AI-driven systematic reviews under scrutiny: Is AI-driven literature screening an alternative to human screening? A synopsis of three systematic reviews on the use of ICT in inclusive school settings

In the age of digitization, societal participation is increasingly tied to media competence, making the acquisition of media competence a crucial educational objective - especially within inclusive and special educational needs school settings. As schools – among other stakeholders in education - are responsible for fostering media literacy, educators face the question of which didactic digital approaches help to support inclusion. Empirical research is needed to further leverage the potential benefits/risks of digital media in inclusive/SEN settings. However, empirical research must be preceded by a comprehensive understanding of existing studies on the intersection of digital media and inclusive/SEN education. This necessitates systematic literature reviews, which are traditionally conducted through manual screening. Given the time-intensive nature of this task, AI-assisted screening tools have gained attention as a possible alternative. Yet, AI-assisted screening tools require careful evaluation regarding accuracy and bias. By comparing manual and AI-driven screening processes, this study highlights the need for a balanced approach that integrates technological advancements with human expertise. Therefore, this paper pursues two primary objectives. First, it aggregates the findings of three systematic literature reviews examining the role of digital media in supporting students with special educational needs in primary and secondary education. These reviews included peer-reviewed and non-peer-reviewed empirical studies published in German and English over the past ten years. Criteria focused explicitly on digital media use in inclusive or special education contexts, excluding tertiary education. Second, the paper explores the methodological limitations of aggregating the results of three different reviews and it focusses on implications of using AI for literature screening by comparing the outcomes of manual versus AI-assisted approaches. While AI tools offer efficiency, serious concerns remain - regarding the accuracy of machine decisions and the potential introduction of new biases. The replication of one of the above-mentioned three reviews using AI allows for a direct comparison, highlighting the advantages and limitations of automated screening. This contribution aims to inform both the research community and educators by presenting current findings on digital media and inclusion while also critically reflecting on the evolving role of AI in academic research processes.