Generative AI: Empowering or Excluding Researchers with specific learning difficulties? (Preprint)

UNSTRUCTURED Artificial Intelligence (AI) is rapidly re-shaping how scientific research is conducted, from idea generation to producing written academic outputs. Within this, an additional advantage of AI is its potential to create a more accessible and equitable research environment, particularly for individuals with ‘Specific Learning Difficulties’ (SpLD), such as dyslexia, dysgraphia, dyspraxia, and dyscalculia. Generative AI has the potential to reduce cognitive burden, make information more assessable, improve efficiency and help improve their writing. However, generative AI has the potential to amplify existing biases in academia. Automated triage, AI assisted peer review, and AI detection systems can disproportionately penalise researchers with SpLD. This is due to the biases embedded within training data, which represents neurotypical writing and communication styles. This can lead to writing styles that differ being judged as lower quality, and automated systems used to review manuscripts, grants, or fellowships have the potential to disadvantage researchers with SpLD. Individuals. Alternatively, individuals who use assistive software or have neurodivergent writing styles may get flagged by AI detection models as incorrectly using generative AI, which may lead to questions around their integrity, misconduct and impact their reputation. We concluded while Generative AI has the potential to reduce barriers for researchers with SpLD, it is not guaranteed. Whether Generative AI empowers or excludes researchers who are neurodivergent will depend on how it is governed and used within the scientific community.