Presentations from the PIAAC Conference 2025

The new PIAAC data shows that skills are central to employment, income, social participation and individual well-being. Yet many countries face stagnating or declining skill levels and rising inequalities. To meet the digital and green transitions and demographic change, we must improve how skills are developed, recognised and aligned with labour-market needs. The challenge for governments and social partners is to design education, training systems and workplaces that are fit for this task.

In this first presentation of the symposium, we illustrate findings from a series of studies based on the first cycle of the PIAAC, focusing on adults’ problem-solving skills in technology rich environments (TRE). Our presentation begins with an examination of vocational education and training (VET) adults, where European-level analyses revealed considerable variation in digital problem-solving performance across countries (Hämäläinen et al., 2014). We illustrate how these differences are shaped by national systems and learning contexts. A more detailed analysis conducted in Finland helps explain this variation by identifying key factors such as educational background, exposure to workplace technologies, and integration of digital tools into learning (Hämäläinen et al., 2015). Building on these results, our presentation explores the characteristics of high-performing adults in PIAAC, highlighting the role of cognitive strategies, self-regulation, and motivation (Hämäläinen et al., 2017). We then extend our focus to higher education (HE) adults in Europe, examining how formal education, learning pathways, and access to digital environments contribute to problem solving competence (Hämäläinen et al., 2019). Our findings emphasize the importance of both formal and informal learning settings in supporting these skills (Nygren et al., 2019). In the next phase, we broaden the perspective to include lifelong learning, illustrating how formal, non-formal, and informal learning experiences throughout adulthood are linked to adults’ ability to solve problems in TRE contexts. Most recently, we turn our attention to teachers using both PIAAC and TALIS data to investigate educators’ problem-solving skills as well as their technology-related knowledge, attitudes, and professional development needs. These findings underline the importance of continuous support for teachers, who act as both users and facilitators of digital competence (De Wever et. al., 2023; Hämäläinen et al., 2021). 

In an era of rapid digitalisation and automation, strong cognitive skills are essential for individuals to adapt to technological changes. Lifelong learning plays a crucial role in maintaining and enhancing these skills. While initial education lays the foundation for essential skills, lifelong learning is crucial for equipping individuals with the skills needed to keep pace with evolving technologies. However, not all adults have equal opportunities to participate. Using data from the Programme for the International Assessment of Adult Competencies (PIAAC), this study examines the socio-demographic patterns of participation in lifelong learning in Flanders. Our analysis distinguishes between formal, non-formal, and informal learning (Boeren, 2016). Results confirm that participation in all three forms is positively associated with literacy skill scores (Van Nieuwenhove et al., 2025). The relationship between participation in lifelong learning and adult literacy suggests that engaging in training activities can be a key factor in keeping adults’ skills up to date in a changing society. We explore which socio-demographic groups are more or less likely to engage in lifelong learning and identify barriers influencing participation. Findings indicate that older adults, less educated adults, adults with a migrant background and adults out of labour force participate less frequently in non-formal training activities (Van Nieuwenhove et al., 2025). The most frequently mentioned barriers to (further) participation in non-formal learning were work obligations, followed by family responsibilities and constraints related to the time and location of the course. Identifying these participation gaps and barriers is essential for designing targeted policies that enhance access to lifelong learning and ensure equitable skill development. Without such efforts, digital transformations risk reinforcing existing inequalities. This study contributes to the discussion on lifelong learning by providing empirical evidence on its accessibility across different socio-demographic groups, emphasising the importance of policies that support skill development for all.

Numerous studies have documented gender differences in self-reported personality, suggesting that women score higher than men, particularly on the Big Five domains of Agreeableness and Neuroticism. Previous studies also indicate that gender differences can vary across facets of the same personality domain. Findings regarding the cross-cultural generalizability of – or systematic cultural variations in – these gender differences are inconclusive. This has been attributed to methodological differences in the samples and the questionnaire adaptations. We therefore systematically investigated gender differences in scores on the Big Five domains in 27 countries based on comprehensive samples representative of the respective adult populations (total N = 143,313) and on questionnaires adapted using consistent translation methodology. Results revealed gender differences in all domains except Extraversion. The most pronounced differences emerged for Neuroticism and Agreeableness, with effects of .40 and .35, respectively. These differences varied in magnitude across countries and were more pronounced in more individualistic countries. Investigating gender differences in scores on 15 more-specific, facet-level traits in a subset of countries, we found that typically one facet per domain showed the most substantial gender differences. For example, women scored higher than men on all facets of Neuroticism and Agreeableness, but primarily on Anxiety and Compassion, respectively.

