Research report: An experimental study on the mediating roles of career decision-making self-efficacy between digital media literacy and employment choices of university students

Of the 204 students assigned to the four conditions, 192 completed four assessment waves extending through six months after graduation. The combined condition was associated with more document-verified job offers, a starting salary 18.1% higher than in the control condition, closer externally coded major–job match and greater employment satisfaction. Students were assigned to Media Critical Literacy Training, Career Algorithm Transparency Training, both programs or neither.

The design allowed the investigators to evaluate the two components separately and together. Career Algorithm Transparency Training primarily increased Algorithmic Awareness for Careers, abbreviated AAC, which the researchers specified as a distinct subdimension of digital media literacy. Its estimated effect on AAC was a partial eta squared of 0.128.

Media Critical Literacy Training primarily raised broader critical literacy, with a partial eta squared of 0.096. The two components jointly strengthened career decision-making self-efficacy, abbreviated CDMSE. The trial therefore did not treat digital media literacy as a single undifferentiated skill.

Instead, it tested a program focused on algorithmic transparency alongside a program focused on critical media literacy and examined whether their combination affected CDMSE. This distinction matters because the reported abstract does not provide the numerical effect sizes or uncertainty estimates for every individual employment outcome, nor the sample counts underlying each comparison. Parallel-process latent growth modeling was used to analyze change across the four assessment waves.

This approach represented the repeated measurements as trajectories and estimated whether increases in one measured process preceded increases in another. The model estimated that growth in AAC predicted growth in CDMSE, with a standardized coefficient of 0.42. It also detected the reverse temporal association: growth in CDMSE predicted later AAC, with a coefficient of 0.28 and a 95% confidence interval of 0.13 to 0.43.

These are modeled relationships among changes in the measured variables, rather than evidence that either process inevitably causes the other to change. CDMSE also carried a statistically significant indirect effect on a composite employment outcome. The estimated effect was 0.151, with a 95% confidence interval of 0.077 to 0.238.

In the study’s mediation model, this result connects digital media literacy with employment outcomes through changes in career decision-making self-efficacy. The reported analysis also incorporated reverse-direction feedback, operationalizing the bidirectional mediation structure proposed in the paper’s theoretical framework. The findings support a dynamic relationship in which algorithm awareness and career decision-making confidence may develop together, but the abstract does not describe the exact time ordering of every individual transition.

The six-month employment results provide the practical outcome measure for the intervention. Participants in the combined condition obtained more offers verified through documents than controls, began jobs with starting salaries 18.1% higher, achieved closer alignment between their majors and jobs coded by external evaluators, and reported greater employment satisfaction. These outcomes covered both externally assessed characteristics of employment and a self-reported measure.

Because the source summary does not report their separate effect sizes or confidence intervals, the relative contribution of salary, job match, offer verification and satisfaction cannot be quantified from the available evidence. Several limits constrain how broadly the result can be interpreted. The population was limited to graduating undergraduates enrolled at three universities in southern China, so the findings do not automatically extend to students in other academic years, regions or labor markets.

Twelve of the 204 enrolled participants did not complete all four assessment waves, leaving an analyzed sample of 192 and introducing the possibility of attrition-related bias. The study also tested a particular pair of training programs and an employment context that the abstract does not describe further. Its results therefore concern this factorial intervention and measured outcomes rather than proving that any single awareness skill improves every aspect of career decision-making.

Even with those limits, the experiment offers experimental support for treating Algorithmic Awareness for Careers as a trainable component of digital media literacy. The randomized assignment and comparison of four conditions showed that the transparency program mainly affected AAC, the critical-literacy program mainly affected critical literacy, and the pair jointly strengthened CDMSE. The study therefore identifies a specific educational combination that was associated with improved graduate employment outcomes, without establishing that the effect generalized beyond the participating students and settings or persisted beyond the six-month follow-up.

Subject of Research: A factorial randomized trial of digital media literacy training and employment outcomes among graduating undergraduates in southern China.

Article Title: An experimental study on the mediating roles of career decision-making self-efficacy between digital media literacy and employment choices of university students

Article References: Wu, S. (2026). An experimental study on the mediating roles of career decision-making self-efficacy between digital media literacy and employment choices of university students. Frontiers in Psychology, 17, Article 1963363. https://doi.org/10.3389/fpsyg.2026.1963363

Image Credits: AI Generated

DOI: 10.3389/fpsyg.2026.1963363

Keywords: randomized controlled trial, digital media literacy, algorithmic awareness, career decision-making self-efficacy, university graduates, employment outcomes, starting salary, job satisfaction, latent growth modeling, bidirectional mediation, experimental, mediating