Male high school students demonstrated significantly higher scores in Determination and Exploration dimensions of AI attitudes compared to female students, with moderate and small effect sizes respectively, while no gender differences emerged in Collaboration, indicating that gender-responsive AI education strategies must specifically address agentic and exploratory dimensions to create equitable learning environments that empower all students to thrive in an AI-driven future.
Objective: This study aimed to develop and psychometrically validate a comprehensive scale measuring high school students' attitudes toward Artificial Intelligence, grounded in UNESCO's AI education guidelines and Self-Determination Theory. The research sought to identify the underlying factor structure of students' AI attitudes, examine whether significant gender disparities exist across these dimensions, and assess how AI attitudes relate to students' perceptions of motivational support in their learning environment, ultimately providing educators and policymakers with a reliable instrument for assessing and addressing equity issues in AI education.
Methods: The study employed a rigorous sequential three-phase methodology with 553 students from grades 10, 11, and 12 across public, private, and charter schools in urban, semi-urban, and rural locations in Qatar. Phase 1 involved 223 students for Exploratory Factor Analysis (EFA) using Principal Component Analysis with Varimax rotation to identify latent factor structures, with Kaiser-Meyer-Olkin and Bartlett's test assessing sample adequacy. Phase 2 used a separate dataset of 200 students for Confirmatory Factor Analysis (CFA) through AMOS software to validate the factor structure identified in Phase 1, evaluating model fit through multiple indices including Chi-Square/df, RMSEA, SRMR, CFI, TLI, and IFI, with convergent validity assessed via Average Variance Extracted (AVE) values and reliability measured through Composite Reliability (CR). Phase 3 involved 130 students for descriptive analysis and independent samples t-tests examining gender differences, with Cohen's d calculated to assess effect sizes. The instrument was initially developed through extensive literature review, expert validation involving 10 specialists who rated items using Item-Level Content Validity Index (I-CVI), and iterative refinement, resulting in a final 22-item questionnaire comprising 17 items measuring AI attitudes and 5 items assessing supportive learning environment motivation, all rated on a 5-point Likert scale. Participants were purposively sampled from students enrolled in Computing and Information Technology courses ensuring prior AI exposure, representing diverse demographics including both Qatari nationals and expatriates from Asian and African countries. Data collection occurred from September 2022 to April 2023 via pen-and-paper administration supervised by research team members, with procedural fidelity monitored in 35% of sessions through direct observational recording.
Key Findings: Exploratory Factor Analysis confirmed a robust three-factor structure for students' AI attitudes. The Exploration dimension (7 items, factor loadings 0.656-0.802, Cronbach's α = 0.915) captures students' curiosity and proactive willingness to engage with AI tools and resources, operationalizing autonomy by reflecting intrinsic motivation and self-directedness in exploring robotics competitions, programming activities, and AI discussions. The Determination dimension (7 items, factor loadings 0.544-0.806, Cronbach's α = 0.913) reflects students' persistent commitment to learning and mastering AI-related skills despite challenges, aligning with the competence construct by emphasizing career aspirations in AI, pursuit of higher education in AI fields, and confidence in future employability. The Collaboration dimension (3 items, factor loadings 0.617-0.803, Cronbach's α = 0.772) assesses students' ability and inclination to work with peers in AI-related tasks, operationalizing relatedness through emphasis on social connectedness, peer teamwork, and ethical considerations in group AI projects. Confirmatory Factor Analysis validated this structure with strong fit indices for AI Attitude (SRMR=0.073, TLI=0.885, IFI=0.905, CFI=0.903, RMSEA=0.076, χ²/df=1.966). The supportive learning environment emerged as a unidimensional Motivation construct (5 items, factor loadings 0.660-0.859, Cronbach's α = 0.843) with excellent fit indices (SRMR=0.065, TLI=0.923, IFI=0.978, CFI=0.977, RMSEA=0.029). Composite Reliability values exceeded 0.70 for all factors, and AVE values supported convergent validity. Discriminant validity was confirmed using the Fornell-Larcker criterion, with the square root of each factor's AVE exceeding correlations between factors. Nomological validity was established through structural equation modeling demonstrating positive correlation between students' AI attitudes and their perception of supportive learning environment, confirming that students who feel encouraged and supported exhibit higher engagement and persistence in AI activities. Gender-based analysis revealed statistically significant differences in Exploration (t=2.948, p=.004, Cohen's d=0.523, moderate effect) and Determination (t=2.647, p=.009, Cohen's d=0.470, small effect), with male students scoring significantly higher than females in both dimensions. Notably, no significant gender differences emerged in Collaboration (p=.690, Cohen's d=-0.066) or Motivation (p=.748, Cohen's d=-0.057), suggesting males and females share comparable perceptions of collaborative engagement and motivational support. Correlation analysis demonstrated that all three AI attitude factors significantly intercorrelated for both genders, indicating that enhancing one domain may positively influence others.