The associations of cognitive abilities and personality have been studied intensively showing (a) that both constructs show only small correlations and (b) that these correlations are typically clearest for Openness and Emotional Stability. Even though the body of studies investigating these interrelations is vast, it faces two major limitations: (1) Most studies in this field were based on highly selective samples. (2) Nearly all studies are focused on one culture only. We address these two limitations by investigating the association between personality and cognitive ability based on comprehensive and population representative data sets from 27 countries worldwide (total N of approx. 150,000). Aggregated across these countries results widely support previous findings with regard to the amount of explained variance and the differential findings for the Big Five domains. Results however also clearly showed strong country variations in the amount of variance the Big Five explained in cognitive skills.

Is Conscientiousness a useful construct across cultures? Using the 2012 Program for International Student Assessment data, we examined whether perseverance, a measure of Conscientiousness, was related to achievement and truancy across 62 countries/regions (N > 470,000). We investigated whether these relationships were linear or curvilinear in nature and assessed the utility of item-level information. After establishing partial metric invariance of the perseverance measure across various countries/regions and cultural regions, our findings unveiled that perseverance consistently predicted both math achievement and truancy, with predominantly linear associations. Notably, among the five items of the Perseverance scale, the item reflecting one’s tendency to give up easily in the face of challenges emerged as the strongest and most consistent predictor of math achievement. Further, country-level correlations between perseverance and both math achievement and truancy displayed contradictory patterns compared to individual-level correlations, suggesting the presence of biasing factors in how people respond to these measures. Nonetheless, it appears reasonable to conclude that measures of Conscientiousness are pan-cultural predictors of achievement and truancy.

The PIAAC Cycle 2 frameworks for Literacy and Numeracy were developed with the PISA 2018 Reading and PISA 2022 Mathematics frameworks in mind and made explicit connections where possible. With the recent release of PIAAC Cycle 2 data, the similarities between PISA and PIAAC within these two domains are greater now than they have ever been before. However, there are methodological differences between PISA and PIAAC that should be considered before making data-based comparisons. This presentation will describe some of the operational similarities and differences between PISA and PIAAC administrations ranging from the presentation of the stimuli and to the setting of the data collections. For example, the screen layout, item types, organization of stimulus materials and the general development process share many commonalities that create a comparable test-taker experience. By contrast, the one-on-one, interviewer-based data collection in PIAAC versus the group, school-based administration in PISA could result in motivational differences between the two samples. The order of the background questionnaire and the direct assessment differs across the two surveys which could also differentially affect test-takers. These kinds of methodological similarities and differences and their potential impacts on cross-survey comparisons will be discussed.

This paper explores the conceptual similarities and differences between the approaches to assessing literacy in the Survey of Adult Skills (PIAAC), reading literacy in PISA, and emergent literacy in the International Early Learning and Childhood Well-being Study (IELS). The paper aims to explore the extent to which the OECD has presented a cohesive approach to literacy assessment that spans from early childhood through adulthood, within the context of significant changes in reading practices driven by technological change. While a detailed comparison has been conducted for PIAAC adult numeracy and PISA mathematical literacy (Gal and Tout, 2014), no such comparison has been published for adult literacy and reading literacy. However, several OECD publications provide summary tables illustrating similarities and differences in the definitions and structures of the PISA and PIAAC assessment domains. Synthesising these summaries and other OECD material, this paper traces the developments and changes in the Reading/Literacy assessment frameworks of PISA and PIAAC across their multiple cycles. It posits and then compares and contrasts the directions in which these programmes appear to be moving in their assessment of these domains and examines whether there is convergence or divergence in their developmental trajectories. Additionally, the paper includes a summary of how Emergent Literacy is conceptualised in IELS (across its two cycles) and considers the extent to which there is conceptual continuity across the three programmes. The paper ends with a discussion of whether it is possible to understand these frameworks in a more unified way, as presenting an approach to understanding literacy across the lifespan, while noting several important limitations around what can be validly and practically compared. It also considers whether the developmental trajectories of the assessment frameworks are likely to continue in the same direction in the face of further anticipated technological changes.

