ducation is of central relevance for social and system integration in multilingual European migration societies such as Austria and Germany (Hadjar & Becker 2019). In Austria and Germany, educational inequality can be observed for immigrant students at all stages of education, in correlation with disadvantaged socio-economic status and multilingualism (Dobutowitsch 2020, Döll & Knappik 2015, Ebert & Heublein 2017, Hinz & Thielemann 2013, OECD 2023, Unger et al. 2019).

It seems logical to pick up on recent developments in the field of artificial intelligence (AI) and discuss the potential of assistive AI technology such as natural language processing (NLP) tools for reducing language-based discrimination. AI-based NLP tools have already found their way into educational institutions worldwide: They are used for assessment and evaluation (e.g. feedback), management of learning processes (e.g. learning analytics), as assistants (e.g. for making contact), in the form of intelligent tutor systems and for the design of quasi-authentic meaning-focused tasks (Crompton & Burke 2023), and students use NLP tools to search for articles, translate, structure and edit texts (Garrel & Mayer 2023). It is widely recognized that AI has the potential to increase educational equality, but also carries the risk of making equal participation more difficult (GI 2023). At present, discrimination through AI is mainly discussed in terms of disadvantages due to various forms of algorithmic bias (Baker & Hawn 2021). From a power-critical anti-racist perspective, the question arises as to what extent the institutional regulation of access to assistive AI-supported NLP tools (re)produces inclusion and exclusion in education: Who is allowed to use AI-assisted NLP tools and in which situations? How are restrictions argued?

In recent years, in official German-speaking countries the term linguicism became established to describe language-related discrimination in the context of migration and multilingualism (Skutnabb-Kangas 2015). The term describes "ideologies, structures and practices which are used to legitimate, effectuate, regulate and reproduce an unequal division of power and resources (both material and immaterial) between groups which are defined on the basis of language" (Skutnabb-Kangas 1988: p. 13). Linguicism is therefore more likely to be understood as structural discrimination, which can have effects on the macro, meso and micro levels of education systems. According to Skutnabb-Kangas (2015), if the education policy of a multilingual migration society prioritizes a monolingual education system, this is linguicism at the macro level. At the meso and micro level, linguicism can occur in various forms of direct and indirect institutional discrimination (Dovidio et al. 2010, Gomolla 2023), e.g. by banning specific languages on campus or when lecturers also take linguistic aspects such as accents, sociolects or the fact of a multilingual biography into account when assessing academic performance (Döll & Knappik 2015, Dobutowitsch 2020).

Following the understanding of linguicism as a social structure, it has to be assumed that students will be allowed to use AI-supported generative NLP technology to improve the production and reception of texts to varying degrees depending on their and their family’s migration and language biography. For multilingual students from immigrant families, the strongest restrictions tend to be expected, especially in nation-state contexts such as Germany and Austria, which are characterized by neo-assimilationism (Nieke 2006, Döll 2019).

At a time when universities around the world are discussing how to deal with AI, we will use the example of two universities from Austria and Germany to examine the extent to which linguicist tendencies are emerging in the discourses on AI-supported generative NLP technology at the meso and micro level of higher education.

In order to reconstruct the processes of (re)production of linguicism in connection with AI-based generative NLP technology, in an exploratory and open-ended qualitative research project based on grounded theory (Charmaz 2006) data has been continuously collected on an occasional basis since spring 2023. The open multi-method approach makes it possible to capture the dynamic discourses and developments on the topic. So far, participant observations have been carried out in five courses for lecturers at the two universities with a focus on the thematization of language-related discrimination. The field notes taken were first analyzed in terms of content and then specific situations were examined using key incident analysis, which reveals practices of a social group without applying a complete ethnography (Erickson 1986). In addition, the policy papers and information on AI-based generative NLP technology in university teaching for university lecturers and students are analyzed using critical discourse analysis (CDA, Wodak & Meyer 2016). In order to be able to describe the lecturers' ways of approaching AI-based generative NLP tools, including the implementation of the universities' guidelines, in their courses and examinations with descriptive statistics, a quantitative survey by means of an online questionnaire for students is prepared for spring term 2024. If beneficial to our research project, in-depth interviews or group discussions will be conducted in the autumn term to clarify the statistical results.