Implications: The findings carry substantial implications for designing equitable and inclusive AI education. The validated three-dimensional structure grounded in Self-Determination Theory provides educators with a theoretically robust framework for understanding how psychological needs (autonomy, competence, relatedness) shape students' AI engagement. The significant gender disparities in Exploration and Determination dimensions reveal that current AI educational approaches may inadvertently favor male students in agentic and self-regulatory aspects of AI learning, likely reflecting differential early exposure, confidence levels, and societal expectations about technological competence. These findings align with literature suggesting girls face barriers including limited exposure to AI-relevant tools, fewer relatable role models, and prevailing stereotypes depicting AI as male-dominated. However, the absence of gender differences in Collaboration suggests that team-based, socially engaging AI learning tasks create more egalitarian spaces where students contribute equitably regardless of gender, highlighting the inherently inclusive nature of collaborative learning environments. This finding emphasizes that interventions targeting collaboration may naturally buffer gender-based confidence gaps. The positive relationship between AI attitudes and supportive learning environment confirms that teacher encouragement, school resources, and motivational climate significantly influence student engagement, emphasizing the critical role of autonomy-supportive, competence-enhancing, and socially engaging instructional practices. The interconnectedness of the three attitude dimensions suggests holistic interventions supporting multiple aspects simultaneously may prove most effective, with ripple effects from increasing exploration opportunities potentially enhancing both persistence and peer collaboration. Practically, the study provides a reliable, validated tool for assessing students' AI attitudes that can guide targeted pedagogical interventions, inform curriculum development, and enable ongoing monitoring of equity in AI education implementation. The proposed framework suggests multiple intervention pathways including hands-on exploratory tasks combined with collaborative reflections for all students, differentiated strategies recognizing that males may prefer learning-by-doing while females may value deeper understanding before application, informal learning opportunities building confidence beyond classrooms, and parental/family support providing resources and encouragement equitably to both genders.
Limitations: Several limitations contextualize the findings. First, the sample consisted exclusively of students enrolled in Computing & IT courses, purposively selected to ensure AI exposure, which may restrict generalizability to students from other academic backgrounds with less AI familiarity, suggesting future validation across diverse student populations and academic streams is needed. Second, the cross-sectional self-report survey design limits ability to capture attitude development over time or in response to specific interventions, with longitudinal or experimental designs needed to understand dynamics of attitude formation and change. Third, the instrument was designed specifically for high school students, with suitability for younger learners or those at early AI exposure stages remaining unexplored, requiring development of age-adapted versions and validation for developmental appropriateness across age groups. Fourth, exclusive reliance on surveys may not fully capture complexity of students' experiences and motivations, suggesting qualitative approaches like interviews, focus groups, or classroom observations could provide richer interpretations. Fifth, the conventional Likert format beginning with "Strongly Agree" may introduce minor acquiescence bias risk among respondents, though this standard ordering supports response consistency and ease of understanding particularly for younger participants. Sixth, while effect sizes for gender differences were small to moderate, suggesting subtle yet meaningful disparities, the practical significance of these differences in real-world educational contexts requires further investigation. Seventh, the geographic concentration in Qatar with culturally diverse participants including Qatari nationals and Asian/African expatriates provides some diversity, but findings may not generalize across other geographic regions, international contexts, or cultural settings with different technological infrastructure, educational systems, or gender norms. Finally, exploring alignment between students' attitudes and actual behaviors such as elective course choices, AI project participation, or career interests would enhance the scale's predictive utility and practical relevance.
Future Directions: Several crucial research directions emerge from this study. First, longitudinal research is essential to understand how AI attitudes develop over multiple years, examining whether early gender disparities persist, narrow, or widen through secondary education and how specific educational interventions influence attitude trajectories over time. Second, cross-cultural validation studies are needed to examine whether the three-factor structure, gender patterns, and relationships with motivational support generalize across different cultural contexts, educational systems, and countries with varying approaches to technology education and gender equity. Third, experimental intervention studies should test specific pedagogical strategies designed to reduce gender gaps in Exploration and Determination dimensions, such as providing female students with visible AI role models, early hands-on technology experiences, confidence-building activities, and socially relevant design projects, while measuring whether these interventions successfully narrow observed disparities. Fourth, research examining the instrument's applicability to younger students in elementary and early middle school grades would support earlier identification of emerging attitude patterns and enable preventive interventions before disparities become entrenched. Fifth, mixed-methods studies combining quantitative scale data with qualitative interviews, focus groups, and classroom observations would provide richer understanding of the mechanisms through which gender, instructional practices, and peer dynamics shape AI attitudes. Sixth, predictive validity studies examining relationships between measured AI attitudes and actual behaviors including course selections, extracurricular participation in AI clubs or competitions, career aspirations, and eventual pursuit of AI-related higher education would demonstrate real-world relevance and practical utility. Seventh, research investigating how AI attitudes intersect with other dimensions of diversity beyond gender, including socioeconomic status, race/ethnicity, disability status, and prior academic achievement, would provide more comprehensive understanding of equity issues. Finally, studies examining optimal combinations of pedagogical approaches for different student profiles could identify whether differentiated strategies (hands-on exploratory for males, reflective structured for females) or universal inclusive approaches (collaborative team-based for all) prove more effective in promoting equitable AI engagement while avoiding reinforcement of stereotypes.
Title and Authors: "Exploring Students' Attitudes Toward Artificial Intelligence (AI): Psychometric Validation of AI-Attitude Scale" by Almaas Sultana, Nafilah Abdul Latheef, Nitha Siby, and Zubair Ahmad, Qatar University.
Published On: October-December 2025
Published By: SAGE Open (DOI: 10.1177/21582440251378375)