Youth work represents a critical non-formal educational space where professionals support young people's development in increasingly digitalized environments. As artificial intelligence (AI) tools become embedded in young people's everyday lives and professional practices, youth workers face the challenge of developing AI literacy – not only to engage effectively with digital technologies themselves but also to support young people's critical engagement with AI systems. This study examines how changing technological conditions are reshaping both the learning environments of youth work and the professional competencies required within these spaces.

Drawing on the AI@youthwork research project (2024–2025), this paper presents findings from a quantitative online survey of 303 Austrian youth work professionals conducted in autumn 2024. The study investigates practitioners' attitudes towards AI integration, their; current knowledge about AI, and their professional learning needs as they navigate the digital transformation of non-formal educational spaces.

The findings reveal differences between current AI users (n = 142) and non-users (n = 149). While 44.6% of all respondents believe AI could enhance youth work – particularly for administrative tasks such as case documentation (68.3%) and reporting (66.7%) – practitioners express considerable uncertainty about AI applications in client-facing activities, with 53.1% viewing AI-based counselling as risky. Current users primarily identify technical and ethical concerns – incorrect AI diagnoses (56.2%), data security issues (52.3%), and difficulties in AI interpreting human emotions (52.3%) – while non-users cite knowledge gaps (69.9%) and uncertainty about applications (63.0%) as primary barriers to adoption.

The professional learning needs assessment demonstrates that youth workers require substantial support to develop AI literacy: 84.1% need knowledge about application possibilities, 73.1% seek formal training opportunities, and 62.9% require organizational guidelines. Notably, 55.3% emphasize the importance of collegial spaces for reflecting on AI-related decisions, highlighting the social dimension of professional learning in this domain.

These findings illuminate how social transformations driven by AI are reconfiguring the conditions of educational practice and research in non-formal settings. Youth work professionals view AI tools not as replacements for human expertise but as supplements that require critical evaluation and contextual judgment – particularly when working with vulnerable populations. The study contributes to understanding how educational spaces must respond to technological change by supporting practitioners in developing both technical competencies and critical reflexivity regarding AI's capabilities, limitations, and potential to perpetuate social inequalities.

The paper concludes by discussing implications for professional development frameworks in non-formal education, arguing that AI literacy must be understood as an organizational priority requiring dedicated time, resources, and opportunities for collaborative learning. As youth work navigates the intersection of social pedagogy and digital transformation, this research offers insights into how educational spaces can respond to technological change while maintaining their core commitment to supporting young people's development and agency.

Introduction

Teachers in secondary education consistently face emotionally demanding situations in their daily practice. Empirical research indicates that it is often not singular crises that cause severe emotional exhaustion, but rather the cumulative effect of continuous, smaller tensions—ranging from disruptive student behavior to challenging interactions with parents. Despite the pressing need for professional self-regulation, schools frequently lack structured, just-in-time support mechanisms for their staff. While digitalization has introduced tools for administrative tasks, the potential of technology to serve as a pedagogical "thinking assistant" for complex emotional challenges remains critically underexplored.

The theoretical foundation of this study connects knowledge management, pedagogical reflection, and conflict management. The proposed system utilizes Retrieval-Augmented Generation (RAG) to draw upon curated pedagogical concepts, ensuring professional accuracy. Deliberately departing from prescriptive models, it employs a Socratic dialogue architecture. Through targeted meta-prompts, the AI encourages "teacher noticing"—the professional ability to perceive relevant classroom events and interpret them based on pedagogical knowledge. This separates objective observations from subjective emotional interpretations, fostering critical distancing rather than cognitive offloading.

This Design Science Research (DSR) study aims to develop and evaluate a specialized AI chatbot, conceptualized as a "virtual colleague," to guide educators through structured reflective processes. The central research question is: To what extent is an AI-supported chatbot perceived by teachers as a helpful tool for reflecting on and managing school conflicts? Sub-questions include: (1) In which specific areas of school conflicts do teachers experience the greatest need for support? (2) Which contents of the chatbot are perceived as supportive? (3) Which functional and dialogical properties influence its perceived usability?

