In the rapidly evolving landscape of artificial intelligence, the question of how well the general public understands AI has become profoundly significant. Now, a groundbreaking advancement in AI literacy measurement promises to illuminate this critical gap. Researchers have introduced the Scale of Artificial Intelligence Literacy for All (SAIL4ALL), a meticulously designed tool aimed at quantifying knowledge about AI across diverse adult populations. Unlike prior instruments that often focused on specific groups or relied on self-assessed competencies, SAIL4ALL offers a systematic, evidence-based approach to evaluating factual AI literacy, confronting the complexity of AI understanding head-on.
SAIL4ALL, inspired by the influential framework proposed by Long and Magerko in 2020, intricately dissects AI literacy into four distinct but interconnected dimensions. These dimensions encompass core conceptual knowledge about AI, perceptions of AI’s potential capabilities and limits, comprehension of AI’s underlying operational mechanisms, and ethical considerations surrounding its use. By addressing these thematic pillars independently, the scale respects the inherent multidisciplinary and multifaceted nature of AI literacy, a construct too broad to be captured by a single aggregate score. This nuanced approach signals a paradigm shift in how AI literacy is conceptualized and measured, reflecting the state-of-the-art understanding of this essential cognitive domain.
The first theme, “What is AI?”, is subdivided to capture both the recognition of AI as a concept and its broader interdisciplinary connections, alongside distinctions between general AI and narrow, task-specific AI. This theme probes foundational comprehension that transcends superficial awareness, challenging respondents to differentiate between popular misconceptions and technically accurate understandings. With fourteen targeted items, this section rigorously evaluates the cognitive frameworks people deploy when they encounter AI-related topics in daily life, media, and education.
Complementing this foundational theme is “What can AI do?”, which explores the perceived strengths and weaknesses of AI technologies. This dimension acknowledges that public understanding extends beyond technical definitions to include evaluative judgments—how people imagine AI’s capabilities and limitations in real-world contexts. Employing a bifactorial structure, this segment offers a granular capture of optimism and skepticism, providing vital insights into public sentiment and the potential biases that shape technological adoption and policy preferences.
Perhaps the most technically rich component is the “How does AI work?” scale, standing as a unidimensional measure of an individual’s knowledge of AI’s mechanistic foundations. This 23-item section delves into algorithmic principles, data dependencies, machine learning paradigms, and the socio-technical systems embedded within AI functionality. Its complexity reflects the rigorous challenge of translating intricate scientific knowledge into accessible but sophisticated assessment items. Responses here reveal the depth of understanding that differentiates superficial familiarity from nuanced competence.
Lastly, the scale addresses “How should AI be used?”, introducing a critical ethical lens into the literacy framework. This dimension evaluates awareness of principles such as fairness, transparency, privacy, and accountability. In a world where AI applications increasingly impact personal and societal domains, understanding these ethical imperatives is not merely academic but foundational to responsible citizenship and policy-making. This thematic arena highlights the sociotechnical symbiosis that defines modern AI landscapes.
SAIL4ALL’s dual-format design caters to diverse research and practical needs by offering both a binary true/false option and a more nuanced five-point Likert scale. The binary format focuses sharply on factual correctness, providing clear yes/no answers suited for contexts where certainty and simplicity are paramount. Meanwhile, the Likert scale enriches data granularity by incorporating respondent confidence, capturing subtle gradations in knowledge certainty and revealing deeper layers of cognitive and affective engagement with AI concepts. This adaptability ensures the scale’s utility across academic, educational, and policy-oriented settings.
Crucially, the scale development integrated a rigorous three-phase process, including expert feedback, pilot testing, and advanced quantitative analyses aligned with established psychometric guidelines. This comprehensive methodology safeguards the robustness and validity of the instrument, ensuring that it measures what it purports to across diverse adult populations. The emphasis on factual knowledge rather than self-reported skills marks a methodological advancement in AI literacy research, moving away from subjective bias toward objective competence evaluation.
The implications of SAIL4ALL’s deployment stretch beyond individual assessment. Researchers applying the scale have uncovered important insights about invariance in measurement across gender and educational backgrounds. While gender invariance holds firm, suggesting equitable interpretation between men and women, the scale reveals variability across educational levels, particularly in the conceptual understanding of AI. This finding signals that educational experiences fundamentally shape how AI knowledge is internalized, underscoring the importance of tailored pedagogical approaches to foster equitable AI literacy.