The OECD Programme for the International Assessment of Adult Competencies (PIAAC) has transformed international research and policy debates on the assessment of adult skills. Although research using PIAAC data is accumulating, little is known about how these data are used and what contributes to developing the various disciplines interested in adult skills. In this study we use a data-driven approach to examine PIAAC-based international research to date. Drawing on a comprehensive analysis of PIAAC-related publications from 2008-2024, the review finds that the field of PIAAC research is young and geographically diverse, with dominant contributions from the United States and Germany. While PIAAC research relies on a diverse pool of researchers with high collaboration rates, only a quarter of publications involve international collaboration. Our analyses also revealed that the field has developed grounded on four interrelated disciplines (education, sociology, psychology, economics) and three differentiated historical paths: theoretical and methodological approaches to the measurement of adult skills (e.g. response processes in computer-based assessments), cognitive skills and problem solving in technology-rich environments at the workplace, and the role of adult literacy skills for societal and economic development. Moreover, the PIAAC literature addresses a broad range of topics, including cognitive, non-cognitive and basic skills (e.g. literacy and numeracy), human capital, occupational mismatch, migration, returns to skills, informal learning and large-scale assessment methodologies. Implications for the further development of PIAAC research for users of PIAAC data, data-providing institutions and policymakers are discussed.

This paper addresses key methodological challenges in the latest cycle of the PIAAC survey, with a particular focus on the measurement of adult education and learning activities – specifically, Module B: Education and Training. It draws upon both longstanding critiques within the field of adult education and learning, as well as an analysis of the updated survey, including its questionnaire and accompanying background report. While the new generation of PIAAC provides valuable insights into adult skills in contemporary societies, significant methodological concerns persist regarding its capacity to accurately capture adult education and learning. First, the second cycle of PIAAC has modified the wording and operationalisation of key items used to measure non-formal education (NFE). While Cycle 1 employed a sequence of questions regarding various types of NFE (e.g., courses, workshops, private lessons), Cycle 2 utilised only the one master category – “training activities”. This conceptual simplification restricts our understanding of NFE. Moreover, it limits the possibility of meaningful longitudinal comparisons with data from the first cycle and the earlier IALS survey. This constraint is particularly apparent in differences in reported participation rates between Cycle 1 and 2 in countries such as South Korea, which appears to see a significant decline in participation rates according to PIAAC. Second, although the PIAAC Cycle 2 included a validated scale to measure personality traits, it notably omits established and validated scales for assessing motivation and barriers to participation – despite the availability of such instruments. This omission is especially striking given that these constructs are central to current theoretical understandings and are regularly employed in empirical studies examining participation and non-participation in adult education and training. Furthermore, the sequencing of questionnaire modules addressing motivation and barriers is limited to two specific groups: (a) those who participated in training for job-related reasons in the case of motivation and (b) only for those who wanted to participate in but did not in the case of barriers. This design choice may significantly impede our capacity to grasp the broader dynamics that drive or constrain participation in activities that leads to acquisition and development of skills across life.