Methods

Methodologically, this research embraces the iterative Design Science Research (DSR) paradigm to ensure the artifact is rooted in actual pedagogical needs. As the project has just commenced, the design begins with an explorative qualitative needs assessment involving semi-structured interviews with lower secondary school teachers. Purposive sampling ensures the inclusion of varying levels of professional experience. This foundational data informs the prototypical development of the chatbot. Subsequently, the artifact will undergo two iterative phases of qualitative evaluation where educators test the system using exemplary conflict situations from their practice. Data collected from guide-based interviews will be rigorously evaluated using qualitative content analysis according to Kuckartz and Rädiker. A combined deductive-inductive category system will allow for a nuanced understanding of the system's impact.

Results and Discussion

As this research is currently in its early stages, final empirical results are pending. However, based on the robust theoretical framework, it is anticipated that the Socratic intervention will successfully redirect educators from reactive, emotionally charged states toward analytical noticing. We expect the qualitative data to reveal that the structured dialogue helps teachers broaden their repertoire of action and gain necessary emotional distance. The future discussion will focus on evaluating the specific dialogical features of the AI that are perceived as genuinely helpful versus those that feel mechanical. A critical point of discussion will be the boundary of AI capabilities, specifically exploring the balance between technological support and the irreplaceable need for human empathy.

Educational Significance of the research.

This study contributes a practically validated design artifact and theoretical insights into human-computer interaction in educational settings. It addresses a critical gap in teacher well-being by providing a scalable, low-threshold tool for modern schools. By establishing a safe "role-play sandbox," the system allows educators to practice de-escalating communication strategies without real-world consequences. Ultimately, this research demonstrates how adaptive technology can actively catalyze professional development, functioning as a valuable complement to—rather than a replacement for—essential human mentoring.

In today’s information-driven society, the Internet serves as a rich repository of valuable information but also a space where malicious actors create and disseminate misinformation. This poses a substantial challenge, undermining public support for critical topics like climate policies and health measures, and generating a climate of uncertainty. Individual-level interventions mainly address context, content, or source of misinformation with each intervention type having notable limitations from small, short-lived effects to extensive professional involvement. This highlights the need for approaches that bundle these skills, teaching citizens when to use which strategy. We designed a short curriculum that integrates these three pillars, using pedagogical principles (e.g., Cognitive Apprenticeship, educational design) to teach (1) inoculation, (2) lateral reading, and (3) intellectual humility.

We address the main question: “To what extent can a short, bundled curriculum increase participants’ credibility discernment, sharing discernment, analytic processing, and intellectual humility?”

We also investigate the role of several covariates (e.g., conspiracy thinking, persuasion knowledge, climate knowledge, political opinion).

Methodology

While data collection for the main study (N=900) is ongoing, we have already conducted a pilot study with N=276 German secondary school students (35% female, 2% non-binary, 27% did not report gender, Mage=15.72). Classrooms were randomly assigned to a waitlist control condition or experimental condition. The experimental condition received the curriculum, consisting of three one-hour modules on judging the context (based on inoculation principles), investigating the source (based on lateral reading), and reflecting on the content (based on reappraisal and intellectual humility). The study included a pretest before the curriculum and an immediate posttest after each intervention.

Dependent variables were credibility discernment, sharing discernment, and analytic thinking for the context (8 social media posts) and source intervention (4 news articles) as well as intellectual humility and the third-person effect for the content intervention.

Results

To answer our question, we estimated Bayesian linear mixed models with weakly informative priors. The curriculum proved effective. The context intervention improved credibility discernment, β=0.69, sharing discernment, β=0.52, and analytic thinking, β=0.67. The source intervention improved source credibility discernment, β=0.87, source sharing discernment, β=0.73, and analytic thinking, β=0.63. The content intervention increased intellectual humility, β=0.21, and reduced the third-person effect, β=-0.58. Discussion

Our findings indicate that our short curriculum may be an effective step towards building a holistic “misinformation toolbox” for students, letting them identify information credibility better, make better sharing decisions, and be more cautious about information uncertainty. This multi-skill approach moves beyond isolated interventions to provide a comprehensive framework. Given its scalability and effectiveness, this curriculum holds promise for large-scale integration into formal education.

In our presentation, we will present the findings of our main study, including a delayed posttest and a refresher intervention to investigate the longevity or our curriculum and investigate potential synergies of lateral reading and inoculation. We will discuss the implications of these findings for designing interventions that can build lasting critical competencies to foster critical information skills.