Towards responsible use of AI tools
Bron Eager · 2024
This chapter examines key risks and limitations of AI tools for academic research, focusing on hallucinations, biases, and data privacy concerns. Scenarios illustrate how AI-generated inaccuracies, fabrications, and prejudices can undermine scholarship and damage reputations. Mitigation strategies include thorough fact-checking, critical review, transparent AI use, and continuous learning. This chapter also explores how biased training data, underrepresentation, labelling bias, and feedback loops can introduce unfair prejudices into AI outputs, and offers reflective prompts to help identify and attempt to counteract biases. Data privacy risks, such as unauthorised access, identity theft, reputational harm, and legal violations, are discussed in the context of sharing sensitive research data with AI tools, and lastly, best practice suggestions are offered for data handling as well as considerations for choosing an AI provider. Overall, this chapter encourages responsible, well-informed AI use in academia.