This talk aims to critically examine selected aspects of academic studies that use PIAAC data and that involve numeracy (and literacy) proficiency scores (i.e., based on the direct assessment) or numeracy practices (based on BQ scales). The talk is based in part on results of a scoping review of recent research on adult numeracy and mathematics, in part on observations developed through my long-standing activities as part of the PIAAC community (e.g., Chair of a national PIAAC steering committee, Chair of the Numeracy Expert Group for PIAAC Cycle 1). The talk will reflect on uses of the PIAAC data from Cycle 1, using examples from academic studies that: use complex multivariate analyses or seek to create profiles involving numeracy and literacy proficiency scores (but in ways that do not enable their results to necessarily connect with policy goals), studies that pool results across all levels of proficiency (and ignore the qualitative differences between different proficiency levels, or literacy-numeracy dependencies), studies that analyse numeracy and literacy scales regarding practices (but do not necessarily recognize the nature and limitations of BQ items examining practices), and more. The goal of the talk is not to be critical of specific studies, which will not be individually referenced, nor focus on statistical methodologies or technical nuances. Rather, the discussion and conclusions will aim to sketch in broad strokes areas of concern that focus on the nature of the underlying analytical assumptions or the interpretations of the results, and how these may affect the conclusions from certain studies. Overall, the talk intends to contribute to improving the future use of PIAAC datasets (both the new Cycle 2 datasets and analyses pertaining to changes from Cycle 1 to Cycle 2), so that they can better inform policy formation and future multidisciplinary research regarding adult skills and their determinants and correlates.

The results of the OECD Survey of Adult Skills 2023 provide a once-in-a-decade opportunity to understand adults who have low literacy but are not currently being reached by literacy services. Charitable organizations, state education providers and other groups working in adult literacy can use this up-to-date information to effectively target outreach and communications. There are also significant implications for policymakers. This presentation will examine the risk factors for both having low literacy and not engaging in lifelong learning. The statistical analysis will include age, gender, family status, immigration status, socio-economic status, labour force status, early school leaving and educational attainment to date. We will analyse Ireland as a specific country example, as we have expert knowledge of the Irish social and policy context, and we will compare findings for Ireland with findings for the 31 countries who participated in PIAAC as a whole. We will identify typical characteristics and circumstances of adults with low literacy who are not in education or training and reflect on self-reported barriers to participation in education and training in the light of these findings. We will examine if there any differences between adults with low literacy who are participating in education or training, and those who are not. All of this information together should provide a clear picture of those whose literacy needs are not being met.

International institutions call for installing financial education strategies (OECD, 2024; Ansar et al., 2023). The German Federal Ministries of Finance and of Education and Research launched an initiative for Financial Literacy in 2023. This initiative was criticized for not paying sufficient attention to vulnerable groups (Höhne, 2024) and for attributing the financial difficulties of groups of people to a lack of financial skills and supposedly inadequate financial behaviour, like a low commitment to financial matters. We use data from the 2nd PIAAC cycle to examine questions of adult financial practices focusing on economically vulnerable groups. From this perspective, financial practices serve as a proxy for financial behaviour, which is one component of financial literacy (OECD & INFE, 2020; Beckker, 2020). Secondary analysis of the PIAAC data (1st cycle) showed that adults in strained financial situations deal with everyday financial issues (e.g. calculating costs, prices, budgets; variable G2_Q03a) more frequently than economically better-off groups (Grotlüschen et al., 2019). This relativized the stereotypical assumption that groups with very low incomes adults cause their economic situation through passivity in financial matters (blaming the victim). Current PIAAC data now allows for expanding the analysis. With the variable G2_Q03b (“How often do you use information to make financial decisions?”) the dataset offers information on the way adults deal with economic issues from a more strategic perspective. Research questions for the contribution are whether the (a) frequency of practices differs depending on whether they are more every day, or more strategic, and (b) whether there are differences between countries. Analyses of the frequency of financial practices are carried out in multivariate form in relation to economically vulnerable groups (low income, unemployment), controlling for socio-demographic factors and numeracy skills in a country comparison.

