AI technological disruption awareness, design thinking mindset, and design innovation competence among design students

Synthesis: Wang, Hou, Luo, Du and Yuan (2026) asked whether design students' awareness that AI may replace parts of design work helps or damages their innovation competence, and whether regulatory focus carries the association while AI critical thinking sets its strength. 568 design students at Chinese universities answered a 28-item questionnaire in one wave in early 2026 through convenience and snowball sampling, and the authors fitted a moderated dual-path model. Awareness was positively associated with competence (β = 0.375, p < 0.001), and the two indirect paths nearly offset each other: 0.0929 through promotion focus and -0.0884 through prevention focus, summing to 0.0045 with a confidence interval spanning zero. Critical thinking strengthened the promotion path and, unexpectedly, the prevention path too, flipping that slope from -0.1688 to 0.3270. The ordering is theoretical, not observed: the double-edged sword names two offsetting correlations, not a mechanism.

Key Findings

1. Awareness was positively associated with competence before mediators entered (β = 0.375, p < 0.001, R² = 0.142) and stayed positive with both foci controlled (b = 0.349, p < 0.001; R² = 0.307).
2. Promotion focus carried a positive indirect association of 0.0929, 95% CI (0.0532, 0.1416); prevention focus carried a negative one of -0.0884, 95% CI (-0.1229, -0.0532), or 26.8% and -25.5% of the total effect.
3. The indirect paths summed to 0.0045, 95% CI (-0.0433, 0.0596), so the total indirect effect straddled zero while the direct effect of 0.3424, 95% CI (0.2681, 0.4167) carried 98.7% of the total.
4. Prevention focus went with lower competence (b = -0.393, p < 0.001), the largest coefficient in the model, correlating at r = -0.262.
5. The critical thinking interaction was significant on both paths: b = 0.2088, p < 0.001 for promotion focus, where simple slopes rose from 0.2033 to 0.7578 between -1 and +1 SD, and b = 0.1867, p < 0.001 for prevention focus.
6. On the prevention path the slope ran -0.1688, p = 0.0001 at low critical thinking and 0.3270, p < 0.001 at high; the conditional indirect effect moved from 0.066, 95% CI (0.017, 0.107) to -0.129, 95% CI (-0.187, -0.075), index -0.073, 95% CI (-0.101, -0.043).

One questionnaire, one sitting

The 28 items cover five constructs on a 7-point scale, piloted on 102 design students and trimmed for weak loadings. Reliability was high (alpha 0.897 to 0.932), average variance extracted ran 0.636 to 0.687, and the five-factor model fit well (χ²/df = 1.46, CFI = 0.986, TLI = 0.984, RMSEA = 0.028) against one- to four-factor alternatives, all with CFI and TLI below 0.90. Sampling was convenience plus snowball, covering design students in comprehensive universities and art and design institutions across regions of China: 632 returns, 568 valid (89.87%).

The design point matters more than the psychometrics. All measures were self-reports taken in one sitting, so the four equations re-describe a single covariance matrix. The paper concedes this in its abstract, then writes as though ordering had been seen, with "mediating" in the title and "mechanism" throughout. A path model on one wave can show the association is not fully shared with either focus; it cannot show that awareness came first.

Two paths that cancel

The headline result is arithmetic. Because the indirect effects are close in absolute size and opposite in sign, their sum is 0.0045 and the direct path absorbs nearly everything. The authors define a double-edged sword as two significant opposite-signed indirect effects plus a near-zero total indirect effect; by that standard the data deliver. The mediators account for about 1.3% of the total association: awareness correlates with self-rated competence, and the two routes through which it might act pull against each other.

Prevention focus is where the paper is least consistent. H4 predicted that awareness raises prevention focus, and the average coefficient (β = 0.228) supports it. The moderation analysis shows that average is a blend across a crossing interaction: one standard deviation below the mean on critical thinking, more awareness went with less prevention focus. The authors report the negative slope but interpret only the high end.

Critical thinking as amplifier, not buffer

The expectation was a buffer: critical thinking should weaken the conversion of perceived threat into defensiveness. The interaction runs the other way, and the authors make the reversal their main theoretical contribution, reading high-critical-thinking prevention focus as reasoned vigilance rather than anxious avoidance. That reading fits critical thinking's correlation with prevention focus (r = 0.493), higher than its correlation with competence (r = 0.159), but it is untested: item content cannot separate prudence from worry. Both second-stage interactions were null (b = 0.0170, p = 0.500 and b = -0.0140, p = 0.577), confining the claim to appraisal rather than output.

What this means for practice

For instructors, the usable claim concerns appraisal, not exposure. On the paper's reading, raising or lowering awareness is the wrong lever, since both motivational routes are live in one classroom; the lever is how students are taught to read what they notice. That points to verification of AI output, ethical reasoning about bias and authorship, and integration with domain knowledge, differentiated by profile: low-threat tasks for students who hear disruption as a personal verdict, ethics and quality-assurance work for students whose caution is already reasoned. One subgroup result: the awareness to prevention association was weaker among students with formal AI training (β = 0.186, p < 0.01) than without it (β = 0.300, p < 0.001).

Limitations

• One self-report wave cannot separate "awareness raises prevention focus" from "defensive students notice AI more"; the authors concede longitudinal or experimental work is needed before the chain can be called causal.
• Recruitment was convenience plus snowball from Chinese design programs, with no probability frame, no early-versus-late comparison, and no specialty variation.
• The outcome is self-rated competence; no portfolio, studio critique or instructor rating was collected. Method variance is reduced but not eliminated: the first unrotated factor took 31.54% of variance, just under the 40% heuristic.

Citation

Wang, X., Hou, J., Luo, T., Du, J., & Yuan, Z. (2026). The double-edged sword effect of AI technological disruption awareness on design innovation competence: the mediating role of regulatory focus and the moderating role of AI critical thinking. Frontiers in Psychology, 17, 1883577.