The researchers set out to answer a question that has long hovered over educational technology: does the technology itself drive improvement, or do human qualities such as emotional skill, ambition, and mental health do the heavy lifting? To find out, they built a model linking five constructs: AI usage, emotional intelligence, career aspiration, psychological well-being, and academic performance. AI usage in this context meant students’ use and perceived use of AI-enabled tools to support learning activities, such as enriching study materials and solving academic problems. Academic performance was measured with a standardized performance scale, while psychological well-being drew on Ryff’s classic multidimensional framework covering autonomy, purpose in life, personal growth, and emotional balance.

Methodologically, the team employed a quantitative cross-sectional design, collecting data at a single point in time from students in ten public secondary schools selected for evidence of AI integration in teaching, learning, or academic support. A multistage sampling procedure produced a final sample of 200 participants, and questionnaires were checked for completeness, with invalid or substantially incomplete responses excluded. Five standardized scales formed the instrument: an AI usage measure adapted from Ng and colleagues’ validated AI literacy questionnaire, the Schutte Emotional Intelligence Scale, the Career Aspiration Scale, Ryff’s Psychological Well-Being Scale, and an academic performance scale. Experts in psychology, computer science, and education reviewed the adapted items for clarity and cultural appropriateness before deployment.

Before the main study, the instruments were pilot tested with 71 secondary school students who were excluded from the primary analysis. All five scales showed acceptable to good internal consistency, with Cronbach’s alpha coefficients ranging from 0.84 to 0.89, and a handful of items were rephrased for clarity based on participant feedback. No items were removed. In the main sample, confirmatory factor analysis conducted in AMOS showed standardized factor loadings between 0.70 and 0.78, composite reliability values from 0.87 to 0.92, and average variance extracted between 0.54 and 0.56, all meeting accepted thresholds for indicator reliability, internal consistency, and convergent validity.

The structural model itself fit the data well. The normed chi-square was 2.12, below the conventional cutoff of 3.00, while the Comparative Fit Index reached 0.94 and the Tucker-Lewis Index 0.91, both above the 0.90 benchmark. Absolute fit indices were similarly strong, with the Goodness-of-Fit Index at 0.92 and the Adjusted Goodness-of-Fit Index at 0.90. Error indices told the same story: a Root Mean Square Error of Approximation of 0.051 and a Standardized Root Mean Square Residual of 0.046, both comfortably below the 0.08 threshold, indicating only small discrepancies between the observed and model-implied covariance matrices.

The substantive findings were striking. AI usage had a significant direct positive effect on academic performance, with a standardized path coefficient of 0.41, and an even stronger positive effect on psychological well-being, at 0.47. Emotional intelligence predicted performance with a coefficient of 0.29, and career aspiration with 0.26, all statistically significant at p < 0.001. Psychological well-being, in turn, significantly predicted academic performance with a coefficient of 0.34. The overall regression model accounted for roughly 46 to 47 percent of the variance in academic performance, a substantial share for an educational outcome, and the model as a whole was highly significant.

The most theoretically interesting result concerned mediation. Using bootstrapping with 5,000 resamples, the team estimated the indirect effect of AI on performance through psychological well-being at 0.16, significant at p = 0.003, with a bootstrapped 95 percent confidence interval that excluded zero. Because the direct effect of AI on performance remained significant alongside this indirect pathway, the authors concluded that psychological well-being partially mediates the relationship. In plain terms, AI appears to help students learn in two ways: directly, presumably through personalized assistance, immediate feedback, and easier access to resources, and indirectly, by bolstering feelings of competence, autonomy, and emotional balance that themselves feed into better schoolwork.

The study is candid about its limits. The cross-sectional design precludes causal claims, and longitudinal work is needed to establish temporal ordering. All data came from student self-reports collected at a single time point, raising the possibility of common method bias, although Harman’s single-factor test found the first unrotated factor accounted for only 34.5 percent of variance, below the 50 percent criterion of concern. The sample of 200 yields a participant-to-item ratio of about 5.56 to 1, below the often recommended 10 to 1, which may limit statistical power for detecting small effects. The researchers also note that they did not distinguish among types of AI applications, did not control for prior academic achievement, and drew participants only from Awi Zone, restricting generalizability to other Ethiopian regions or beyond.

Context matters for interpreting these findings. The study was conducted in a setting where AI adoption is constrained by digital infrastructure, internet connectivity, device access, digital literacy, and institutional capacity. Prior research on Ethiopian schools has identified limited AI understanding among educators and called for curriculum development and teacher training, and English-dominant AI systems may not adequately serve Ethiopia’s multilingual classrooms. That a measurable, statistically robust association between AI use and both performance and well-being emerged in this environment makes the result notable, suggesting the benefits of thoughtful AI integration can materialize even where resources are stretched.

The practical upshot, according to the authors, is that schools should not treat AI as a plug-and-play solution. Their conclusion recommends pairing technological investment with deliberate strategies that support psychological well-being, strengthen emotional intelligence, and nurture career aspirations, since these human factors independently and jointly shape how much students gain from AI-enabled learning. Students with clear career goals, the data suggest, engage more actively with technological innovations, viewing AI as a resource for future development rather than a source of uncertainty. For education systems worldwide racing to deploy AI tools, the message from Awi Zone is that the algorithm is only half the equation: the psychological state of the learner sitting in front of it determines much of what the technology can achieve.

Subject of Research: The influence of AI use on secondary students' academic performance, psychological well-being, emotional intelligence, and career aspiration

Article Title: The influence of artificial intelligence on students’ academic performance through career aspiration emotional intelligence and psychological well being among secondary school students in Awi Zone

Article References: Abebe, A., Geremew, Y., & Muluye, T. (2026). The influence of artificial intelligence on students’ academic performance through career aspiration emotional intelligence and psychological well being among secondary school students in Awi Zone. Discover Education, 5(1), Article 1087. https://doi.org/10.1007/s44217-026-02183-5

Image Credits: AI Generated

DOI: 10.1007/s44217-026-02183-5

Keywords: artificial intelligence, education, academic performance, psychological well-being, emotional intelligence, career aspiration, structural equation modeling, Ethiopia, secondary schools, educational technology, mediation analysis, Discover Education