The proliferation of digital devices and the advent of generative artificial intelligence (AI) have transformed communication and information access in contemporary society. Adults with low literacy could be excluded from social practices that require a certain reading proficiency. A more nuanced picture emerges from available research. Data from a national assessment (LEO 2018) shows that for the German-speaking population aged 18 to 64, adults with low literacy use digital devices regularly (Buddeberg & Grotlüschen, 2020). Audio-visual practices, like voice messages or video tutorials, are used more frequently by people with low literacy than by people with higher literacy (Buddeberg & Grell, 2023). Furthermore, Adults with low literacy can prompt in generative AI and even use strategies for prompting intuitively (Skowranek, in review). In Germany, there are 6.2 million adults with low literacy (Grotlüschen et al., 2020), 47.4 per cent of whom grew up speaking a language other than German (Heilmann & Grotlüschen, 2020). Consequently, the role of migration background should be considered when assessing the utilization of digital devices by adults with low literacy. Previous data on the usage of digital devices by adults with low literacy (PIAAC level one and below) could not account for the effects of the COVID-19 pandemic on digital transformation due to the timing of the survey. The data from LEO 2018 was collected in 2017/2018 and should be updated. As they relate to Germany, an international comparison has yet to be made. Especially since the digitalization in Germany can be improved compared to other countries (European Commission, 2022). This contribution shows how digital devices are used by adults with low literacy based on data from PIAAC 2023. For the descriptive analysis, the migration background is taken into account, and a comparison is made between different countries within the EU with more and less developed digitalization.

The production and certification of skills is a primary task of educational systems. One aim of the OECD’s Programme for the International Assessment of Adult Competencies (PIAAC) is to evaluate how effectively and equitably educational systems produce cognitive skills. Educational attainment is, as the most important predictor of adult skills, a crucial background variable in PIAAC. However, due to the varying organization of education across countries, it is also the most challenging variable to measure comparatively. This paper presents how PIAAC Cycle 2 measured and harmonized educational attainment across countries, and related opportunities and challenges. While national adaptation and international ex-post harmonization were already in place in PIAAC Cycle 1, the international PIAAC consortium refined these procedures as well as the target education variable for Cycle 2. Regarding procedures, these included a formal education consultation between countries and the consortium, enhancing the practice established for the European Social Survey. In contrast to experience from the ESS, national adaptations in PIAAC were very complex in some countries, requiring disproportionate time and effort for consultation, data processing, and quality control. Furthermore, countries’ implementation of ISCED does not always lead to highly comparable measures. Then, the paper introduces the target international coding scheme, which extends the International Standard Classification of Education 2011, and the various derived variables made available for analysis. These will be illustrated empirically. However, data protection requirements, together with a lack of provisions such as data usage contracts, led to the suppression of many education variables in the public use files for different countries, limiting the potential of the data for scientific research. The paper will conclude with some reflections on how the measurement of education can be improved and facilitated. 

Dialogue between educational institutions and the business sector has centred on how to most effectively prepare young people for the modern labour market. This has led educators to continuously revise curricular content, with an emphasis on narrowly defined, subject specific knowledge and skills. However, such a focus overlooks essential soft skills, including teamwork, communication, negotiation, critical thinking, and problem-solving. Among these, problem-solving is particularly valued by employers. The PIAAC 2023 survey in Croatia, shows below-average results in the domain of adaptive problem-solving (APS), while results in the problem-solving domain of the PISA 2012 study point to the same conclusions. Therefore, the focus of this paper is on pupils and students who participated in the PIAAC study, aiming to answer the question of which predictors contribute to explaining results in the APS domain. Although the constructs are not entirely aligned, the paper also reflects on Croatian students’ PISA results to identify broader trends in problem-solving skills over the last decade. These findings contribute to the ongoing debate about how the education system addresses problem-solving, especially in light of evidence that both individual characteristics and school-level factors significantly contribute to differences in problem solving skills. The paper is expected to yield several key recommendations for policymakers in the Republic of Croatia from both sociological and pedagogical perspectives. Sociological insights highlight the role of socio-economic and cultural family background in shaping (educational) outcomes, while pedagogical insights advocate for greater use of constructivist, self-regulated, andragogical, and transformative (or even transgressive) teaching and learning approaches. Emphasis is placed on the problem-solving method as a central in contextual and interactive teaching strategies for effectively developing this key 21st-century skill.

