Rethinking AI Literacy Education in Higher Education: Bridging Risk Perception and Responsible Adoption
Authors: Shasha Yu, Fiona Carroll, Barry L. Bentley
Organizations: School of Professional Studies, Clark University, 950 Main Street, Worcester, 01610, MA, USA · Cardiff School of Technologies, Cardiff Metropolitan University, Llandaff Campus, Western Avenue, Cardiff, CF5 2YB, Wales, UK · Harvard Medical School, 25 Shattuck Street, Boston, 02115, MA, USA
As AI becomes increasingly embedded across societal domains, understanding how future AI practitioners, particularly technology students, perceive its risks is essential for responsible development and adoption. This study analyzed responses from 139 students in Computer Science, Data Science/Data Analytics, and other disciplines using both explicit AI risk ratings and scenario-based assessments of risk and adoption willingness. Four key findings emerged: (1) Students expressed substantially higher concern for concrete, explicitly stated risks than for abstract or scenario-embedded risks; (2) Perceived risk and willingness to adopt AI demonstrated a clear inverse relationship; (3) Although technical education narrowed gender differences in risk awareness, male students reported higher adoption willingness; and (4) A form of "risk underappreciation" was observed, wherein students in AI-related specializations showed both elevated explicit risk awareness and higher willingness to adopt AI, despite lower recognition of risks in applied scenarios. These findings underscore the need for differentiated AI literacy strategies that bridge the gap between awareness and responsible adoption and offer valuable insights for educators, policymakers, industry leaders, and academic institutions aiming to cultivate ethically informed and socially responsible AI practitioners.
Figures & tables
Risk domain
Scenario example (Survey item)
General AI risks
No direct scenario counterpart
Privacy / Data security
Smart-home assistant learning user habits (Q13)
Bias / Fairness
Job-matching platform ranking applicants (Q14)
Security vulnerabilities
AI financial advisory system analyzing spending patterns (Q15)
Job displacement
AI-driven automation in manufacturing (Q16)
Misinformation / Manipulation
AI-curated personalized news feeds (Q17)
Table 1: Mapping of Explicit AI Risk Domains to Scenario-Based Survey Items
Count
%
Age
18–24 years
88
63.3
25–34 years
48
34.5
35–44 years
3
2.2
Gender
Male
68
48.9
Table 2: Demographic Information of Participants ( N=139 )
Scale
Explicit Risk
Scenario Risk
Adoption Willingness
Items
12
10
10
Cronbach’s α
0.93
0.88
0.86
Mean (SD)
3.70 (0.29)
3.10 (0.35)
3.40 (0.28)
Item mean range
2.99–4.01
2.39–3.46
2.95–3.86
Inter-item correlations
0.30–0.76
0.18–0.66
0.09–0.64
Item-total correlations
0.50–0.77
0.48–0.70
0.39–0.68
Table 3: Reliability and Descriptive Statistics for Survey Scales
Risk Category
Male
Female
CS
DS/DA
Other
Total
General AI risks
3.04 (1.20)
2.93 (1.19)
3.02 (1.30)
3.09 (1.09)
2.61 (1.23)
2.99 (1.19)
Privacy/data security
4.19 (1.03)
3.77 (1.31)
4.25 (1.18)
3.97 (1.04)
3.43 (1.50)
3.98 (1.20)
Bias/fairness
3.35 (1.09)
3.39 (1.29)
3.56 (1.40)
3.38 (1.05)
2.96 (1.07)
3.37 (1.19)
Security vulnerabilities
3.87 (1.11)
3.80 (1.20)
4.10 (1.13)
3.84 (1.09)
3.26 (1.21)
3.83 (1.15)
Job displacement
3.76 (1.28)
3.62 (1.35)
4.08 (1.15)
3.62 (1.27)
3.09 (1.54)
3.69 (1.31)
Misinformation
3.91 (1.12)
3.99 (1.17)
4.08 (1.13)
3.87 (1.18)
3.91 (1.04)
3.95 (1.14)
Table 4: Mean (SD) of Explicit AI Risk Awareness by Gender and Specialization ( N=139 )
Figure 1: Explicit AI Risk Awareness by Academic Specialization. CS and DS/DA students showed significantly higher scores than the Other group.
Risk Category
Male
Female
CS
DS/DA
Other
Total
Personal privacy
2.96 (1.43)
3.32 (1.20)
3.13 (1.48)
3.34 (1.30)
2.22 (1.09)
3.14 (1.33)
Bias
3.12 (1.47)
3.08 (1.42)
3.25 (1.60)
3.07 (1.34)
2.87 (1.39)
3.10 (1.44)
Security vulnerability
3.09 (1.32)
3.06 (1.34)
3.35 (1.36)
2.81 (1.34)
3.26 (1.10)
3.07 (1.33)
Job displacement
2.79 (1.40)
2.85 (1.25)
3.02 (1.42)
2.76 (1.33)
2.57 (1.04)
2.82 (1.32)
Misinformation
3.41 (1.40)
3.18 (1.41)
3.75 (1.28)
3.07 (1.40)
3.00 (1.48)
3.29 (1.40)
Autonomous weapons
3.40 (1.39)
3.49 (1.42)
3.54 (1.44)
3.32 (1.32)
3.61 (1.41)
3.45 (1.37)
Table 5: Mean (SD) of Scenario-Based AI Risk Awareness by Gender and Specialization ( N=139 )
Figure 2: Scenario-Based Risk Awareness and Adoption Willingness Across Application Contexts. A clear inverse relationship appears: scenarios rated as higher risk consistently show lower willingness to adopt, whereas lower-risk scenarios show higher willingness.
Risk Type
Risk Awareness
Adoption Willingness
Mean
SD
Mean
SD
Personal Privacy
3.14
1.33
3.44
1.19
Bias
3.10
1.44
3.31
1.23
Security Vulnerability
3.07
1.33
3.22
1.25
Job displacement
2.82
1.32
3.55
1.22
Misinformation
3.29
1.40
3.17
1.32
Table 6: Scenario-Based AI Risk Awareness and Adoption Willingness ( N=139 )
Figure 3: Scenario-Based Adoption Willingness by Academic Specialization. CS and DS/DA students reported significantly higher willingness than students in Other disciplines.
Risk/Scenario
Z
p
Privacy in automation
-5.409
<.001
Bias in hiring
-1.822
.068
Security in banking
-5.074
<.001
Job displacement in manufacturing
-5.858
<.001
Manipulation in social media
-4.616
<.001
Autonomous weapons in drones
-2.211
.027
Table 7: Wilcoxon Signed-Rank Test: Explicit vs. Scenario-Based AI Risk Awareness
AI literacy develops people's ability to understand, use, evaluate, and shape AI systems responsibly across everyday, educational, and professional settings.