Moreover, the scale’s correlation with affective components such as AI acceptance, affinity, and fear further enriches the theoretical landscape of AI literacy, linking factual knowledge with attitudinal dimensions. Such integrative perspectives enable a deeper exploration of how knowledge and emotion intertwine to influence public engagement with AI technologies, which is crucial for designing communication strategies and interventions intended to bridge gaps between technological innovation and societal readiness.
The debate between using binary and Likert formats is illuminated by the scale’s evidence. While the binary approach offers clarity and simplicity, particularly valuable in large-scale screening or educational contexts, the Likert format’s higher internal consistency and nuanced feedback make it advantageous in research settings demanding richer data. This insight empowers researchers, educators, and practitioners to select response formats aligned with their specific goals, balancing precision and practical utility.
Assessment of measurement invariance reveals methodological subtleties, particularly in the “How does AI work?” dimension, where factor loadings vary across demographic groups. Such disparities encourage further refinement and sensitivity analysis to ensure that the scale accurately captures the intended constructs uniformly across all segments of the population. Addressing these complexities will enhance the scale’s fairness and generalizability in future iterations.
Gender and education emerge as consistent influencers on AI literacy levels, with educational attainment positively correlating with literacy across most dimensions. Men reported higher scores in certain thematic areas, particularly when using the Likert format, suggesting nuanced gender-based differences in knowledge confidence or acquisition. These demographic patterns emphasize the necessity for inclusive educational frameworks and outreach initiatives that recognize and address systematic disparities in AI understanding.
One particularly noteworthy outcome is the identification of potential ceiling effects, indicating that many adult respondents already demonstrate high AI literacy levels. While this is encouraging, it also points to limitations in the scale’s sensitivity for distinguishing among higher-literacy individuals, advocating for the development of more challenging items or complementary assessment methods to capture advanced expertise more effectively.
Beyond research, SAIL4ALL offers valuable practical applications. Its deployment in educational settings can inform curriculum development by pinpointing knowledge gaps across AI’s conceptual, operational, and ethical dimensions. For policymakers, the tool’s ability to reveal population-level literacy insights can guide the design of inclusive AI education programs and public awareness campaigns that transcend superficial technology use to embrace critical understanding and ethical stewardship.
While the scale’s development benefited from a diverse UK sample, its application beyond this context demands caution. Cultural and linguistic differences may affect the interpretation and relevance of items, necessitating cross-cultural validation and adaptation before global deployment. Nonetheless, the careful avoidance of culturally specific references and incorporation of multinational expert feedback lay a promising foundation for future international expansion.
Limitations embedded in the study design include reliance on an online participant pool through Prolific, which may skew demographics toward more technologically savvy individuals, potentially inflating literacy scores and limiting generalizability. The exclusion of less digitally connected populations highlights the need for more inclusive sampling approaches in subsequent research to ensure representation of broader societal segments.
Future research trajectories abound. Beyond further psychometric refinement, longitudinal studies could track AI literacy evolution as technologies and social attitudes change over time. Qualitative investigations, integrating interviews and ethnographic methods, are poised to complement quantitative findings by revealing the lived experiences and interpretive frames individuals apply when engaging with AI. Moreover, extending application into varied professional and educational environments promises to deepen understanding of how contextual factors shape AI literacy acquisition and utilization.
In sum, SAIL4ALL stands as a monumental advancement in the measurement of AI literacy, blending rigor, multidimensionality, and practical adaptability. Its comprehensive approach not only expands academic frontiers but also equips educators, policymakers, and communicators with an unprecedented tool to foster informed, ethical, and empowered public engagement with artificial intelligence, an imperative as AI continues to weave itself inexorably into the fabric of modern life.
Article References:
Soto-Sanfiel, M.T., Angulo-Brunet, A. & Lutz, C. The scale of artificial intelligence literacy for all (SAIL4ALL): assessing knowledge of artificial intelligence in all adult populations. Humanit Soc Sci Commun 12, 1618 (2025). https://doi.org/10.1057/s41599-025-05978-3