The literature documents a higher risk of over-education among younger workers with age believed to act as a linear corrective over time. However, two recent studies have challenged this assumption, finding a U-shaped relationship between age and over-education in the US and Israel. Here, I extend that insight to a cross-national level, using data from the 1st and 2nd cycles of PIAAC to examine over-education among 38,644 tertiary-educated workers across 39 country-years. Employing realized match measures – which compare workers’ education level to the modal level in their country and occupation – I find a U-shaped relationship between age and objective over-education. However, when I employ a subjective measure – where respondents evaluate the education level needed for their job – the relationship reverses, showing an inverted U-shape. This divergence shows that the perceived and structural aspects of over-education evolve differently across the life course. Results are derived from fixed-effect linear probability models, controlling for average skill level, gender, working time status, recent training, and sector (and job satisfaction in subjective models). These patterns remain robust across two distinct objective measures and suggest that older workers may become increasingly over-educated relative to peers, even as they report lower levels of perceived over-education. Moreover, I examine job change as a moderator. Findings suggest that late-career job changes are associated with increased objective over-education, while younger workers benefit from such transitions. However, this effect is not mirrored in subjective responses, hinting at a disconnect between occupational downgrades and worker perceptions. This paper provides the first cross-country documentation of U-shaped age patterns in over-education, demonstrating that both the direction and interpretation of results depend strongly on the choice of measurement strategy.

The Programme for the International Assessment of Adult Competencies (PIAAC) serves as a critical indicator of adult knowledge and skills across countries, informing educational policy decisions. Over the past decade, substantial advancements in information and communication technology (ICT), particularly accelerated by the COVID-19 pandemic and AI-driven tools, have reshaped digital literacy and problem-solving abilities. Computer-based PIAAC domain assessments enable the tracking of adults’ responses, strategies, and time allocation patterns during the test, providing new evidence into adults’ cognitive processes. This study examines changes in response patterns between PIAAC Cycle 1 (2012) and Cycle 2 (2022), focusing on how U.S. adults solve literacy and numeracy items in digital environments. We will compare behavioural trends including response times and time to first actions across 28 literacy and 32 numeracy trend items, which were administered across both cycles. The aims of this study are twofold: (1) to track response pattern in adults’ problem-solving approaches, controlling for proficiency levels and demographic variables, and (2) to identify key factors influencing these shifts. The research occurs in two phases. The first phase analyses whether time allocation and reaction time patterns have significantly evolved, accounting for demographic factors such as gender, ICT proficiency, and education level. We examine age-cohort progression, which offers insights into the development of cognitive skills over time. The second phase applies machine learning techniques to uncover underlying drivers of response pattern variations between cycles. Preliminary results indicate an average decrease of 5 seconds in literacy item completion times, with distinct time allocation trends across proficiency groups. Findings from this study contribute to a deeper understanding of how technological and societal changes shape cognitive processing in educational assessments. 

While advances in Generative AI have made it increasingly accessible to build chatbots for complex domains like large-scale assessments, robust and cost-effective methods to evaluate these tools remain underdeveloped. This study addresses two core challenges for the PIAAC research community: improving chatbot performance for technical tasks and establishing scalable methods for evaluating these systems. This study presents a Retrieval Augmented Generation (RAG)-based chatbot specifically designed to support analysis of large-scale assessments. The chatbot assists users with complex methodological questions – such as the application of plausible values and sampling weights – and can provide step-by step guidance for analysing PIAAC data using R. To benchmark performance, we compared the customized chatbot against GPT-4o across dimensions of correctness, completeness, and communication. Preliminary results show the tailored chatbot significantly outperforms GPT-4o, with an overall pass rate of 86%. To address the lack of scalable evaluation frameworks, we compared an LLM-as-a-judge approach with human raters to assess chatbot responses. This LLM-as-a-judge method demonstrates high agreement with expert human raters, offering a practical and low-cost solution for chatbot evaluation. Importantly, this framework is model-agnostic and can be applied to other potential chatbot developed by PIAAC institutes or users, enabling continuous quality monitoring without the resource intensity of manual expert review. This study contributes to both the technical development of effective chatbots for PIAAC data use and the methodological advancement of chatbot evaluation. As AI tool development becomes more accessible, standardized and scalable evaluation methods – such as the LLM-as-a-judge approach as a complement to human evaluation – are essential for ensuring quality, usability, and trustworthiness across educational data systems.

