation, with final analysis performed using R Studio's metafor package. AI-generated results were necessarily verified by researchers to prevent errors, adhering to the ethical principle that AI serves as an auxiliary tool while final judgment rests with the researcher.
Of the 23 finally selected articles, 17 concerned hair, 11 concerned scalp, and 5 addressed both scalp and hair. Study subjects included 6 male-only, 6 female-only, and 9 mixed-gender studies, with ages ranging from 19 to the 60s. Intervention periods ranged from 24 hours to 6 months, though 4 to 8 weeks was most common, and application frequency was most often once or twice daily. The overall effect size for hair care showed Hedges' g of 0.898 (p<.001, 95% CI [0.659, 1.137]), indicating a large and significant improvement effect. By gender, all showed large effects: females 1.147, males 0.852, mixed 0.785. By age group, large effects were confirmed across all ages: 60s 1.340, 20s 0.969, 40s 0.881. By plant extract, medium to very large effects were shown: soybean 1.558, white fungus 1.301, arborvitae leaf 0.821. By product type, large effects were shown: shampoo 1.085, scalp applicant 0.901. The intervention period of 8 to 12 weeks was most s표, and application frequency showed large effects: once daily 1.068, twice weekly 0.836. The overall effect size for scalp care showed Hedges' g of 1.165 (p<.001, 95% CI [0.913, 1.417]), indicating a very large and significant improvement effect. By dependent variable, very large to medium effects were shown: moisture and barrier function 2.612, sebum control 2.129, keratin and surface 1.100. By gender, large to very large effects were shown: males 1.478, mixed 1.197, females 0.797. By age group, all showed large effects: 30s 1.458, 40s 1.261, 50s 1.179. By plant extract, large to very large effects were shown: Lespedeza cuneata 2.227, rooibos 1.605, turmeric 1.245. Mid-term intervention of 4 to 8 weeks was most s표. Application frequency showed large effects: twice daily 1.421, once daily 1.098.
This study ethically utilized AI (LLM) model processes in meta-analysis and quantitatively demonstrated the effects of plant extract-based scalp and hair cosmetics. Results confirmed that AI (LLM) can be utilized as an auxiliary tool throughout the meta-analysis process and can significantly enhance research efficiency, though researcher verification is essential. Plant extract cosmetics showed statistically significant and clinically meaningful improvement effects with Hedges' g of 0.898 for hair care and 1.165 for scalp care. The academic significance of this study lies in being the first in Korea to systematically integrate plant extract scalp and hair cosmetics research, applying advanced meta-analysis methodology in the beauty field, and presenting research methodology utilizing AI (LLM). The practical significance lies in providing scientific evidence applicable in hair salons, scalp care specialty shops, and home care settings, and presenting effective product types, intervention periods, and application frequencies. The industrial significance lies in quantifying effects by plant extract to provide foundational data for functional cosmetics development, presenting product development directions, establishing grounds for subscription-based marketing strategies, and opening possibilities for K-Beauty and K-Cosmetics expansion into Asia. This study made important academic, practical, and industrial contributions by ethically utilizing AI technology to enhance beauty research efficiency and quantitatively demonstrating the scientific effects of plant extract-based scalp and hair cosmetics.