At the moment, we assume that we will be able to present the results of the CDA of the guidelines and the key incident analysis as well as the initial results of the quantitative survey. In line with the mechanisms of structural discrimination in democratic states, we assume that there won’t be linguicist inequality between monolingual and multilingual or native and immigrant students in connection with AI-based generative NLP tools in the meso-level guidelines, as this would contradict the democratic principle of equal treatment. However, the interim results of the analyses of the field notes from the participant observations indicate a limited awareness of the potential for discrimination of AI-based generative NLP tools among both university lecturers and further education lecturers, so that we assume that linguicist speech and actions are experienced at the micro level, i.e. in the interaction between students and lecturers. Due to the similar migration histories and migration discourses in Austria and Germany, we do not expect any national differences at present, but this assumption still needs to be checked with the data. In any case, our work, which is located at the intersection of educational science, linguistics and the sociology of technology, offers initial findings on the question of whether linguicist routines are becoming established in higher education institutions in connection with AI-based generative NLP tools and raises new research questions in this field.

Baker, R. S., & Hawn, A. (2021). Algorithmic Bias in Education. International Journal of Artificial Intelligence in Education 32, 1052-1092. Charmaz, K. (2006). Constructing grounded theory. London. Crompton, H., & Burke, D. (2023). Artificial intelligence in higher education: the state of the field. Int J Educ Technol High Educ 20, 22. Dobutowitsch, F. (2020). Lebensweltliche Mehrsprachigkeit an der Hochschule. Münster. Döll, M., & Knappik, M. (2015) Institutional mechanisms of inclusion and exclusion in Austrian pre-service teacher education. Tertium comparationis 21 (2015) 2, 185-204. Döll, M. (2019). Sprachassimilativer Habitus in Bildungsforschung, Bildungspolitik und Bildungspraxis. ÖDaF, 1+2/2019, 191-206. Dovidio, J. F. et al. (2010). Prejudice, stereotyping and discrimination: Theoretical and empirical overview. In J. F. Dovidio et al. (Ed.), The Sage handbook of prejudce, streotyping and discrimination (pp 3–28). Los Angeles. Ebert, J., & Heublein, U. (2017). Studienabbruch bei Studierenden mit Migrationshintergrund. Hannover. Erickson, F. (1986). Qualitative methods in research on teaching. In M. C. Wittrock (Ed.), Handbook of research on teaching (pp 119—161). New York. Garrel, J., & Mayer, J. (2023). Artificial Intelligence in studies—use of ChatGPT and AI-based tools among students in Germany. Humanities and Social Sciences Communications, 10(1), 1-9. Gesellschaft für Informatik (GI) (2023). Künstliche Intelligenz in der Bildung. Positionspapier. https://gi.de/fileadmin/GI/Hauptseite/Service/Publikationen/GI_Positionspapier_KI_in_der_Bildung_2023-07-12.pdf Gomolla, M. (2023). Direkte und indirekte, institutionelle und strukturelle Diskriminierung. In A. Scherr et al. (Ed.), Handbuch Diskriminierung (2nd edn, pp 171-194). Wiesbaden. Hinz, T., & Thielemann, T. (2013). Studieren mit Migrationshintergrund an einer deutschen Universität. Soziale Welt, 64(4), 381–399. OECD (2023). PISA 2022 Results (Volume I): The State of Learning and Equity in Education, PISA. Paris. Nieke, W. (2006). Anerkennung von Diversität als Alternative zwischen Multikulturalismus und Neo-Assimilationismus? In H.-U. Otto & M. Schrödter (Ed.), Soziale Arbeit in der Migrationsgesellschaft (pp 40-48). Lahnstein. Skutnabb-Kangas, T. (1988). Multilingualism and the Education of Minority Children. In T. Skutnabb-Kangas & J. Cummins (Ed.), Minority education: from shame to struggle (pp 7-44). Clevedon. Skutnabb-Kangas, T. (2015). Linguicism. The Encyclopedia of Applied Linguistics. Malden, MA. Unger, M. et al. (2019). Studierenden-Sozialerhebung 2019. Wien. Wodak, R., & Meyer, M. (Ed.) (2016). Methods of critical discourse studies (3rd edn). London.