In the current climate of rapid workplace changes mainly due to automation and AI, is there an emerging new cohort of adults who need targeted literacy and numeracy (L&N) support? Traditionally adult L&N policies and programs have targeted adults at the lower levels of performance, as there is so much evidence that demonstrates that investing in improving their L&N skills makes a positive difference to them as individuals, citizens and workers. Business, industry and the economy benefit as well from these improvements. But there are reasons to believe that current technological changes come with additional L&N requirements. For instance, citizens are now expected to deal with online information, repositories, forms and AI agents, as opposed to talking to a clerk at a counter. The extensive use of big data necessitates critical numeracy skills. Is there now a new cohort that should be targeted beyond basic L&N skills in relation to work, the economy, but also social and civic life? This presentation will report on the outcomes of a 2024 research investigation based on PIAAC Cycle 2, where a small group of members of the literacy and numeracy PIAAC Expert Groups, under the guidance of ETS colleagues, collaborated to research this issue. The research considered what levels of L&N skill performance are required for workers and citizens to be able to take better advantage of automation and AI. We asked what interventions would help someone progress from one level to another to achieve the required skills. Based on the proficiency descriptions and an analysis of the features of all the items, including performance data, the study yielded an analytic picture of the skills that support a transition from level 2 to 3 of the L&N proficiency scales. In addition, a comparison of the drivers of progress in the two domains revealed interesting commonalities. The analysis also provided several hints regarding the instructional strategies that could support this transition.

Why are there substantial cross-country differences in the nature of work, even within the same occupations? In this paper, we investigate how exposure to automation technologies shapes the task content of occupations across countries. Using PIAAC data between 2011 and 2023, we construct internationally comparable measures of task content for 37 occupations across 32 countries from 2011 to 2023. We measure occupational exposure to automation by identifying automation patents and linking them to occupational descriptions from O*NET using NLP tools. Our empirical analysis links automation exposure to task content at the country–occupation–year level, controlling for occupation-by-year fixed effects. We find that countries with greater exposure to automation innovation tend to exhibit higher levels of non-routine analytical (NRA) and non-routine interactive (NRI) tasks, and lower levels of routine cognitive (RC) and manual (MAN) tasks. These effects vary across sectors, occupations, and demographic groups. For instance, task adjustments are larger in manufacturing than in services, and male, low-educated, and older workers experience more significant task changes. Decomposing variation reveals that over 60% of the cross-country differences in task content arise from within-industry changes, rather than shifts across industries. To explore underlying mechanisms, we identify two channels: task-based sorting and task-oriented training. Countries with higher automation exposure exhibit larger wage premiums for NRA tasks and penalties for manual tasks, encouraging workers to shift toward higher-return tasks. Additionally, automation exposure increases training focused on non-routine tasks, reinforcing the shift in occupational task content. Together, these findings suggest that automation reshapes the nature of work by altering the composition of tasks within jobs, primarily through within-industry adjustments and changes in worker behaviour and training priorities.

Challenging technological determinism, this paper introduces the concept of ‘generative jobs’–characterised by high learning requirements, active problem solving, and creativity-oriented work design–as a novel lens to examine how the interaction between technological change and job design enhances workplace performance and supports the augmentation of workers’ skills. This analysis draws on cross-country data from OECD’s PIAAC Cycle 2, which uniquely includes direct assessments of workers’ cognitive skills proficiency. This enables a robust investigation of the skill-enhancing effects of technology in different work environments. Focusing on Singapore, a country with high levels of technology adoption, the paper also draws comparative insights from selected digital frontrunners like the US, Switzerland, and Finland, each representing distinct work organisation models. Multivariate stepwise regression analysis is used to test the interaction effect between exposure to technological change and generative job design on job outcomes across skill levels. Interestingly, once worker and firm characteristics are controlled for, cognitive skills proficiency does not differ significantly across different technological and job environments. However, workers exposed to technological change in generative environments report better outcomes, including higher wages, greater job satisfaction, and higher likelihood of receiving training in response to technological changes. These findings highlight that the impact of technology is not determined solely by its deployment, but by how it is integrated into job design to support human skill development. The paper acknowledges two dataset limitations: first, it predates recent developments in generative AI; second, it lacks detail on types and uses of technologies. Drawing on the authors’ experience designing similar instruments, the paper proposes enhancements to future PIAAC cycles to better capture evolving technology-skill dynamics.

