AI Virtual Human-Augmented Game-Based Teaching to Enhance Emotional Intelligence in Nursing Students: Mixed Methods Action Research Study (Preprint)

<sec> <title>BACKGROUND</title> <p>Nursing students need opportunities to recognize and manage emotion while communicating with patients in psychologically complex situations. A two-cycle teaching project combined game-based activities with AI virtual-human cases in a psychiatric nursing course in Taiwan. The study was previously published as a registered-report protocol.</p> </sec> <sec> <title>OBJECTIVE</title> <p>This study evaluated changes in emotional intelligence across two action-research cycles, described students' learning experiences, and integrated quantitative, qualitative, measurement, and implementation evidence.</p> </sec> <sec> <title>METHODS</title> <p>This nonrandomized mixed methods action research study used a single-group pretest-posttest design in two 18-week course cycles. The Adult Emotional Intelligence Scale (AEIS) was administered at pretest and posttest. Paired data were analyzed with Wilcoxon signed-rank tests because total-score change was nonnormal; standardized paired mean differences (Cohen dz) and linear mixed models were also reported. Sensitivity analyses addressed duplicate responses, missingness, multiple imputation, baseline-adjusted cycle comparisons, and floor or ceiling effects. End-of-course reflections underwent inductive thematic analysis. A joint display compared data streams. Three experts reviewed virtual-human scenario content separately from AEIS reliability analysis.</p> </sec> <sec> <title>RESULTS</title> <p>Of 57 students with pretest data, 56 had matched pretest-posttest AEIS data. AEIS total scores increased from mean 52.96 (SD 9.70) to 60.29 (SD 10.48; mean change 7.32; Cohen dz=0.80, 95% CI 0.50-1.10; P&lt;.001). Each domain had higher posttest scores, and all overall paired Wilcoxon tests remained P&lt;.001 after Holm adjustment for the four domain tests. The time-by-cycle interaction for AEIS total was not statistically significant (F1,54.80=1.09; P=.300), so the study did not detect a difference in score change between cycles. Five themes in 59 reflection narratives described students’ professional development trajectory: discovering the self, reframing the helping mindset, understanding others through empathy, growing through uncertainty, and becoming the nurse they aspire to be. AEIS total internal consistency was alpha=.914 at pretest and alpha=.933 at posttest. For virtual-human scenarios, expert-review S-CVI/Ave was .92 for content suitability and .92 for clinical interaction fidelity.</p> </sec> <sec> <title>CONCLUSIONS</title> <p>Course participation was associated with higher self-reported emotional intelligence in two cohorts. Reflection narratives described discovering the self, reframing the helping mindset, understanding others through empathy, growing through uncertainty, and becoming the nurse they aspire to be. The noncontrolled design, small second cycle, and absence of objective competence or longitudinal outcomes require cautious interpretation and motivate a controlled, multisite evaluation.</p> </sec> <sec> <title>INTERNATIONAL REGISTERED REPORT</title> <p>RR2-10.2196/80290</p> </sec>