Tackling domain loss: a critical step

Domain loss occurs when a language loses its function in specialized contexts (e.g., research) to a dominant language (often English). Terms to describe new concepts are not created in the local language, which may result in the local language becoming restricted to informal or private settings, reducing its functional scope. Not only does this lack of specialized terminology hinder researchers who want to publish or present in their own language, it also shuts out non-English speakers who want to access that information. To counter domain loss, terminology planning is essential. Terminology planning ensures that timely translation equivalents are available to prevent a default strategy such as borrowing English terms. Collaborations between linguists and researchers in different domains are beginning to bear fruit, such as the TZOS project at the University of the Basque Country (Spain), the Termportalen project at the University of Bergen (Norway), and the LexiConcordia project at Concordia University (Quebec). These projects aim to ensure that specialized terminology is readily accessible to researchers working in other languages. Once the terminology is available, it can be used by researchers directly, as well as being incorporated into resources for training generative AI or machine translation tools so that they can process specialized content more accurately.

Incorporating AI translation into a field’s peer-reviewed publication process

Some fields are finding ways to overcome language bias in their written research processes. For instance, while journals that publish in multiple languages are not new, it is rarer to find multilingual review processes. Members of the Association for Information Systems Taskforce on AI Translation for Inclusive, Impactful Science have demonstrated how AI translation tools can be used to support a multilingual review in five languages in the Australasian Journal of Information Systems (AJIS). Their experience confirms that applying AI-supported multilingualism in the review and publication process is complex for many reasons, including:

• Scoping the problem (e.g., Should we translate only accepted or also submitted papers? What languages should we support?)
• Practicalities of the publication environment (e.g., How should we deal with Digital Object Identifiers and copyright restrictions?)
• Characteristics of AI (e.g., How do we deal with hallucinations or embedded cultural bias?)
• Nuances in written language (e.g., How can the translation approach address issues such as tone and consistency?)
• Institutions of science reinforcing the status quo (e.g., What if researchers now prefer writing in English, partly because of domain loss?)
• Philosophical questions (e.g., conception, articulation and interpretation problems).

Although the taskforce uncovered numerous complexities, they viewed them as opportunities to contribute. The taskforce is also working with major conferences in their field (e.g., the Americas Conference on Information Systems, and the Pacific Asia Conference on Information Systems) to demonstrate the feasibility of providing AI-translated papers and summaries in conference proceedings in different languages. The goal is to inspire a field-level change in the written record.

Incorporating AI translation into a field’s conference activities   

Beyond the written record, science is a social and verbal process that is highly dependent on participation in activities such as conferences, where English may again be a lingua franca, particularly for international events. As illustrated in a recent study, many early-career researchers in environmental sciences who are non-native English speakers regularly decide not to attend or not to present at international conferences owing to language barriers. Organizers of the 2025 International Congress for Conservation Biology (ICCB 2025) wanted to improve the conference experience for speakers of other languages in order to increase linguistic and cultural diversity and to facilitate the exchange of knowledge on biodiversity conservation among culturally and linguistically diverse participants. To this end, they undertook several initiatives, including the use of AI and more:

• Surveying past and potential ICCB participants to understand the barriers and types of desired support (63.4% of respondents identified language accessibility measures as being important or very important to their decision to attend or present)
• Translating the conference website in its entirety into a dozen languages
• Prioritizing scientific rigour over language issues
• Providing guidelines for producing linguistically inclusive oral presentations
• Providing live transcriptions (in English) and translations (in over 60 languages) during presentations using the AI tool Wordly
• Implementing a mentorship program to help mentees feel comfortable and confident delivering their presentations in English.

A post-conference survey was also conducted to evaluate the effectiveness of the initiatives with a view to refining the offers for future.

Moving forward

The ways in which a lingua franca model of scholarly communication reinforces information privilege have become increasingly evident, but scholars around the globe are taking steps to counter this issue. AI tools hold promise, but other actions are needed too. We hope this short piece will inspire additional efforts to counter information privilege in scholarly communication.