Due to the dynamic nature of modern life, enhancing adaptive problem solving (APS) skills has become essential. APS involves using cognitive and metacognitive strategies to define problems, gather relevant information, and apply solutions effectively by engaging digital, social, and physical tools in evolving contexts (Greiff et al., 2021). Since one of the most crucial skills for successful integration into society is APS, researchers are interested in the factors predicting APS. Scholars indicate that socio-emotional skills (SES) are related to individuals’ problem-solving processes (Sullivan et al., 2010). In addition to individual-level factors, country-level factors such as Human Development Index (HDI), Gross Domestic Product (GDP) and Government expenditure on education (GEE) could play a role in predicting the APS (Greiff et al., 2017). Furthermore, the predictors of APS might differ across various education attainment and gender groups. Thus, the current study aims to examine how the individual- and country-level predictors of APS vary across different educational attainment and whether gender moderates the strength of this relationship in PIAAC. Method The study will use 2nd Cycle of PIAAC dataset that has participants from 31 countries. SES and APS were measured via BQ and cognitive module of the Survey of Adult Skills, respectively. Using multi-level regression, the study analyses the individual- and country-level predictors of APS skills. Individual-level variables are dimensions of SES and country-level variables are HDI, GDP and GEE. Whether the relationships differ according to different levels of educational attainment and gender groups will be examined. Sample weights and plausible values will be handled by Mplus. Expected Outcomes It is expected that the role of SES and country-level factors predicting APS differ between less educated and high educated adults as well as males and females. The findings will be discussed based on regions.

At the individual level, the Practice Engagement Theory postulates that individuals’ reading and writing competencies emerge as a result of their participation in everyday reading and writing activities. Simultaneously, these competencies influence the level of involvement in such practices. This relationship was shown based on German PIAAC data (Reder, Gauly & Lechner 2020). The first cycle of PIAAC showed a decline in literacy practices compared to the 1990s IALS survey. This finding has been used to explain the surprisingly declining or stagnating literacy proficiency in many countries despite technological progress and educational expansion at that time (Desjardins 2017). In PIAAC Cycle 2, a stagnation or decrease in literacy is also observed in many countries, raising questions about the underlying causes. Therefore, this presentation aims to investigate how literacy practices have developed over time, based on the data from PIAAC Cycle 1 and Cycle 2. It will include those countries that participated in both Cycle 1 and Cycle 2. The analysis will consider the indices “Read at Home” and “Read at Work” which are provided as trend variables. The descriptive analysis will pay special attention to the development of literacy practices according to PIAAC levels. Additionally, the analysis of literacy practices will be supplemented with considerations regarding migration, professional qualifications, and participation in adult education.

The economic downturn due to the COVID-19 pandemic has prompted speculation about workers’ participation in job-related training, with both optimistic and pessimistic scenarios. The latest release of PIAAC data covering 29 countries for 2021/22 allows for an empirical examination of the pandemic’s impact on inequalities in job-related training. The study explores how inequalities in job-related non-formal training have changed during the pandemic and examines whether and how the determinants of training participation have shifted – in particular, whether labour market allocation remains the main factor explaining the relationship between education and training (as found for PIAAC 2012/14, see Hornberg 2024, Hornberg et al. 2024), or whether worker characteristics (such as learning disposition and socio-demographics) have gained in importance. Preliminary results show that training participation declined across all education groups in many countries during the pandemic. While differences in participation between education groups persist, the magnitude of these differences, particularly of the lower training gap, has decreased in several countries. However, these reductions should not be seen as a positive development, as they are mainly due to declines in participation among intermediate-educated workers. Despite shifts in participation rates of education groups, labour market allocation remains the key factor driving the education-training relationship in most countries.