One in four children in Germany fails to reach minimum standards in reading at the end of primary school (IGLU 2021). These children cannot read well enough for knowledge acquisition – a dramatic limitation given that education crucially depends on this skill. There also is substantial heterogeneity between children, posing a big challenge for teachers. Social status has a strong impact, so the issue is also closely tied to educational equity. While interventions have been designed to address these challenges, realizing them puts high demands on educators, and scaling up is infeasible due to the personnel resources this would require.

The AI-LIT project will develop a fully scalable AI-supported intervention based on the dialogic reading intervention, which was created to systematically promote language interactions in a picture book reading setting (Whitehurst et al., 1988) captured by the acronym PEER: A Prompt stimulates speech production by the child (e.g., Was haben die Kinder gemacht?), Evaluate analyzes the production (e.g., Ball putt), Expand extends it, often with some reformulation (e.g., Ja! Die Kinder haben den Ball kaputt gemacht.), and Repeat continues the dialog (e.g., Wie ist denn das passiert?). This supports a natural dialog in which a child can also acquire structures and vocabulary in their zone of proximal development (Vygotsky, 1986). Studies and meta-analyses (Flack et al., 2018; Mol et al., 2009; Swanson et al., 2011) show the effectiveness of dialogic reading on receptive and productive vocabulary as well as on educational language and narrative skill development, and it has been successfully evaluated with preschoolers in Germany (Ennemoser et al., 2013; Ennemoser & Hartung, 2017; Besca et al., 2024).